AAAS

BME Frontiers
Volume 2022, Article ID 9857485, 21 pages
https://doi.org/10.34133/2022/9857485

Review Article
Impedance Imaging of Cells and Tissues: Design and Applications

Raziyeh Bounik , Fernando Cardes , Hasan Ulusan , Mario M. Modena ,

and Andreas Hierlemann
ETH Zürich, Department of Biosystems Science and Engineering, Basel, Switzerland

Correspondence should be addressed to Raziyeh Bounik; [email protected]

Received 2 April 2021; Accepted 28 March 2022; Published 9 June 2022

Downloaded from https://spj.science.org on August 31, 2023

Copyright © 2022 Raziyeh Bounik et al. Exclusive Licensee Suzhou Institute of Biomedical Engineering and Technology, CAS.
Distributed under a Creative Commons Attribution License (CC BY 4.0).

Due to their label-free and noninvasive nature, impedance measurements have attracted increasing interest in biological research.
Advances in microfabrication and integrated-circuit technology have opened a route to using large-scale microelectrode arrays for
real-time, high-spatiotemporal-resolution impedance measurements of biological samples. In this review, we discuss different
methods and applications of measuring impedance for cell and tissue analysis with a focus on impedance imaging with
microelectrode arrays in in vitro applications. We first introduce how electrode configurations and the frequency range of the
impedance analysis determine the information that can be extracted. We then delve into relevant circuit topologies that can be
used to implement impedance measurements and their characteristic features, such as resolution and data-acquisition time.
Afterwards, we detail design considerations for the implementation of new impedance-imaging devices. We conclude by
discussing future fields of application of impedance imaging in biomedical research, in particular applications where optical
imaging is not possible, such as monitoring of ex vivo tissue slices or microelectrode-based brain implants.

1. Introduction trode arrays (MEAs) [9] and their use for impedance analy-
sis enabled high-sensitivity and spatially highly resolved
Recent progress in in vitro cellular and molecular analysis measurements at subcellular resolution [10–13]. The label-
has greatly advanced our understanding of human physiol- free and high-resolution nature of impedance detection
ogy in healthy and diseased states. Owing to their label- offered by high-density MEAs has enabled to produce 2D
free and noninvasive nature, impedance-based detection impedance images of biological samples to, e.g., monitor
methods have been used for quantitative long-term charac- how cell attachment and mobility of adherent cells are
terization of live cells and tissues. In comparison to other affected by drug treatment [11] or to recognize the spatial
routine analysis methods, such as optical imaging, imped- organization of cells in live, ex vivo brain slices [12].
ance measurements typically require less costly equipment, In this review article, we first present different examples of
the respective setups can be miniaturized, and experiments how impedance measurements and impedance imaging is
can be parallelized [1, 2]. applied in neuroscience and biomedical research. We discuss
First reports on the use of impedance-based detection different methods that have been introduced for the character-
methods in biology date back to the 1920s, when impedance ization of cell and tissue models in vitro, with a particular focus
measurements were used to characterize membrane capaci- on methods and approaches for two- and three-dimensional
tance and resistance of blood cells [3, 4]. Technical improve- (2D and 3D) impedance-based imaging, and we present
ments and the establishment of theoretical models have then examples of impedance imaging in biomedical research. We
enabled the use of impedance methods for the analysis of cell then discuss the different detection methods that are used
volumes in suspension [5], for discerning different cell pop- for impedance analysis as well as related circuitry implementa-
ulations [5], and for monitoring cellular adhesion and tions and discuss the advantages and limitations of the differ-
growth [6–8]. The development of high-density microelec- ent approaches. Afterwards, we summarize the design options
2 BME Frontiers

that need to be considered during the conceptualization of mittivity, which is associated with its ability to store energy.
impedance-based MEA sensors for different applications. As the energy of an applied electric field can either be dissi-
Finally, we end the review with an outlook on trends in devel- pated or stored, the impedance of a dielectric material is
oping MEAs for impedance-based imaging of biological directly related to its permittivity and can be expressed as
entities. [17]

2. Impedance Measurements: Methods σ − jωϵ 0 ϵ

and Applications z= , ð2Þ
σ2+ ðωϵ 0 ϵ Þ2
2.1. Impedance Measurements and Impedance Model of Cells
and Tissues. Impedance extends the concept of Ohm’s resis- where σ is the characteristic conductivity at DC, ω is the
tance to alternating-current (AC) circuits and features both, radial frequency of the applied electric field, ϵ 0 is the permit-
magnitude and phase, unlike Ohmic resistance, which has tivity of vacuum, and ϵ is the relative permittivity of the
only magnitude. Impedance is a complex number with the respective dielectric material. Typically, the relative permit-
same unit as resistance, for which the SI unit is Ohm (Ω), tivity of a material is frequency-dependent and tends to fall
its symbol is usually Z. In analogy to Ohms law for directed as the frequency of the applied field increases [18]. For cells
current (DC), the impedance (Z) is the ratio of the electrical and tissues in solution, the relative permittivity extends over
potential difference (V), applied to a conductor, and the four principal dispersion regions, i.e., frequency ranges in
resulting current (I) through it (Equation (1)). Impedance which the relative permittivity features large variations
can, therefore, be expressed as the combination of a resis- (Figure 1(b)): the α-dispersion, in the ~100 Hz-10 kHz
tance (R, the real component) and a reactance (X, the imag-

Downloaded from https://spj.science.org on August 31, 2023

region, which is associated with ion diffusion across the cell
inary component): surface; the β-dispersion, in the sub-MHz–10 MHz range,
which is related to charge accumulation at the cellular mem-
jV je j θ V brane; the δ-dispersion in the sub-GHz regime, in which
Z= = jZ je jθZ = R + jX, ð1Þ
jI je j θ I rotation of macromolecular side-chains takes place; the γ-
dispersion in the ~10 GHz regime, where dipolar rotation
of water molecules is dominant [18, 19]. Therefore, mea-
where j∙j and θð∙Þ represent the modulus and phase of the
surements of cell and tissue impedance at different frequen-
respective complex number, and j is the imaginary unit. cies can be used to investigate different phenomena. At low
Impedance sensing is based on measuring the absolute frequencies, in the α-dispersion frequency range, the cell
value or relative change in the impedance of a cell or tissue impedance is extremely high so that impedance measure-
to extract information on sample properties. Impedance ments provide information on the cell or tissue volume,
measurements can be carried out at a specific frequency or which obviates current conduction. Therefore, measure-
over a broad frequency range, the latter being termed electri- ments in this frequency range can be used to estimate cell
cal impedance spectroscopy (EIS). Figure 1(a) shows a sim- or tissue volume or size. Upon increasing the frequency,
plified electrical equivalent-circuit model of a single cell information on cell and tissue membranes can be extracted,
between a pair of electrodes. The overall impedance of the as the cell membrane polarization is in the β-dispersion
system includes three main components: (i) the cell imped- range. Higher frequencies, in the sub-GHz and GHz regime,
ance, which consists of membrane contributions (Cm and can provide information on the water content or protein
Rm in parallel) and cytoplasm contributions (Rc and Cc in concentrations within cells [20, 21]. However, these fre-
parallel); the cell equivalent circuit is known as the single- quency ranges are rarely used for investigating cells and tis-
shell model [14]; as the resistive impedance component of sues, as measurements at such high frequencies require
the cell membrane and the capacitive impedance component careful electronic design, and as cell/tissue impedance con-
of the cytoplasm are usually orders of magnitudes larger tributions cannot be easily separated from, e.g., medium
than the capacitive component of the cell membrane and contributions, due to the dominant effect of dielectric relax-
the resistive component of the cytoplasm, the contributions ation of water in the GHz regime [22].
of these electrical equivalent-circuit components are typi-
cally neglected in the electrical equivalent-circuit models 2.2. Impedance Measurement Methods. Impedance measure-
[15]; (ii) the impedance of the electrode-electrolyte inter- ments are typically carried out by applying an electrical
faces, Z el , which consists of the double layer capacitances stimulus to the sample of interest and by measuring the sam-
(main contribution for polarizable electrodes, such as plati- ple response as a function of the frequency of the applied
num (Pt) or gold (Au) electrodes) in parallel to the charge- signal. Two different implementations are possible: a known
transfer resistances (main contribution for nonpolarizable test voltage is applied to the sample, and the resulting cur-
electrodes, such as silver/silver chloride (Ag/AgCl) elec- rent is measured, or, alternatively, a test current is injected,
trodes) [16]; (iii) the solution resistance Rsol [16]. and the resulting voltage drop across the sample is detected.
When an electric field is applied to a dielectric material, The two approaches are conceptually equivalent from a the-
such as a cell membrane, the field causes a polarization of oretical point of view and are based on the generalized
the material. The polarization response of a dielectric mate- Ohm’s law (Equation (1)). However, the respective imple-
rial to an externally applied field is characterized by its per- mentations require different electronic circuitries and may
BME Frontiers 3

Electrode
Zel
108
Cell 𝛼

Permitivity
106
Cc Rc

104 𝛽
Rsol

102 𝛿
Cm Rm
𝛾

1
Zel 1 102 104 106 108 1010
Electrode Frequency (Hz)

(a) (b)
Figure 1: (a) Electrical equivalent-circuit model of a cell in suspension between a pair of electrodes. The overall measured impedance
includes three contributions: the cell impedance, which consists of the cytoplasm and membrane capacitive and resistive contributions
(C c , Rc and C m , Rm ), the electrode impedances (Z el ), and the medium resistance (Rsol ). (b) Idealized spectrum of the relative permittivity
of cells and tissues, showing four main dispersion regions [18, 19]; for details, see text.

Downloaded from https://spj.science.org on August 31, 2023

feature different susceptibility to unwanted effects during brane integrity in reference to cell size by determining the
sample characterization. ratio of impedances measured at high (cell membrane char-
The power and frequency of the stimulation signal are acteristics) and low (cell size) frequencies (Figure 2(b)), the
chosen depending on the composition of the sample, the so-called electrical opacity [25, 26], and to differentiate dif-
measurement features of interest, and the limitations of the ferent cell populations in mixed samples [27]. Simultaneous
instrumentation. Signal power must be high enough to pro- multifrequency monitoring requires either hardware that
duce a response that can be detected by the readout electron- can independently generate and process signals at each fre-
ics, however, without saturating the amplifiers. Moreover, quency [28] or sequentially measure the sample responses
large stimulation currents can damage a sample, as it is at different frequencies, however, at the cost of increased
intentionally done in electronic wound-healing assays [6], acquisition time [29].
while high voltages can cause undesired electrochemical In EIS, impedance characteristics are measured at multi-
reactions in an aqueous phase, such as electrolysis. Stimula- ple frequencies across several orders of magnitude [30].
tions can be carried out by using sinusoidal waves, the signal These measurements can be used to fit the electrical equiva-
power of which is concentrated at one specific frequency, or lent circuit of the sample [30, 31] or to use a dedicated math-
the stimulus can consist of a multifrequency signal, where ematical model to extract specific features of interest [32].
the signal power is spread across a broader frequency EIS measurements can be performed by applying a sinusoi-
spectrum. dal stimulus, whose frequency is sequentially altered. While
Finally, impedance detection can also be carried out in this technique is unparalleled in terms of accuracy, the mea-
an indirect way, for example, through impedance-to- surement acquisition time can amount to up to several
frequency conversion. This conversion method and its minutes, if there is a large number of frequency steps, espe-
implementation in integrated systems will be discussed in cially in the low-frequency range. Faster measurements can
more detail in Section 3.5. be obtained by using more complex stimuli and by distribut-
ing the signal power across a wide spectral range, e.g., by
2.2.1. Stimulation Frequency Selection. Single-frequency applying a chirp stimulus [33], rectangular pulses [34], or
measurements are commonly used when the target feature white noise [35]. These stimulation methods are sometimes
is known to be observable at a specific frequency referred to as “time-domain” techniques, as the readout cir-
(Figure 2(a)). For example, cellular attachment to an elec- cuit directly records the signal over time, and frequency
trode surface can be detected at 1 kHz [23], cellular micro- analysis is carried out during postprocessing, typically by
motions across electrodes can be observed using a 4 kHz calculating the Fourier transform of the time-domain signal
sinusoidal signal [7], or parasite motility in vitro can be [36]. However, spreading the power of the stimulus over a
detected with a 500 kHz signal [24]. Single-frequency moni- wide frequency band may result in low power spectral den-
toring offers significant advantages with respect to hardware, sities, which may reduce the signal-to-noise ratio and ren-
since all information is contained in a narrow-bandwidth ders these techniques less accurate than their narrow-band
signal that can be measured and separated from out-of- counterparts.
band noise using, for example, lock-in amplifiers.
Combining multiple measurements at different frequen- 2.2.2. Electrode Configurations. In its simplest form, imped-
cies allows for investigation of more complex features. This ance characterization can be carried out by using a two-
approach is often used to acquire information on cell mem- electrode configuration (Figure 3(a)). The electrode
4 BME Frontiers

|Z|
|Z|

f
fLF fHF
f Cell type 1
Cell type 2
𝜃Z Cell type 3

Electrical opacity
(|ZHF|/|ZLF|)
f
fm Electrical volume |ZLF|
Cell type 1
Cell type 2
Cell type 3

(a) (b)

Downloaded from https://spj.science.org on August 31, 2023

Figure 2: Examples of impedance measurements at different frequencies. (a) Illustration of the impedance of two samples, the capacitive
contribution of which is dominant at low frequencies while the resistive component is dominating at high frequencies. A single
measurement at f m allows for detection of resistance changes of the sample. Note that phase variations and magnitude variations
manifest at different frequencies. (b) Example of cell classification using two-frequency impedance measurements. Low frequency (f LF )
impedance provides information on the cell size, while the ratio of high-frequency (f HF ) and low-frequency impedance yields the so-
called electrical opacity.

positions influence the electric-field distribution and how reflects the voltage drop across the sample, while the influ-
electrical currents flow across the sample. For coplanar elec- ence of electrode impedance is eliminated. The location of
trodes, as shown in Figure 3(a), the electric field decreases the voltage-sensing electrodes needs to be carefully chosen,
with increasing distance to the electrode plane, and most in particular for large samples, as the measured voltage drop
of the current tend to flow along the surface of the electrode may otherwise not represent the average voltage drop across
plane. Therefore, the measured impedance is significantly the sample.
altered when a sample is placed on the plane in between Electric impedance tomography (EIT) is an impedance-
the electrodes. In contrast, if the same sample is located far based imaging technique, which relies on measurements of
from the electrode plane, it only marginally affects the sample impedance at different positions to investigate the
impedance measured between the electrodes. This internal properties of a sample [38, 39] (Figure 3(c)). A
position-dependence can be avoided by using facing elec- reconstruction algorithm, frequently based on a finite-
trodes, i.e., electrodes placed on opposing planes, which fea- element model, is used to infer the internal impedance dis-
ture uniform electric-field distribution, as in parallel plate tribution that corresponds to the experimentally determined
capacitors. However, the realization of facing electrodes is measurement values. EIT reconstruction is often an ill-posed
more difficult due to a more complex fabrication, and the problem, and algorithms have limitations concerning the
required alignment of the two electrodes. complexity of the reconstructed image. Reconstruction typi-
Regardless of the electrode configuration, the electrode- cally requires a priori information about the expected prop-
electrolyte-interface impedances add in series to the sample erties of the sample [40].
impedance, which may obstruct sample characterization at Electrode arrays allow for both, multiple two-terminal
specific frequencies, where electrode contributions domi- and multielectrode impedance measurements (Figure 3(d)).
nate. To circumvent this limitation, a multiple-electrode Two-terminal measurements can be taken between two elec-
arrangement can be used to reduce or even eliminate the trodes in the array [41, 42], or several electrodes can be
effects of the electrode impedance [29, 37]. In four- grouped to obtain larger pixels to enable multielectrode or
electrode measurements, the two outer electrodes (one per differential measurements [43]. Alternatively, one of the ter-
terminal) provide the stimulation current, while two inner minals can be used as a common reference electrode to per-
electrodes sense the voltage drop across the sample form parallel measurements [10, 44].
(Figure 3(b)). Voltage measurements require that the sens-
ing electrodes do not carry any current so that no voltage 2.2.3. Absolute and Relative Measurements. Absolute imped-
drop occurs at the electrodes. The measured signal at the ance measurements enable a quantitative characterization of
inner electrodes in an arrangement shown in Figure 3(b) the sample features of interest. However, absolute
BME Frontiers 5

I + V – I I I
V1 V2 V3 V4 V5 V6

(a) (c)

+ V – + V1 – + V2 – + V3 –
I I I1 I1 I2 I2 I3 I3

(b) (d)
Figure 3: Illustration of impedance measurement methods using a variable number of coplanar electrodes. (a) Two-electrode setup, in
which a stimulation current (I) is applied through the same electrodes that are used for voltage measurement (V). Note that voltage
stimulation and current measurements are also possible. (b) Four-electrode setup, using one outer pair of electrodes for current
stimulation and a second inner pair of electrodes for voltage measurements. (c) EIT uses two (outermost) electrodes of a microelectrode
array for current stimulation, while the other electrodes are used to simultaneously measure the voltage at different positions of the
sample. (d) Microelectrode array configured for several two-electrode measurements in parallel.

Downloaded from https://spj.science.org on August 31, 2023

measurements are strongly dependent on the ability to sep- of neuronal cells in vitro [47, 48]. The topological complex-
arate the sample impedance contribution from other effects ity of cells and tissue models and the long acquisition times
that may contribute, such as electrode impedance, variation of SEIM have, so far, prevented a broader application of this
in medium conductivity due to evaporation, and contribu- technique for cell and tissue characterization. However, sim-
tions of parasitic circuitry elements. To reduce these ple two-electrode setups have been widely used to character-
unwanted effects on the measurement, differential measure- ize a large variety of biological samples, ranging from single
ments can be carried out by comparing the sample measure- cells to tissue models and multicellular organisms, without
ments to reference measurements under the same acquiring spatial information. We here briefly discuss these
environmental conditions and acquired with the same read- systems, as they demonstrate the large variety of information
out circuitry. Differential impedance values can either be that can be gained through impedance characterization of
obtained by determining impedance differences between biological application.
independent recordings during digital postprocessing [45], The seminal work of Giaever and Keese in the 1980s
for example, by comparing sample impedance recordings and 1990s demonstrated the possibility of monitoring the
to impedance recordings at a reference time point, or by adherence and micromotion of adherent cells by using a
using a differential acquisition scheme including a sample pair of coplanar gold electrodes patterned at the bottom
and reference [44, 47]. of a cell culture well [7, 49]. The electrode arrangement
Calibration is also key to improve the accuracy of imped- consisted of a “small” electrode (~10-2 mm2) and a large
ance measurements. During calibration, the impedance counter electrode (~102 mm2), so that the overall imped-
readout circuit is tested with different known samples across ance would be dominated by the electrolytic interface
the frequency spectrum of interest to verify the response of between the small electrode and the medium solution.
the system and to enable signal corrections and normaliza- The presence of cells on top of the small electrode then
tions during data analysis [29]. altered the electrode-medium interface, which could be
readily detected by monitoring impedance changes [50].
2.3. Applications of Impedance Measurements. Owing to its The technique, which was later termed Electrical Cell-
label-free and noninvasive nature and its potential for scal- substrate Impedance Sensing (ECIS®), was further devel-
ability, integration, and automation, impedance spectros- oped, and a variety of electrode arrangements have been
copy has attracted considerable interest for investigating devised to monitor different parameters of interest, such
biological and biomedical samples. Here, we will focus on as cell confluency or stem-cell differentiation using
the application of impedance spectroscopy and imaging for spectroscopy-based ECIS [51].
in vitro studies of cell and tissue models. Impedance analysis has also been widely used for the
analysis of single cells or small tissue models by using elec-
2.3.1. Two-Electrode Setups. To obtain a two-dimensional trodes, the size of which was comparable to that of the sam-
impedance representation or “impedance image” of a sample ple of interest. The liquid containing the samples was flown
of interest with a two-electrode setup, the electrode pair over or between the electrodes (coplanar or facing elec-
needs to be moved across the sample surface. This approach, trodes), a technique now known as impedance cytometry
which is called scanning electrical impedance microscopy [52, 53], or confined in the sensing area, e.g., by hydrody-
(SEIM), was used, e.g., for a morphological characterization namic trapping [54] or physical barriers [24, 55–57].
6 BME Frontiers

To increase the signal-to-noise ratio, impedance cytom- MEAs for real-time imaging of adherent cells is the work of
etry is often carried out using a differential detection scheme Laborde et al., where the authors presented a CMOS high-
by employing, e.g., a three-electrode setup [53]. Differential density microelectrode array (HD-MEA) with 65,536 elec-
measurements enable to directly measure the dielectric trodes of 90 nm radius on a 0:6 μm × 0:89 μm grid [11]. To
properties of the sample against the suspension medium increase the detection range to a few hundreds of microns
and to remove any effect caused by drifts in electrode perfor- from the sensor surface, the authors used a 50 MHz detection
mance, medium evaporation, or temperature variations. frequency to overcome the screening effect of the electrical
Owing to the relatively high throughput of impedance double layer (EDL) at the electrode interface (Figure 4(a)).
cytometry (~100-1000 cells per minute), this technique can While this approach enables to extend the sensing region of
be used to provide a snapshot of the sample condition at a the electrodes while reducing the impedance of the double-
defined time point and has been used to, e.g., identify cells layer capacitance at the sensing frequency, measuring only
in mixed cell populations [21, 58], recognize differentiated the capacitive contribution makes it impossible to perform
mesenchymal stem cells [59], measure the proportion of impedance spectroscopy for cell characterization and differen-
activated platelets in whole blood [60], and discriminate tiation. Nevertheless, the developed system enabled to follow,
activated T-cells [61]. Conversely, sample confinement in in real-time and at subcellular resolution, the adhesion,
the active area is a prerequisite to continuously monitor spreading, and dynamic attachment of cancer cells, which
dynamic processes on the same sample or specimen. As an were cultured on the sensor surface (Figure 4(b)).
example, hydrodynamic trapping was employed to capture A primary use of MEAs includes localized voltage
single yeast cells to follow yeast growth and budding by recordings, for which the systems have been used to perform
monitoring impedance variations over a large range of fre- long-term measurements of intra- and extracellular electrical
quencies [54]. Impedance characterization has also been potentials of electrogenic cells, in particular of neurons and car-

Downloaded from https://spj.science.org on August 31, 2023

used to quantify the nonalcoholic fatty liver disease progres- diomyocytes in vitro [68–70]. With the realization of multifunc-
sion in 3D liver-microtissue models or to monitor the effi- tional electrode arrays, i.e., arrays of electrodes that can be
cacy of candidate drug compounds in vitro, e.g., by utilized with multiple detection schemes (e.g., voltage, imped-
recording the growth of cancer microtissues [57, 62] or the ance, or electrochemical measurements), impedance-based
motility of human parasite larvae [24]. imaging has been integrated in several systems to provide com-
plementary information on the sample under investigation.
2.3.2. Multielectrode Configurations. Two-dimensional Chi et al. presented a sensor array featuring 144 indepen-
“impedance imaging” of biological samples can be achieved dently addressable electrodes for extracellular voltage record-
by using a MEA, which features a matrix of independently ing, impedance mapping, and optical detection, either via
addressable microelectrodes [63, 64]. As for the previously shadow imaging or bioluminescence [71]. The authors dem-
mentioned ECIS sensors, cell adherence to the microelec- onstrated the suitability of their system for the investigation
trode causes an increase in electrode impedance, which can of different cell types, namely, human cardiomyocytes, mouse
be recorded by the 2D array electrodes to yield an image of neurons, and ovarian cancer cells. Through impedance imag-
cell dispersion and adherence over the array. The spatial res- ing, the authors could monitor the cell attachment to the sen-
olution that can be attained with microelectrode arrays is sor array, while the integrated optical detection scheme
defined by the electrode dimensions, the electrode pitch, provided information on the location and distribution of cells.
and, in the z-direction, the ionic strength of the solution Furthermore, the authors treated the cardiomyocytes with iso-
and the detection frequency [11]. Furthermore, both, elec- proterenol, a drug against bradycardia, and were able to detect
trode pitch and readout multiplexing capabilities, which
an increase in the beating rate by using the integrated extracel-
define the spatial and temporal resolution, strongly depend
lular voltage recording. Cell attachment to the array was not
on the technological approach used for the development of
affected by drug dosage, as could be detected in parallel by
the MEA: passive MEAs with off-chip readout circuitry fea-
using impedance analysis. However, although the integrated
ture strong limitations in the number of electrodes that can
be used in parallel and typically have additional space optical and impedance-based capabilities provided an image
between the electrodes for routing and leads; active MEAs of the sample fraction that was in direct contact with the elec-
include addressing and signal-conditioning circuitry and trode array, the spatial resolution was limited to ~90 μm by the
are fabricated in complementary-metal-oxide-semiconduc- electrode pitch and by the nonuniform electrode distribution
tor (CMOS) technology; they feature large numbers of elec- in the active sensing area.
trodes at a very small pitch and enable highly parallel Our group presented a multifunctional HD-MEA, which
impedance recordings from many electrodes. More details featured 59,760 microelectrodes with an electrode pitch of
on the advantages and disadvantages of the two MEA cate- 13.5 μm (Figure 5(a)) [12, 44]. The system enabled to record
gories will be discussed in the following sections. impedance spectra with up to 32 electrodes in parallel in a fre-
Impedance-based detection with MEAs has been used quency range between 1 Hz and 1 MHz. The small electrode
to monitor the growth and distribution of biofilms on pitch provided subcellular spatial resolution, for extracellular
electrode arrays [65], to detect the hybridization of DNA voltage recording and impedance measurements. The HD-
strands that have been previously attached to the sensor sur- MEA was used to monitor the differentiation of mouse embry-
face [10, 66, 67], and to measure the dynamic attachment oid bodies (EBs), which were plated on the array. Impedance
and micromotions of cells [11]. A notable example of using imaging enabled to follow the adhesion, spreading, and
BME Frontiers 7

1.0 y 100 nm 10 kHz

x
0.32
(i) (ii)
Potential (a.u.)

0.30

Cexp (fF)
0.5 0.28 (iii)
10 𝜇m (iv)
50 MHz
0.26

0.24
0 10 15
t (min)
0.0

(a) (b)
Figure 4: Impedance-imaging with HD-MEAs for the detection of cellular adhesion and micromotion. (a) Simulated spatial distribution of
the electrical potential in 150 mM salt solution at 10 kHz and 50 MHz stimulation frequencies. The fields are simulated for a 90 nm diameter
Au electrode and the presence of a 4.4 μm polystyrene bead (dashed line) positioned on the array. At low frequencies, the electric field only
interacts with the bead when the bead is located within ~10 nm distance from the electrode, due to the screening effect of the electrode-
electrolyte double layer. By increasing the stimulation frequency to 50 MHz, the electric field extends further into the solution. (b)

Downloaded from https://spj.science.org on August 31, 2023

Capacitance values measured with a single-electrode (right), marked in red, during (i) PBS washing, (ii) introduction of cells in medium,
(iii) cell attachment, and (iv) cell micromotions; adapted with permission from Laborde et al., Nat. Nanotech. 2015 (11).

growth of EBs on the array for five days. After five days of 1 MHz. Impedance recordings at 1 kHz stimulation fre-
spreading and differentiation on chip, extracellular voltage quency were used to evaluate the adhesion and growth of
recordings were used to measure the spontaneous beating of primary hippocampal neurons on the chip (Figure 5(b))
cells that had differentiated into cardiomyocytes [12]. The [13]. The combination of voltage and impedance recordings
HD-MEA system was also used to acquire an impedance provided information on cell distribution to assess whether
image of an acute mouse cerebellar slice and enabled to iden- the absence of electrical activity in specific regions on the
tify four different regions in the slice, namely, (i) white matter, sensor surface corresponded to areas devoid of cells or
(ii) the granular cell layer, (iii) the Purkinje cell layer with large whether the regions were populated by cells without electri-
electrophysiological activity, and (iv) the molecular layer that cal activity. The impedance recording of the whole array at a
contained the dendritic trees of the Purkinje cells single frequency could be acquired in ~2 minutes, while
(Figure 5(a)). Impedance imaging enabled a label-free detec- imaging of the whole sensor surface with a confocal micro-
tion of different cell layers in the cerebellar slice, which could scope required ~30 minutes. The system enabled high spatial
be recognized by their different impedance signatures (i.e., resolution within the 480 × 480 μm2 sensing areas owing to
magnitude and phase). To perform a complete impedance the small electrode pitch of 15 μm. However, the spatial
scan of the whole sensing area, multiple sequential measure- arrangement of the electrode clusters, which featured a
ments were required due to the large number of active elec- 1020 μm intercluster distance, was optimized for multiwell
trodes and the limited number of impedance-detection packaging, which obviated to perform high-resolution elec-
circuitry modules. Although recording at medium/high fre- trical detection over large areas, e.g., for large tissue slices.
quencies could be performed within a few minutes, low- Finally, MEAs can be used to perform localized electro-
frequency measurements required an acquisition time of ~1 chemical measurements, such as amperometry and voltamm-
hour, which is comparable to the time required for confocal etry, to provide real-time two-dimensional electrochemical
imaging. Impedance imaging still offers the advantage to imaging [44, 73–75]. As an example, a HD-MEA with
record in real time from live cell and tissue models, as there 17:5 × 15 μm2 interdigitated electrodes was used to detect
is no risk of potential damages by phototoxic effects during the spatiotemporal characteristics of neurotransmitter release
long-term optical and fluorescence imaging and as impedance (norepinephrine, epinephrine, and dopamine) in acute
imaging is label-free. murine adrenal-tissue slices upon stimulation with caffeine
Lopez et al. recently presented a multifunctional HD- [75]. A microfluidic channel was used to achieve temporally
MEA featuring 16,384 electrodes, which were arranged in defined chemical stimulation of the tissue slices so that the
16 active areas of 1024 electrodes each at an electrode pitch neurotransmitter release could be measured at high spatial
of 15 μm [72]. The system featured current and voltage stim- and temporal resolution. Electrochemical measurements
ulation, intracellular and extracellular voltage recordings, were also used to image the secretion of metabolites in bacte-
and two modes for impedance measurement: fixed fre- rial films of P. aeruginosa [74]. Interestingly, the use of
quency (1 and 10 kHz) measurements for fast impedance potential-sweep methods (square-wave voltammetry), as
monitoring (0.1-1 ms temporal resolution), and impedance opposed to fixed-potential detection, enabled to detect the
spectroscopy in a frequency range between 10 Hz and presence and concentration of multiple phenazine
8 BME Frontiers

Impedance magnitude (MΩ)

2.4 10

Array y (mm)
GCL
WM
GCL
Electrophysiology recording units (1024 ch.) PCL
1.6
5
ML ML
0.8
Impedance (4) Impedance (4) 1 mm
0
Waveform gen.

Impedance (4)
Electrode array Impedance (4) 0 1 2 3 4

Digital ctrl
Impedance (4) Impedance (4)
Array x (mm)
50 𝜇m
1 mm
Impedance (4) Impedance (4)
Electrophysiological activity [Hz] Impedance phase (°)
Ref. el. 2.4 100 GCL
40
PCL

Array y (mm)
1.6
Electrophysiology recording units (1024 ch.) 50 0
0.8 ML

0 –30
0 1 2 3 4 0 1 2 3 4
Array x (mm) Array x (mm)

(a)

Downloaded from https://spj.science.org on August 31, 2023

32 stim. units

1 2 5 6
512 channels 512 channels
32 MUXs 3 4 7 8 32 MUXs
32 ADCs 32 ADCs
digital control 9 10 13 14 digital control Voltage (𝜇V)

11 12 15 16
0 100 200 300 400 500 600
32 stim. units

16 pixels
1 pixel
E1 E2
16 pixels

E3 E4
|Z| variation
15 𝜇m
0 0.2 0.4 0.6 0.8
(b)
Figure 5: Multifunctional HD-MEAs. (a) A micrograph of the HD-MEA chip with the main functional blocks highlighted in the picture
(left). The array featured 59,760 platinum electrodes at a 13.5 μm pitch. The HD-MEA was used for impedance imaging and
electrophysiological recordings of an acute brain slice (right). The electrical recordings are compared to a micrograph of the slice;
adapted with permission from Viswam et al., IEEE Trans. Biomed. Circuits Syst. 2018 (12). (b) HD-MEA with 16 active areas, each of
which includes 1024 TiN electrodes. The main functional blocks are indicated in the chip schematic, including 64 stimulation units and
1024 voltage-recording units, 64 multiplexers (MUX), and 64 analog-to-digital converters (ADC). Impedance recordings were used to
reconstruct the spatial distribution of primary neurons that were cultured on chip and to correlate their positions with their
electrophysiological activity (right); adapted with permission from Miccoli et al., Front. Neurosci. 2019 (13)–Creative Commons
Attribution License (CC BY).

metabolites released by the bacterial film. This MEA allowed uses a set of electrodes, placed at specific locations of the
for monitoring the gradient distribution of metabolite secre- sample, to inject small alternating currents and to measure
tion over a large sensing area of 8 × 8 mm2 at 225 μm spatial the resulting voltages [76]. Currently, EIT is used in medi-
resolution, which corresponds to the electrode pitch. Up to cine and industry, for example, for label-free and continuous
38 electrodes could be read out in parallel, and each potential monitoring of patient respiratory parameters during
sweep was carried out in 0.2 seconds. Impedance imaging can mechanical ventilation [77, 78], while its use for in vitro bio-
be used to provide information that is complementary to logical applications has also been explored [76, 79–81].
functional analysis through voltage recordings or electro- A first example of EIT application in pharmaceutical
chemical imaging for an in-depth analysis of cell and tissue research included its use to monitor the dissolution of phar-
models in vitro. maceutical tablets [76]. 80 electrodes were integrated within
Electrode impedance tomography (EIT) is a technique to a test vessel, and the dissolution of sodium chloride tablets in
generate 2D or 3D impedance-based images of a sample. EIT water was monitored in real time by detecting local
BME Frontiers 9

0.53
0.318
0.106
–0.106

GND
–0.318
Chamber –0.53
Common ground
(A) (B) (C) (D)
+ Imaging
software
1
10 mm

15 mm 0.9

Correlation coefficient
0.8
0.7
0.6
0.5
15 mm ID Socket to EIT
0.4
system 0.3
2 6
1 5 0.2
0.1
0
0.07 0.16 0.28 0.44 0.64 0.87 1.14 1.44 1.78 2.15 2.56 3.00
Area ratio between the sensor and the spheroid (%)

(a) (b)
Figure 6: EIT sensor chip. (a) Schematics and picture of a planar EIT sensor at the bottom of a 15 mm diameter tissue-culture well. (b) The
sensor enabled to reconstruct the 3D shape of microtissues of different size positioned within the well (top). The correlation between the
EIT-reconstructed image and microscopy image for each microtissue size was reported (bottom); adapted with permission from Wu
et al., Analyst 2018-Published by The Royal Society of Chemistry [79].

Downloaded from https://spj.science.org on August 31, 2023

variations in conductivity caused by the dissolved salt. EIT development of reconstruction algorithms for each applica-
enabled to follow the dissolution process without interfering tion and setup is not trivial, which currently limits the wide
with the stirring of the solution during the test. However, the adoption of this imaging technique.
analysis imposed strict requirements on the sample type and
experimental conditions: the dissolution experiment needs 3. Circuitry for Impedance Measurements:
to be carried out in solutions with low conductivity, such Circuit Topologies
as distilled or tap water, and sample dissolution needs to
cause a local increase in conductivity, which restricts the Impedance sensing relies on measuring the relationship
analysis to salts or charged forms of drugs. between the voltage drop across a sample and the current
A prototype system featuring a cuboidal sample con- flowing through it. Different circuit topologies can be
tainer, equipped with 18 electrodes on opposing faces for selected according to the most relevant requirements of the
current injection and 360 voltage-sensing electrodes on three detection task, such as desired accuracy, frequency selectiv-
“imaging” sides, was proposed as a potential approach to ity, measurement speed, parallelizability, or simplicity of
monitoring tissue models in vitro [80]. The system was the implementation.
tested with different physical models of dimensions of MEAs can be assigned to two main categories: (i) passive
several millimeters to assess the performance of the micro- electrode arrays, which feature metal electrodes on top of
EIT system and the reconstruction algorithms. These pre- glass or silicon substrates that are then connected to external
liminary experimental tests were used for the optimization impedance measurement equipment; (ii) active electrode
of micro-EIT test systems, which could theoretically provide arrays, mostly fabricated in CMOS technology, which fea-
a spatial resolution of less than 100 μm for monitoring the ture monolithic integration of electrode array and, at least,
growth of small tissue models in real time. parts of the readout circuitry on the same chip. CMOS-
Imaging of breast cancer microtissues was achieved with MEAs offer the possibility to devise high-density electrode
a planar EIT sensor, which consisted of 16 microelectrodes, arrays with tens of thousands of electrodes that can be used
placed along the circumference of the base of a cylindrical for multiple functions [83, 84] and enable a high level of par-
cell-culture well of 15 mm in diameter, plus one grounded allelization. In this section, we will review the main circuit
microelectrode at the center of the well (Figure 6) [81]. By topologies and implementations for impedance recording
exploiting a novel image reconstruction algorithm, the sys- with MEAs and HD-MEAs.
tem was able to produce a 3D image representation of a
microtissue spheroid of 550 μm diameter, which corre- 3.1. Potentiostats. The operation principle of a potentiostat is
sponded to ~3.7% of the diameter of the EIT system. The based on controlling the potential difference between a
same system was then used to monitor, in real time, the loss working and a reference electrode by applying a current
of cell viability of a cancer spheroid exposed to Triton-X, through a counter electrode (Figure 7(a)) in a classical
which is known to rapidly permeabilize the cell membrane three-electrode setup. An operational amplifier with negative
[79]. New reconstruction algorithms based on machine feedback is used to adjust the current and to counterbalance
learning have been recently proposed to further improve any deviation from the target voltage. Potentiostats are
the spatial resolution of micro-EIT systems [82]. Finally, widely used for electrochemical measurements and in com-
while EIT enables to provide information on the internal mercial impedance analyzer devices [85–88]. The sample
impedance distribution within 3D cellular constructs, the impedance is calculated from the ratio between the applied
10 BME Frontiers

VDD IS IS-IR
Rs VOut
+ Is ADC
IS VDD IR
– + Vs –
CE Фc
Vin IR
WE
Фc
RE WE
RE Фd CR Фd Cint Фd

(a) (d)
RE R
LPF
VI
Vin WE Vc VOut
RS
WE
TIA
90° LPF
VQ RE

Ccell
(b)
BPF
RE VOut1
ADC t

Downloaded from https://spj.science.org on August 31, 2023

VOut
Vin WE VHi
BPF Vc
VOut2 VLo
TIA ADC
t

(c) (e)
Figure 7: Simplified schematics of the circuit topologies commonly used for impedance sensing. (a) Potentiostat-based impedance sensing.
The voltage difference between reference (RE) and working electrode (WE) is controlled by an operational amplifier, which injects a current
through the counter electrode (CE). The injected current that is required to keep the sample voltage equal to the stimulation voltage (V in )
depends on the sample impedance. The injected current can be measured by using a known resistance (Rs ) in series with the WE and by
measuring the potential drop (V s ) across it. (b) Lock-in detection. A stimulation voltage is applied to the sample, and the resulting
current is converted to a voltage using a TIA. The TIA output is multiplied with in-phase and quadrature-phase reference signals (RS) to
demodulate the signal at the frequency of interest. The demodulated signals are then low-pass filtered to eliminate out-of-target-
frequency noise and to calculate the signal amplitude and phase. (c) Simultaneous multiple-frequency detection. The applied stimulation
signal can be a square pulse or a linear combination of multiple sinusoidal waveforms. The response of the sample is measured with a
wide-range amplifier, which is then followed by multiple band-pass filters (BPFs), and digitized by a set of analog-to-digital converters
(ADCs) in parallel to simultaneously measure the sample response at multiple frequencies. (d) Charge-based capacitive sensing. The
sample capacitance and a reference capacitor (C R ) are charged to a constant voltage (V DD ) using two switches (Φc ), while measuring the
current flowing through each capacitor (I S and I R , respectively). The difference between these two currents is injected into a third
capacitor, which acts as an integrator (C int ). The resulting voltage is proportional to the difference between sample and reference
capacitances. After digitizing the output voltage, all capacitors are discharged (Φd ). (e) Capacitance-to-frequency conversion based on a
relaxation oscillator. The sample acts as a capacitor, which is repeatedly charged and discharged between two threshold voltages (V Lo
and V Hi ) by a comparator with hysteresis and a feedback resistor (R). The oscillation period is proportional to the time constant of the
resulting RC circuit, which depends on the sample capacitance.

voltage and the measured current, i.e., by using a known high dynamic range, and suitability for electrochemical mea-
resistor in series to the sample. surements. This detection method is not common in
Goikoetxea et al. designed an impedance measurement CMOS integrated circuits for impedance detection, as it
system, which featured a 20-channel, passive MEA with typically requires high circuit complexity and large silicon
TiN electrodes [87], where the electrode impedance was area and potentially entails high power consumption, all
monitored with a benchtop potentiostat by sequentially mul- features that ultimately limit parallelizability and large-
tiplexing between the electrodes. Zhang et al. presented a scale integration.
single-cell impedance measurement system with 128 passive
electrodes that used dielectric forces to locate and drive the 3.2. Lock-In Amplifiers. A lock-in amplifier enables to extract
cells on top of the measurement electrodes [88]. Moreover, low-amplitude signals with known periodicity from a noisy
the authors used an external potentiostat-based impedance background and to measure single-frequency signals with
analyzer to carry out the impedance recordings. remarkable accuracy. Lock-in detection is carried out by
Potentiostat-based detection is commonly used in com- multiplying the readout signal with in-phase and quadrature
mercial equipment owing to its wide frequency spectrum, (i.e., 90°-shifted) reference periodic signals, the periodicity of
BME Frontiers 11

which equals that of the signal of interest. The multiplication range. A similar sequential approach was also used in the
results in a demodulation of the signal information at low integrated HD-MEA, developed by Lopez et al. [72]. How-
frequencies, typically referred to as in-phase/quadrature (I/ ever, here, the authors used square-wave current-stimulation
Q) demodulation, where undesired components and out- signals, while the cell impedance was measured by detecting
of-target-frequency noise can be easily eliminated using a voltages at the sensing electrodes. This impedance measure-
low-pass filter. By multiplying with both, the in-phase and ment structure takes advantage of current generators and
quadrature reference signals, lock-in detection enables to amplifiers that were already implemented in the HD-MEA
measure the real and imaginary component of the signal for cellular stimulation and voltage recording, which resulted
[89, 90]. The same periodic signal can be used for stimula- in a compact and power-efficient design. Electrode voltage
tion and for generation of the reference signals for lock-in recording was performed by using an amplifier, equipped
demodulation, which enables to measure the absolute mag- with choppers, which allowed to demodulate the impedance
nitude and phase delay of the detected signal (Figure 7(b)) signal to low frequencies (<10 kHz) and to subsequently dig-
[10]. Typically, a sinusoidal voltage stimulation is generated itize those on-chip. The system featured 64 channels, which
by the lock-in amplifier. The current flowing through the test could be configured to provide either the in-phase or
system is converted into a voltage using a TIA before feeding it quadrature-phase signals so as to calculate the real and imag-
back to the lock-in detector for multiplication with the refer- inary parts of the sample impedance off-chip. The spatial res-
ence signals and subsequent low-pass filtering. Impedance olution of the system was limited by the multiplexed structure
values can then be extracted by calculating the ratio of the of the array, on which electrodes were arranged in pixels with
applied stimulation voltage and the measured current. four electrodes per pixel. While the electrode pitch was
Lock-in detection is widely used in active CMOS-MEAs 15 μm, only one of the four electrodes in each pixel could
for obtaining impedance recordings of cells on electrode be recorded at a given time.

Downloaded from https://spj.science.org on August 31, 2023

arrays. Chi et al. proposed an impedance sensing circuit
for an integrated MEA system, where the reference signal 3.3. Simultaneous Multifrequency Detection. Analyzing the
was applied to one electrode and the sensing current was impedance at multiple frequencies using single-frequency
measured from four adjacent electrodes [71]. The sensed lock-in detection can be time consuming, especially at low fre-
current was then multiplied with in-phase and quadrature quencies, as each frequency of interest must be measured
reference current signals, low-pass filtered, and amplified sequentially. Therefore, different detection methods have been
with a variable-gain amplifier on chip. The low-pass-filtered developed to enable the characterization of impedance across
and amplified signal was digitized off-chip using a data acqui- a wide range of frequencies at higher temporal resolution. To
sition (DAQ) board, which enabled to record up to nine sens- enable wideband detection, the stimulation signal must pres-
ing units in parallel. Impedance signals were then extracted by ent a power spectrum, distributed across multiple frequencies,
calculating the ratio between the voltage signal applied at the e.g., by using a pulse stimulation, a linear combination of sev-
reference electrode and the average current flowing through eral sinusoidal signals, or a random signal. Dedicated, special-
the four sensing electrodes. Modified versions of this imped- ized detection circuits have, therefore, been designed to be able
ance sensing circuit, all based on off-chip digitalization, have to simultaneously process the system response across a wide
been presented by the same research group [91, 92]. However, range of frequencies of interest (Figure 7(c)).
off-chip digitalization increases the complexity of the experi- Liu et al. demonstrated a pulse-based impedance-
mental setup and strongly limits the number of electrodes that spectrum-measurement system using a four-electrode setup
can be simultaneously recorded from. and off-chip components [93]. A short current impulse sig-
Viswam et al. demonstrated an impedance spectroscopy nal was applied to a cell using a pair of stimulation elec-
system, based on an integrated lock-in amplifier. The on- trodes and amplified with a differential amplifier. The
chip waveform generator provided a sinusoidal voltage stim- amplified signal was then band-pass filtered and digitized
ulation with a tunable frequency in a frequency range using a high-speed ADC. Impedance values across the spec-
between 1 Hz and 1 MHz. The current flowing through the trum of interest were then calculated by Fourier transforma-
sensing electrode was then converted to a voltage via an inte- tion of the digitized signal during postprocessing.
grated TIA, multiplied with the in-phase and quadrature ref- Bragos et al. proposed a multiple-frequency detection
erence signals, and subsequently, low-pass filtered and system that used broadband, burst stimulation signals, such
digitized on-chip using a sigma-delta analog-to-digital con- as multisine waveforms with distributed frequencies, which
verter (ADC) [12, 44]. The system included 32 impedance were provided by an arbitrary-waveform generator [94].
measurement units in parallel, which could be connected The authors used a four-electrode configuration, with two
to any of the 59,760 platinum electrodes of the array, to mea- current-stimulation electrodes and two voltage-readout elec-
sure both the real and imaginary components of the imped- trodes. The front-end circuit included a voltage-controlled
ance in parallel. However, the system only allowed to detect current source for generating the stimulation current, a dif-
the impedance at a single frequency at a time for all con- ferential amplifier for measuring the tissue-sample response,
nected electrodes. To perform impedance spectroscopy anal- and a current-to-voltage converter for monitoring the
ysis, it is required to sweep the excitation signal across applied stimulation current. All recorded signals were then
multiple frequencies sequentially, which resulted in long captured by an external digital oscilloscope, and impedance
acquisition times per frequency sweep depending on the calculation was then performed during data postprocessing.
number of detected frequency points and the frequency Although responses of the system at multiple frequencies
12 BME Frontiers

could be obtained from a single measurement, postproces- circuit used the CBCM technique to observe the capacitance
sing of the signals to extract the impedance values compli- change on the electrodes and to obtain real-time capacitance
cated data acquisition. Furthermore, since the stimulus images of cells. Finally, another example of capacitive sensing
signal contained power at multiple frequencies, the power for impedance analysis was described by Prakash et al. [100].
spectral density at each frequency inevitably was low, which The authors presented a charge-sharing-based capacitive sen-
rendered the output signal more sensitive to noise. The accu- sor for tracking cell adhesion and cell health. The authors
racy of the measurement may be even more decreased dur- could differentiate healthy and diseased cells by detecting the
ing postprocessing due to the crosstalk between signals at differences in cell adherence to the electrode surface. The
different frequencies. stronger adherence of healthy cells led to a higher capacitive
Hamilton et al. proposed a cell impedance sensor based coupling at the electrode, which could be detected by charging
on a “silicon cochlea” to provide simultaneous impedance and discharging the electrode capacitance.
sensing over a wide frequency range (from Hz to MHz) These examples show how capacitive coupling can be
[95]. The simulated circuitry included both, a multiple- used to obtain high-resolution electrical imaging of cells on
frequency signal generator (current stimulator) and an ana- electrode arrays and information on cell adherence. How-
lyzer part. In this implementation, the voltage of a sensing ever, the use of high-frequency detection limits the amount
electrode was fed to a voltage-to-current converter, and the of information that can be extracted with this technique.
corresponding current was split up into components at dif-
ferent frequencies with cascaded low-pass filters. However, 3.5. Impedance-to-Frequency Conversion. With the previous
the multiple analog outputs of the circuit require digitaliza- methods, impedance is quantified by measuring the current
tion for each frequency output of the electrodes, which mas- through (or voltage across) a sample to which a stimulus is
sively increases the complexity of the detection system. applied. However, impedance can also be estimated by inte-

Downloaded from https://spj.science.org on August 31, 2023

Simultaneous multifrequency detection has not been grating the sample as part of an electrical oscillator, so that
widely adopted for integrated CMOS systems, probably the oscillator frequency is directly dependent on the sample
due to the lower accuracy in comparison to well- impedance. Measuring oscillation frequencies can be carried
established single-frequency techniques such as I/Q demod- out in a relatively simple way, especially with modern CMOS
ulation. However, simultaneous multifrequency detection technology, where hundreds of oscillators and frequency-
may be suitable for applications where fast impedance mon- measurement units can be combined into the same integrated
itoring has priority over high accuracy [72]. circuit. Therefore, this detection method is of great interest for
achieving massively parallelized measurements. The topology
3.4. Capacitive Sensing. For many applications, measuring of the oscillator and the type of impedance to be analyzed
only the imaginary component of the impedance may be suffi- directly influence the oscillator sensitivity (i.e., the relationship
cient for extracting the information of interest, such as the pres- between oscillation frequency and impedance) and phase
ence of a cell on an electrode or cell adhesion [96]. Typically, noise (i.e., random frequency fluctuations), which dictate the
capacitive measurements can be carried out using simpler cir- fundamental limit of accuracy of this technique [101, 102].
cuit schemes than those used for absolute impedance detection. Relaxation oscillators are a common choice for capacitive
Figure 7(d) shows an example of a charge-based capacitive sensors, since the oscillation is generated by charging and dis-
measurement (CBCM) circuit to measure variations of the charging a capacitor between two voltage thresholds
electrode capacitance. First, the sensing electrode capacitance (Figure 7(e)). Van der Goes and Meijer presented a readout
(C S ) and reference capacitance (C R ) are charged (through Φc circuit based on a relaxation oscillator that has been sequen-
) to a fixed voltage, while the charging currents are sensed. tially connected to a capacitive sensor and fixed capacitors
The reference current (I R ) is then subtracted from the sensing [45]. An off-chip microcontroller was used to measure the
current (I S ), and the difference is accumulated in the integrat- oscillation frequencies, and the sample capacitance was then
ing capacitor (Cint ). The resulting voltage is afterwards digitized calculated from the differences in oscillation frequency. The
with an ADC and provided as the output. Finally, the system is use of fixed capacitors as reference enabled a continuous auto-
reset by discharging (through Φd ) the capacitors. calibration of the readout circuit, rendering this solution more
Nabovati et al. developed an 8 × 8 MEA, where the robust against undesired drifts caused by fluctuations in envi-
capacitance difference between sensing and reference elec- ronmental conditions or against fabrication process variations.
trodes was measured by using the CBCM system as illus- However, the use of an off-chip microcontroller for frequency
trated in Figure 7(d) [97, 98]. The output voltage was then measurement impedes parallelization and the integration of
digitized with an on-chip sigma-delta ADC and was ana- multiple units in an array. Stagni et al. presented a DNA sen-
lyzed to estimate the capacitance change during postproces- sor array featuring 128 oscillator-based, differential capacitive
sing. Despite featuring a comparably large electrode pitch sensors with on-chip counters for the parallelized monitoring
(~180 μm), which limited spatial resolution, the fully inte- of the respective oscillation frequencies [66].
grated MEA enabled to detect small capacitive changes
caused by single cells growing atop the array owing to the 4. Design Considerations for
interdigitated electrode structure. A much higher spatial res- Different Applications
olution was obtained by Widdershoven et al., who developed
a HD-MEA of 65,536 gold-copper nanocapacitor electrodes In addition to circuit topology for sensing, several other fac-
for high-frequency impedance imaging [11, 99]. The readout tors need to be considered for the design of a MEA or HD-
BME Frontiers 13

Table 1: Applications, circuitry architectures, and design options of some representative MEA-based impedance measurement systems.

Number Electrode
Frequency Electrode Electrode Circuitry MEA
Application of pitch
range (Hz) size material architecture type
electrodes (μm)
Potentiostat-
Liu et al. 2009 Detection of cell adhesion, spreading,
1-1 M 100 80 μm∗ Pt 200 based Passive
[120] and proliferation
sensing
Impedance-
Mucha et al. Detection of cell reactions and 55 × 55
— 64 Au — to-frequency Active
2011 [121] adhesion μm2
converter
Potentiostat-
Mamouni et al. Detection of cell adhesion, spreading, 15 × 15
10-1 M 50 Au — based Passive
2011 [122] and proliferation μm2†
sensing
Impedance imaging of cells, detection
Widdershoven Capacitive
of dynamic attachment, and cellular 1 M–70 M 65,536 180 nm∗ Au 0:6 × 0:89 Active
et al. 2015 [99] sensing
micromotion
Detection of cell adhesion,
Dragas et al. 3 × 7:5 Lock-in
differentiation, and spreading, 1-1 M 59,760 Pt black 13.5 Active
2018 [44] μm2 detection
impedance imaging of tissue slices
2:5 × 3:5

Downloaded from https://spj.science.org on August 31, 2023

μm2
Detection of cell contractility with 4:5 × 4:5
impedance monitoring module,
Lopez et al. μm2 Lock-in
detection of cell morphology, 10–1 M‡ 16,384 TiN 15 Active
2018 [72] 6:5 × 7 detection
differentiation, and adhesion with
impedance spectroscopy module μm2
10:5 × 11
μm2
Potentiostat-
Goikoetxea
Characterization of biofilm structure 1-100 k 64 60 μm∗ TiN 100 based Passive
et al. 2018 [87]
sensing
Jung et al. 2019 Detection of cell distribution, Lock-in
15 k-500 k 21,952 8 × 8 μm2 Au 16 Active
[92] growth, and proliferation detection
Nabovati et al. 5 × 25 Capacitive
Detection of cell-surface binding 1-100 k 64 PEM§ ~180 Active
2019 [98] μm2† sensing
∗Diameter of circular electrodes. †Interdigitated electrodes. ‡1 Hz and 10 kHz for fast impedance monitoring at fixed frequency. §Polyelectrolyte multilayer
films.

MEA system for impedance detection. These factors include increase of the electrode impedance as a consequence of
the geometry and material of the electrodes, the MEA topol- the smaller surface area and the smaller electrical double-
ogy, signal digitalization, and multiplexing approaches, as layer capacitance at the electrode-electrolyte interface [50].
well as sensitivity and noise characteristics. The different The effects of a high initial impedance of small electrodes
design options will be discussed in the following sections. may superimpose to or even obscure cell-dependent varia-
We then summarize a variety of representative solutions that tions at low frequencies, in particular, when the electrode
have been reported in literature in a table which compares diameter is below ~50 μm. For such small electrodes, the
different approaches to designing MEA or HD-MEA sys- electrode impedance in physiological solutions can be in
tems (Table 1). the MΩ range for frequencies up to ~10 kHz, if no surface
modification is applied to increase the effective electrode
4.1. Electrode Geometries. Electrode size, the surface-area area [104]. The selection of a suitable electrode size strongly
ratio between reference and sensing electrodes, and the elec- depends on the application of interest. For example, elec-
trode pitch are key parameters that directly affect the signal trodes of hundreds of microns in diameter can be employed
quality, detection sensitivity, and lateral resolution. for monitoring the formation of a confluent cell layer [7, 50];
Cells sedimenting on or attachment to an electrode such large electrodes feature a large sensing area and allow
causes an increase of the electrode impedance, as they block for using simplified readout circuitry schemes, however
the electric currents through fractions of the electrode sur- strongly affecting the spatial resolution and signal sensitivity
face. The measured impedance increases as the ratio between attainable.
the electrode area and cell dimension decreases [103]. Gen- The electrode size and pitch directly define the spatial
erally, a decrease in electrode size or area results in an resolution that can be attained in impedance imaging. The
14 BME Frontiers

development of integrated MEAs towards HD-MEAs featur- measurements of different sensors. Finally, porous TiN coat-
ing large numbers of small electrodes with dimensions com- ings have also been used for HD-MEAs in in vitro and
parable to the size of cells at low pitch enabled them to attain in vivo applications, as this material has shown electrical
more detailed information including tissue morphology and performance comparable to that of Au and Pt and high sta-
spreading [11]. bility under physiological conditions [13, 109]. A recent
Impedance imaging is usually carried out in conjunction trend towards the use of 3D in vitro models has also fueled
with other types of measurements, such as voltage measure- the development of 3D-MEA structures. Metal-based elec-
ments, so that an electrode size needs to be found that yields trodes (typically Au and Pt) can be fabricated on different
sufficient sensitivity and acceptable noise characteristics for micro- and nanostructures, such as flexible polyimide pillars
all implemented detection methods. Simply decreasing elec- [110] or silicon-based shanks [111], to realize 3D electrode
trode size and increasing the electrode number to improve arrays, which then can be interfaced with standard, planar
spatial resolution results in complex readout architectures readout circuitry.
to handle the large number of electrodes or increased mea- Metal materials, however, are nontransparent and pre-
surement time in case of a limited number of readout units, vent optical access to the sample with inverted microscopes,
as well as higher microfabrication costs. which represents a major limitation for standard in vitro
Moreover, the counter electrode in three-electrode investigations of biological samples. Therefore, transparent,
setups or the reference electrode in two-electrode setups conductive materials, such as indium tin oxide (ITO) or
has to be much larger than the sensing or working elec- functionalized iridium oxide (IrO2), have been explored as
trodes to provide low-impedance paths for current injec- alternative electrode materials for passive MEAs on glass
tion [50]. In MEA architectures that use alternating array wafers to enable simultaneous optical and electrical charac-
electrodes as counter/reference electrodes for each sensing terization of cells in vitro [112–115].

Downloaded from https://spj.science.org on August 31, 2023

electrode, the size of the counter/reference electrode is Recently, the use of hydrogel-based electrodes, which
defined by the array electrode size and pitch. However, more closely match the mechanical and physical properties
for MEA architectures featuring shared reference elec- of biological samples, has been proposed [116]. Hydrogel-
trodes, the reference electrode size can be chosen with based electrodes have not yet been widely applied in MEAs;
more freedom, as such shared electrodes are usually placed however, the interest in integrating hydrogel electrodes is
outside the array so that their size is not limited by the rapidly growing, also for potential use in implants in vivo
size and pitch of working or array electrodes. Finally, the [117]. Finally, new fabrication approaches, such as silicon
distance between sensing and reference electrodes also nanowires [118] or multilayered metal nanostructures [70],
affects impedance measurements [50, 103]. If working have been reported in literature to realize nanometer-
and counter or reference electrodes are placed too close dimension 3D electrodes that enable highly parallel intracel-
to each other, the current directly flows between the elec- lular recordings without damaging cellular structures.
trodes without being affected by cells or biological samples
so that the measured impedance only depends on elec- 4.3. Electrode Array Topologies. MEA systems include con-
trode and medium resistance [105, 106]. ductive electrodes and the readout circuitry to record,
amplify, and filter the signals. In passive MEAs, electrodes
4.2. Electrode Materials. Electrodes for impedance measure- are fabricated on a solid substrate (usually glass or silicon)
ments of biological samples have been realized with different [119] and connected via leads and wires to external readout
materials, such as gold (Au), platinum (Pt), silver/silver circuitry. In contrast, active or integrated MEAs or HD-
chloride (Ag/AgCl), titanium nitride (TiN), iridium oxide MEAs are usually monolithic systems, including microelec-
(IrO2), or ultrananocrystalline diamond [2, 44, 87, 107]. trodes and readout circuitry on the same silicon substrate
Gold and platinum are common material choices for MEA and taking advantage of the high level of integration pro-
fabrication, as these materials feature high biocompatibility vided by CMOS technology. The design and fabrication of
and stability and low resistivity, and they are compatible passive MEAs are simple, fast, and considerably less expen-
with microelectronic postprocessing, which is a fundamental sive than that of integrated MEAs or HD-MEAs, which facil-
requirement for the realization of highly integrated HD- itates rapid prototyping and optimization of electrode size
MEA systems. Au and Pt surface properties can be readily and shape for different applications. However, the long con-
altered by surface modification to, e.g., increase the effective nections between electrodes and readout circuits in passive
electrode surface area by increasing surface roughness, MEAs typically entail higher noise levels, and the number
which leads to a reduction of the electrode impedance by and density of electrodes in the array are limited. In contrast,
orders of magnitude [104, 108]. Examples include the depo- long and noisy connections are avoided in active MEA solu-
sition of gold nanoparticles and carbon nanotubes on gold tions, as the readout circuits are on the same substrate and at
microelectrodes [108] or the electrodeposition of platinum very short distance (typically less than a few millimeters)
black (Ptb) on platinum electrodes [104]. Surface modifica- from the sensing electrodes, which results in better noise
tion imposes an additional step in the fabrication process, performance [84]. Furthermore, reading from tens of thou-
and to achieve a uniform deposition across the whole elec- sands of (sub-)micron size electrodes at low pitch, as they
trode array can be challenging. Surface properties can vary are available in modern HD-MEAs, requires dedicated on-
among electrodes or deposition runs, and characterization chip addressing, signal processing, and sampling or multi-
of individual electrode properties is required to compare plexing schemes that can be fabricated with state-of-the-art
BME Frontiers 15

CMOS technologies [11–13]. Therefore, the availability of of high-performance ADCs, both in terms of resolution and
CMOS HD-MEA technology is key for the realization of speed, to avoid losing information during signal digitaliza-
impedance-imaging systems, as the spatiotemporal resolution tion and multiplexing. However, complex signal processing,
of the imaging is determined by electrode numbers and size, such as signal filtering and multiplication, can be imple-
electrode pitch, and the achievable data acquisition rate. mented in the analog domain and the digitalization of the
processed signal can be done afterwards [12]. In such an
4.4. Signal Conditioning, Digitalization, and Multiplexing. As approach, the design challenges mostly concern the design
mentioned above, signal conditioning, multiplexing, and of analog processing circuits, however, with the benefit of
digitalization strategies form part of fundamental design relaxed requirements for the design of the ADCs.
considerations for CMOS HD-MEA-based impedance sen-
sors. Signal conditioning typically includes amplification 4.5. Sensitivity and Noise. The detection limit of a sensor is
and filtering to transform weak and noisy signals, detected determined by its signal-to-noise ratio (SNR), which defines
at the electrodes, into robust signals for processing in subse- the minimum signal that can be reliably detected given the
quent stages. The design of the amplifiers has to be done measurement noise. Here, we refer to measurement noise
with due regard to the expected input signals and the as any fluctuation in the recorded signal that is not consid-
dynamic range and resolution in subsequent stages. The gain ered a signal, i.e., is caused by a change in the sample. High
must be large enough to ensure that the smallest features of SNR requires high sensitivity, i.e., large variations of the out-
interest remain detectable during further processing. How- put signal upon small variations in the sample, and low noise
ever, extremely large gains may cause system saturation in levels, i.e., small signal variations that are not related to
the presence of large input signals. To be able to record input changes in the sample impedance.
signals with different amplitude ranges, amplifiers can fea- Maximizing the sensitivity of impedance measurements

Downloaded from https://spj.science.org on August 31, 2023

ture variable gain, which can be adjusted with respect to requires to find an optimum measurement frequency where
the input signal amplitude [12, 72]. For measurements based a maximum change in signal output is obtained in depen-
on current sensing, signal conditioning stages may include a dence of impedance magnitude and/or phase changes of
TIA for current-to-voltage conversion, since voltage signals the sample. Simulations and analysis of sample impedance
are typically preferred for processing in subsequent stages. models can be used to determine detection frequencies, at
Analog lock-in amplifiers can be integrated as part of the which sample variations produce large impedance varia-
signal conditioning chain, including the demodulation of tions. As the effects of the parasitic elements and connection
high-frequency signals at low frequencies, which eases fur- lines are often not included in the simulations, detection fre-
ther analog processing [10, 12]. Besides amplification, filter- quencies need to be determined experimentally [50, 103,
ing is performed to attenuate out-of-band signals. High-pass 105]. Finally, sensor sensitivity is highly dependent on elec-
filtering is only required, if the expected input signal has trode size and surface properties. Therefore, the electrode
low-frequency (or continuous) components that would satu- properties, in particular the impedance of the electrode and
rate amplifiers or adapt baselines between different stages. its connections, need to be optimized to ensure good sensi-
However, low-pass filtering is required in most designs to tivity in the frequency range of interest [97].
limit the signal bandwidth before sampling and to prevent The most efficient approaches to reduce signal noise rely
aliasing in the next stages. on reducing the intrinsic noise of the read-out components,
Different multiplexing and digitalization schemes filtering out the out-of-band noise, and using differential
impose different requirements on the performance of the measurement configurations. Noise can be introduced at
different components of the readout chain, such as noise any stage of the signal generation and readout, i.e., during
characteristics, bandwidth, and available area for analog pro- stimulation, signal amplification, demodulation, and filter-
cessing units as well as the number, dynamic range, speed, ing, as well as during data conversion and transmission.
and resolution of the ADCs. A first option is to condition For the impedance measurement methods that are based
and digitize the signal directly at the sensing site, followed on applying a stimulus and recording the resulting current/
by multiplexing of the digital signals [66]. This implementa- voltage, such as lock-in detection or the use of potentiostats,
tion provides the possibility of performing parallel measure- the SNR can be improved by increasing the amplitude of the
ments with a large number of electrodes. However, the applied stimulus to the highest possible levels at which the
performance of the sensing and digitalization circuits is lim- sample is not affected and the input range of the readout cir-
ited by the available area in each electrode pixel, which, in cuits is not exceeded.
turn, determines electrode density. Frequency-domain filtering is an effective way to reduce
An alternative approach includes to perform simple sig- the noise by removing out-of-band signal contributions,
nal conditioning at each sensing electrode and to multiplex since specific sample features may only be detected within
the resulting signals to a set of ADCs outside the active area defined frequency ranges. Ideally, the filter should feature a
for signal digitalization, which disentangles the ADC very narrow passband around the frequencies of interest
requirements and electrode pixel area (electrode size and for the impedance measurement. The design of such filters
pitch) [72]. This alternative implementation requires a lower can be challenging in the analog domain, so signals are
number of ADCs, and signal processing (including lock-in frequently postprocessed in the digital domain. However,
detection) can be carried out digitally on chip or during filtering is remarkably effective and simple for impedance
off-chip postprocessing. The challenge here lies in the design sensors based on lock-in detection, since filtering of the
16 BME Frontiers

downmodulated signal after multiplication can be carried analysis units, especially in case of multifrequency measure-
out using a dedicated low-pass filter [12]. Finally, differential ments. Therefore, digitalization and signal preprocessing,
schemes, whether applied during signal acquisition or post- such as Fast Fourier Transformations (FFTs), directly on-
processing, are an effective means to greatly reduce the effect chip will likely play a key role in the development of highly
of power-supply noise or environmental noise. Acquisition parallelized HD-MEAs as they offer the possibility to rapidly
of differential signals, however, requires more circuitry, acquire impedance recordings over a large spectral range at
therefore larger areas for implementation, which could high temporal resolution. However, the development of
impact electrode density or increase fabrication costs. While CMOS-MEAs requires—owing to design complexity and
differential values can often be calculated during data post- fabrication costs of integrated circuits—much larger time
processing, this solution may yield lower accuracy and less and financial investments in comparison to the fabrication
noise suppression than differential signal acquisition. of passive MEAs. Therefore, the development of custom-
designed CMOS-MEAs for specific applications can only
5. Outlook be justified by either unique functionality or performance
that cannot be achieved otherwise or by large numbers of
Impedance imaging is a real-time, label-free, and noninva- devices that can be produced at comparably low costs.
sive measurement method, which can be used for long- Finally, technological and theoretical improvements in
term characterization of cells and tissue models. The small electrical impedance tomography (EIT) may also increase
form factors of the sensor chips and the electronic nature the relevance of impedance techniques for studying biologi-
of the detection render this technique particularly suitable cal samples. Multifrequency impedance tomography (also
for parallelization and automation. Real-time monitoring known as electrical impedance tomography spectroscopy;
of cell and tissue dynamics can be performed very efficiently EITS) is a promising noninvasive technique to analyze

Downloaded from https://spj.science.org on August 31, 2023

and rapidly in comparison to optical microscopy. High sam- frequency-dependent characteristics of a sample [126]. The
pling rates, which can be achieved by measuring the sample complexity of the computations required to reconstruct the
impedance at a single high frequency, enable to detect rapid image of the sample requires novel reconstruction algo-
dynamics, such as cell micromotions, on the sensor surface rithms and efficient hardware implementations. Machine
at high spatiotemporal resolution and over large sensing learning techniques may prove useful for handling and sim-
areas (up to few mm2). Although EIS imaging may feature plifying the high-dimensional data required through EIT
a low level of specificity in the discrimination of different cell [127, 128], especially for imaging approaches involving mul-
types, which only depends on the dielectric properties of the tifrequency measurements. The resolution and robustness of
cells, impedance-based imaging enables to characterize sam- EIT can also be improved by including a priori knowledge
ple optical access to which is not possible, such as tissue about the sample and adjusting image reconstruction algo-
slices on MEAs during electrophysiological recordings or rithms to exclude reconstructed images that are not compat-
brain tissue in vivo while using MEA-based brain implants ible with the known physical properties of the sample [129,
[123–125]. The development of MEA brain implants has 130]. Passive electrode arrays with low numbers of elec-
been mainly focused on the recording of neural activity or trodes can be controlled through field-programmable gate
brain stimulation. However, the presence of hundreds or arrays (FPGAs), on which efficient postprocessing algo-
thousands of implanted electrodes would also enable to rithms can be implemented [131, 132]. CMOS HD-MEAs
study the dielectric properties of the environment in which are well suited for performing EIT measurements owing to
the devices were implanted and to perform imaging under their small pitch and large number of electrodes. However,
conditions of no optical access. Detecting variations in the given that CMOS HD-MEAs are prevailingly implemented
surroundings of the implants could potentially be useful to in older technologies (such as 0.18 μm), the performance of
track the position of electrodes or to monitor electrode sur- on-chip digital circuitry is poor in comparison to modern
face properties and the integrity of the host tissue over lon- nanometer-technology CMOS processors. Therefore, it
ger periods of time. may be more efficient to perform image reconstruction using
Further developments of MEAs or HD-MEAs for separate dedicated electronic chips.
impedance imaging revolve around increasing the number
of electrodes and reducing their pitch to achieve high spatial Conflicts of Interest
resolution. CMOS-based HD-MEAs represent the only via-
ble approach to attain (sub)micrometer resolution and The authors declare that there are no conflicts of interest
highly parallel detection, although the integration of thou- regarding the publication of this article.
sands of compact and power-efficient circuits represents a
major design challenge. As advances in CMOS technology Authors’ Contributions
have greatly improved the efficiency of digital circuits in
comparison to their analog counterparts, an early digitaliza- All the authors wrote the manuscript.
tion of signals and a minimization of analog processing on
chip may provide a viable solution. At the same time, Acknowledgments
increasing the number of measurement units will produce
large datasets that need to be recorded and processed, which This work was financially supported by the European
requires fast data-acquisition systems and powerful data- Research Council Advanced Grant 694829 “neuroXscales”,
BME Frontiers 17

by the Swiss National Science Foundation under Contracts models of epithelial tissues,” Frontiers in Bioengineering and
205320_188910 and CR32I2_166329 and a Marie Heim- Biotechnology, vol. 7, 2020.
Vögtlin grant to R.B., and by an ETH Postdoctoral Fellow- [16] W. Franks, I. Schenker, P. Schmutz, and A. Hierlemann,
ship to F. C. “Impedance characterization and modeling of electrodes for
biomedical applications,” IEEE Transactions on Biomedical
Engineering, vol. 52, no. 7, pp. 1295–1302, 2005.
References [17] K. R. Foster and H. P. Schwan, “Dielectric properties of tis-
sues and biological materials: a critical review,” Critical
[1] A. Yúfera, G. Huertas, and A. Olmo, “A microscopy tech- Reviews in Biomedical Engineering, vol. 17, no. 1, pp. 25–
nique based on bio-impedance sensors,” Procedia Engineer- 104, 1989.
ing, vol. 47, pp. 1049–1052, 2012. [18] G. H. Markx and C. L. Davey, “The dielectric properties of
[2] Y. Xu, X. Xie, Y. Duan, L. Wang, Z. Cheng, and J. Cheng, “A biological cells at radiofrequencies: applications in biotech-
review of impedance measurements of whole cells,” Biosen- nology,” Enzyme and Microbial Technology, vol. 25, no. 3-5,
sors & Bioelectronics, vol. 77, pp. 824–836, 2016. pp. 161–171, 1999.
[3] H. Fricke, “The electrical capacity of suspensions with special [19] C. Gabriel, S. Gabriel, and E. Corthout, “The dielectric prop-
reference to blood,” The Journal of General Physiology, vol. 9, erties of biological tissues: I. Literature survey,” Physics in
no. 2, pp. 137–152, 1925. medicine & biology, vol. 41, no. 11, pp. 2231–2249, 1996.
[4] H. Fricke and S. Morse, “The electric resistance and capacity of [20] L. Y. Zhang, C. B. M. Du Puch, C. Dalmay et al., “Discrimina-
blood for frequencies between 800 and 4½ million cycles,” The tion of colorectal cancer cell lines using microwave biosen-
Journal of General Physiology, vol. 9, no. 2, pp. 153–167, 1925. sors,” Sensors and Actuators A: Physical, vol. 216, pp. 405–
[5] W. H. Coulter, Means for Counting Particles Suspended in a 416, 2014.

Downloaded from https://spj.science.org on August 31, 2023

Fluid, 1953, US Patent 2,656,508. [21] J.-C. Chien, A. Ameri, E.-C. Yeh, A. N. Killilea, M. Anwar,
[6] C. R. Keese, J. Wegener, S. R. Walker, and I. Giaever, “Electri- and A. M. Niknejad, “A high-throughput flow cytometry-
cal wound-healing assay for cells in vitro,” Proceedings of the on-a-CMOS platform for single-cell dielectric spectroscopy
National Academy of Sciences of the United States of America, at microwave frequencies,” Lab on a Chip, vol. 18, no. 14,
vol. 101, no. 6, pp. 1554–1559, 2004. pp. 2065–2076, 2018.
[7] I. Giaever and C. R. Keese, “Micromotion of mammalian cells [22] I. Popov, P. Ben Ishai, A. Khamzin, and Y. Feldman, “The
measured electrically,” Proceedings of the National Academy mechanism of the dielectric relaxation in water,” Physical
of Sciences of the United States of America, vol. 88, no. 17, Chemistry Chemical Physics, vol. 18, no. 20, pp. 13941–
pp. 7896–7900, 1991. 13953, 2016.
[8] I. Giaever and C. R. Keese, “Monitoring fibroblast behavior in [23] D. R. Merrill and P. A. Tresco, “Impedance characterization of
tissue culture with an applied electric field,” Proceedings of the microarray recording electrodes in vitro,” IEEE Transactions
National Academy of Sciences of the United States of America, on Biomedical Engineering, vol. 52, no. 11, pp. 1960–1965, 2005.
vol. 81, no. 12, pp. 3761–3764, 1984. [24] P. S. Ravaynia, F. C. Lombardo, S. Biendl et al., “Parallelized
[9] C. A. Thomas, P. A. Springer, G. E. Loeb, Y. Berwald-Netter, impedance-based platform for continuous dose-response
and L. M. Okun, “A miniature microelectrode array to mon- characterization of antischistosomal drugs,” Advanced bio-
itor the bioelectric activity of cultured cells,” Experimental systems, vol. 4, no. 7, pp. 1900304–1900312, 2020.
Cell Research, vol. 74, no. 1, pp. 61–66, 1972. [25] R. A. Hoffman, T. S. Johnson, and W. B. Britt, “Flow cytomet-
[10] A. Manickam, A. Chevalier, M. McDermott, A. D. Ellington, ric electronic direct current volume and radiofrequency
and A. Hassibi, “A CMOS electrochemical impedance spec- impedance measurements of single cells and particles,”
troscopy (EIS) biosensor array,” IEEE Transactions on Bio- Cytometry, vol. 1, no. 6, pp. 377–384, 1981.
medical Circuits and Systems, vol. 4, no. 6, pp. 379–390, 2010. [26] G. Mernier, W. Hasenkamp, N. Piacentini, and P. Renaud,
[11] C. Laborde, F. Pittino, H. A. Verhoeven et al., “Real-time “Multiple-frequency impedance measurements in continu-
imaging of microparticles and living cells with CMOS nano- ous flow for the evaluation of electrical lysis of yeast cells,”
capacitor arrays,” Nature Nanotechnology, vol. 10, no. 9, Procedia Engineering, vol. 5, pp. 37–40, 2010.
pp. 791–795, 2015. [27] D. Holmes and H. Morgan, “Single cell impedance cytometry
[12] V. Viswam, R. Bounik, A. Shadmani et al., “Impedance spec- for identification and counting of CD4 T-cells in human
troscopy and electrophysiological imaging of cells with a blood using impedance labels,” Analytical Chemistry,
high-density CMOS microelectrode array system,” IEEE vol. 82, no. 4, pp. 1455–1461, 2010.
Transactions on Biomedical Circuits and Systems, vol. 12, [28] S. Gawad, L. Schild, and P. Renaud, “Micromachined imped-
no. 6, pp. 1356–1368, 2018. ance spectroscopy flow cytometer for cell analysis and parti-
[13] B. Miccoli, C. M. Lopez, E. Goikoetxea et al., “High-density cle sizing,” Lab on a Chip, vol. 1, no. 1, pp. 76–82, 2001.
electrical recording and impedance imaging with a multi- [29] A. Yúfera, A. Rueda, J. M. Muñoz, R. Doldán, G. Leger, and
modal CMOS multi-electrode array chip,” Frontiers in Neu- E. O. Rodríguez-Villegas, “A tissue impedance measurement
roscience, vol. 13, pp. 1–14, 2019. chip for myocardial ischemia detection,” IEEE Transactions
[14] S. L. Tsai, M. H. Wang, M. K. Chen, and L. S. Jang, “Analyt- on Circuits and Systems I: Regular Papers, vol. 52, no. 12,
ical and numerical modeling methods for electrochemical pp. 2620–2628, 2005.
impedance analysis of single cells on coplanar electrodes,” [30] J. R. Macdonald, “Impedance spectroscopy,” Annals of Bio-
Electroanalysis, vol. 26, no. 2, pp. 389–398, 2014. medical Engineering, vol. 20, no. 3, pp. 289–305, 1992.
[15] T. Gerasimenko, S. Nikulin, G. Zakharova et al., “Impedance [31] O. Casas, R. Bragós, P. J. Riu et al., “In Vivo and In Situ Ische-
spectroscopy as a tool for monitoring performance in 3D mic Tissue Characterization Using Electrical Impedance
18 BME Frontiers

Spectroscopya,” Annals of the New York Academy of Sciences, [48] R. T. Kurulugama, D. O. Wipf, S. A. Takacs, S. Pongmayteegul,
vol. 873, pp. 51–58, 1999. P. A. Garris, and J. E. Baur, “Scanning electrochemical micros-
[32] M. Gheorghiu, E. Gersing, and E. Gheorghiu, “Quantitative copy of model neurons: constant distance imaging,” Analytical
analysis of impedance spectra of organs during ischemia,” Chemistry, vol. 77, no. 4, pp. 1111–1117, 2005.
Annals of the New York Academy of Sciences, vol. 873, no. 1 [49] C. R. Keese and I. Giaever, “Use of electric fields to monitor
ELECTRICAL BI, pp. 65–71, 1999. the dynamical aspect of cell behavior in tissue culture,” IEEE
[33] K. Darowicki and P. Ślepski, “Determination of electrode Transactions on Biomedical Engineering, vol. BME-33, no. 2,
impedance by means of exponential chirp signal,” Electro- pp. 242–247, 1986.
chemistry Communications, vol. 6, no. 9, pp. 898–902, 2004. [50] X. Zhang, W. Wang, A. N. Nordin, F. Li, S. Jang, and
[34] A. Mejía-Aguilar and R. Pallàs-Areny, “Electrical impedance I. Voiculescu, “The influence of the electrode dimension on
measurement using voltage/current pulse excitation,” in Pro- the detection sensitivity of electric cell-substrate impedance
ceedings XIX IMEKO World Congress: Fundamental and sensing (ECIS) and its mathematical modeling,” Sensors and
Applied Metrology, pp. 662–667, Lisbon, Portugal, 2009. Actuators B: Chemical, vol. 247, pp. 780–790, 2017.
[35] S. Gawad, T. Sun, N. G. Green, and H. Morgan, “Impedance [51] P. O. Bagnaninchi and N. Drummond, “Real-time label-free
spectroscopy using maximum length sequences: application monitoring of adipose-derived stem cell differentiation with
to single cell analysis,” The Review of Scientific Instruments, electric cell-substrate impedance sensing,” Proceedings of
vol. 78, no. 5, article 054301, 2007. the National Academy of Sciences of the United States of
[36] D. Rairigh, A. Mason, and C. Yang, “Analysis of on-chip America, vol. 108, no. 16, pp. 6462–6467, 2011.
impedance spectroscopy methodologies for sensor arrays,” [52] C. Petchakup, H. Li, and H. W. Hou, “Advances in single cell
Sensor Letters, vol. 4, no. 4, pp. 398–402, 2006. impedance cytometry for biomedical applications,” Micro-
[37] J. J. Ackmann, “Complex bioelectric impedance measure- machines, vol. 8, no. 3, p. 87, 2017.

Downloaded from https://spj.science.org on August 31, 2023

ment system for the frequency range from 5 Hz to 1 MHz,” [53] T. Sun and H. Morgan, “Single-cell microfluidic impedance
Annals of Biomedical Engineering, vol. 21, no. 2, pp. 135– cytometry: a review,” Microfluidics and Nanofluidics, vol. 8,
146, 1993. no. 4, pp. 423–443, 2010.
[38] J. Malmivuo and R. Plonsey, Bioelectromagnetism: Principles [54] Z. Zhu, O. Frey, N. Haandbaek, F. Franke, F. Rudolf, and
and Applications of Bioelectric and Biomagnetic Fields, A. Hierlemann, “Time-lapse electrical impedance spectros-
Oxford University Press, New York, 1995. copy for monitoring the cell cycle of single immobilized S.
[39] T. K. Bera, “Bioelectrical impedance methods for noninvasive pombe cells,” Scientific Reports, vol. 5, no. 1, p. 17180, 2015.
health monitoring: a review,” Journal of Medical Engineering, [55] J. Ahn, J. H. Ahn, S. Yoon, M. Y. Son, S. Cho, and J. H. Oh,
vol. 2014, 28 pages, 2014. “Quantification of non-alcoholic fatty liver disease progres-
[40] W. R. B. Lionheart, “EIT reconstruction algorithms: pitfalls, sion in 3D liver microtissues using impedance spectroscopy,”
challenges and recent developments,” Physiological Measure- Biomaterials, vol. 268, article 120599, 2021.
ment, vol. 25, no. 1, pp. 125–142, 2004. [56] A. Mansoorifar, A. Koklu, and A. Beskok, “Quantification of
[41] A. Rothermel, M. Nieber, J. Müller, P. Wolf, M. Schmidt, and cell death using an impedance-based microfluidic device,”
A. A. Robitzki, “Real-time measurement of PMA-induced cellu- Analytical Chemistry, vol. 91, no. 6, pp. 4140–4148, 2019.
lar alterations by microelectrode array-based impedance spec- [57] S. C. Bürgel, L. Diener, O. Frey, J. Y. Kim, and A. Hierlemann,
troscopy,” BioTechniques, vol. 41, no. 4, pp. 445–450, 2006. “Automated, multiplexed electrical impedance spectroscopy
[42] J. S. Park, S. I. Grijalva, M. K. Aziz et al., “Multi-parametric platform for continuous monitoring of microtissue spher-
cell profiling with a CMOS quad-modality cellular interfacing oids,” Analytical Chemistry, vol. 88, no. 22, pp. 10876–
array for label-free fully automated drug screening,” Lab on a 10883, 2016.
Chip, vol. 18, no. 19, pp. 3037–3050, 2018. [58] X. Ren, P. Ghassemi, J. S. Strobl, and M. Agah, “Biophysical
[43] A. Hassibi and T. H. Lee, “A programmable 0.18-$\mu\h- phenotyping of cells via impedance spectroscopy in parallel
box{m}$ CMOS electrochemical sensor microarray for bio- cyclic deformability channels,” Biomicrofluidics, vol. 13,
molecular detection,” IEEE Sensors Journal, vol. 6, no. 6, no. 4, article 044103, 2019.
pp. 1380–1388, 2006. [59] H. Song, J. M. Rosano, Y. Wang et al., “Identification of mes-
[44] J. Dragas, V. Viswam, A. Shadmani et al., “<italic>In vitro</ enchymal stem cell differentiation state using dual-micropore
italic> multi-functional microelectrode array featuring 59 microfluidic impedance flow cytometry,” Analytical Methods,
760 electrodes, 2048 electrophysiology channels, stimulation, vol. 8, no. 41, pp. 7437–7444, 2016.
impedance measurement, and neurotransmitter detection [60] M. Evander, A. J. Ricco, J. Morser et al., “Microfluidic imped-
channels,” IEEE Journal of Solid-State Circuits, vol. 52, ance cytometer for platelet analysis,” Lab on a Chip, vol. 13,
no. 6, pp. 1576–1590, 2017. no. 4, pp. 722–729, 2013.
[45] F. M. L. Van Der Goes and C. Gerard, “A novel low-cost [61] E. Rollo, E. Tenaglia, R. Genolet et al., “Label-free identifica-
capacitive-sensor interface,” IEEE Transactions on Instrumen- tion of activated T lymphocytes through tridimensional
tation and measurement, vol. 45, no. 2, pp. 536–540, 1996. microsensors on chip,” Biosensors & Bioelectronics, vol. 94,
[46] R. F. Taylor, I. G. Marenchic, and E. J. Cook, “An acetylcho- pp. 193–199, 2017.
line receptor-based biosensor for the detection of cholinergic [62] L. Dong, P. S. Ravaynia, Q. A. Huang, A. Hierlemann, and
agents,” Analytica Chimica Acta, vol. 213, pp. 131–138, 1988. M. M. Modena, “Parallelized wireless sensing system for con-
[47] J. M. Liebetrau, H. M. Miller, J. E. Baur et al., “Scanning elec- tinuous monitoring of microtissue spheroids,” ACS Sensors.,
trochemical microscopy of model neurons: imaging and real- vol. 5, no. 7, pp. 2036–2043, 2020.
time detection of morphological changes,” Analytical Chem- [63] T. S. Pui, Y. Chen, C. C. Wong et al., “High density CMOS
istry, vol. 75, no. 3, pp. 563–571, 2003. electrode array for high-throughput and automated cell
BME Frontiers 19

counting,” Sensors and Actuators B: Chemical, vol. 181, [78] C. Putensen, B. Hentze, S. Muenster, and T. Muders, “Electri-
pp. 842–849, 2013. cal impedance tomography for cardio-pulmonary monitor-
[64] Y. Chen, C. C. Wong, T. S. Pui et al., “CMOS high density ing,” Journal of Clinical Medicine, vol. 8, no. 8, p. 1176, 2019.
electrical impedance biosensor array for tumor cell detec- [79] H. Wu, Y. Yang, P. O. Bagnaninchi, and J. Jia, “Electrical
tion,” Sensors and Actuators B: Chemical, vol. 173, pp. 903– impedance tomography for real-time and label-free cellular
907, 2012. viability assays of 3D tumour spheroids,” The Analyst,
[65] C. Begly, D. Ackart, J. Mylius, R. Basaraba, A. J. Chicco, and vol. 143, no. 17, pp. 4189–4198, 2018.
T. W. Chen, “Study of real-time spatial and temporal behav- [80] E. Lee, H. Wi, A. McEwan et al., “Design of a microscopic
ior of bacterial biofilms using 2-D impedance spectroscopy,” electrical impedance tomography system for 3D continuous
IEEE Transactions on Biomedical Circuits and Systems, non-destructive monitoring of tissue culture,” Biomedical
vol. 14, no. 5, pp. 1051–1064, 2020. Engineering Online, vol. 13, no. 1, p. 142, 2014.
[66] C. Stagni, C. Guiducci, L. Benini et al., “CMOS DNA sensor [81] Y. Yang, J. Jia, S. Smith, N. Jamil, W. Gamal, and P.-O. O.
array with integrated A/D conversion based on label-free Bagnaninchi, “A miniature electrical impedance tomogra-
capacitance measurement,” IEEE Journal of Solid-State Cir- phy sensor and 3-D image reconstruction for cell imag-
cuits, vol. 41, no. 12, pp. 2956–2964, 2006. ing,” IEEE Sensors Journal, vol. 17, no. 2, pp. 514–523,
[67] H. M. Jafari and R. Genov, “CMOS impedance spectrum ana- 2017.
lyzer with dual-slope multiplying ADC,” in 2011 IEEE Bio- [82] Z. Chen, Y. Yang, and P.-O. Bagnaninchi, “Hybrid learning-
medical Circuits and Systems Conference (BioCAS), pp. 361– based cell aggregate imaging with miniature electrical imped-
364, San Diego, CA, USA, 2011. ance tomography,” IEEE Transactions on Instrumentation
[68] F. Heer, W. Franks, A. Blau et al., “CMOS microelectrode and Measurement, vol. 70, pp. 1–10, 2021.
array for the monitoring of electrogenic cells,” Biosensors & [83] M. E. J. Obien, K. Deligkaris, T. Bullmann, D. J. Bakkum, and

Downloaded from https://spj.science.org on August 31, 2023

Bioelectronics, vol. 20, no. 2, pp. 358–366, 2004. U. Frey, “Revealing neuronal function through microelec-
[69] M. E. J. Obien, W. Gong, U. Frey, and D. J. Bakkum, trode array recordings,” Frontiers in Neuroscience, vol. 8,
“CMOS-based high-density microelectrode arrays: technol- p. 423, 2015.
ogy and applications,” in Emerging Trends in Neuro Engi- [84] A. Hierlemann, U. Frey, S. Hafizovic, and F. Heer, “Growing
neering and Neural Computation, pp. 3–39, Springer, cells atop microelectronic chips: interfacing electrogenic cells
Singapore, 2017. in vitro with CMOS-based microelectrode arrays,” Proceed-
[70] B. X. E. Desbiolles, E. De Coulon, A. Bertsch, S. Rohr, and ings of the IEEE, vol. 99, no. 2, pp. 252–284, 2011.
P. Renaud, “Intracellular recording of cardiomyocyte action [85] T. A. Nguyen, T.-I. Yin, D. Reyes, and G. A. Urban, “Micro-
potentials with nanopatterned volcano-shaped microelectrode fluidic chip with integrated electrical cell-impedance sensing
arrays,” Nano Letters, vol. 19, no. 9, pp. 6173–6181, 2019. for monitoring single cancer cell migration in three-
[71] T. Chi, J. S. Park, J. C. Butts et al., “A multi-modality CMOS dimensional matrixes,” Analytical Chemistry, vol. 85, no. 22,
sensor array for cell-based assay and drug screening,” IEEE pp. 11068–11076, 2013.
Transactions on Biomedical Circuits and Systems, vol. 9, [86] C. Hildebrandt, H. Büth, S. Cho, H. T. Impidjati, and
no. 6, pp. 801–814, 2015. H. Thielecke, “Detection of the osteogenic differentiation of
[72] C. M. Lopez, H. S. Chun, S. Wang et al., “A multimodal mesenchymal stem cells in 2D and 3D cultures by electro-
CMOS MEA for high-throughput intracellular action poten- chemical impedance spectroscopy,” Journal of Biotechnology,
tial measurements and impedance spectroscopy in drug- vol. 148, no. 1, pp. 83–90, 2010.
screening applications,” IEEE Journal of Solid-State Circuits, [87] E. Goikoetxea, D. Routkevitch, A. de Weerdt, J. J. Green,
vol. 53, no. 11, pp. 3076–3086, 2018. H. Steenackers, and D. Braeken, “Impedimetric fingerprint-
[73] J. Rothe, O. Frey, A. Stettler, Y. Chen, and A. Hierlemann, ing and structural analysis of isogenic E. coli biofilms using
“Fully integrated CMOS microsystem for electrochemical multielectrode arrays,” Sensors and Actuators B: Chemical,
measurements on 32 × 32 working electrodes at 90 frames vol. 263, pp. 319–326, 2018.
per second,” Analytical Chemistry, vol. 86, no. 13, pp. 6425– [88] Z. Zhang, T. Zheng, and R. Zhu, “Microchip with single-cell
6432, 2014. impedance measurements for monitoring osteogenic differ-
[74] D. L. Bellin, H. Sakhtah, Y. Zhang, A. Price-Whelan, L. E. P. entiation of mesenchymal stem cells under electrical stimula-
Dietrich, and K. L. Shepard, “Electrochemical camera chip tion,” Analytical Chemistry, vol. 92, no. 18, pp. 12579–12587,
for simultaneous imaging of multiple metabolites in bio- 2020.
films,” Nature Communications, vol. 7, no. 1, pp. 1–10, 2016. [89] G. Gervasoni, M. Carminati, and G. Ferrari, “Lock-In Ampli-
[75] W. Tedjo, J. E. Nejad, R. Feeny et al., “Electrochemical bio- fier Architectures for Sub-ppm Resolution Measurements,”
sensor system using a CMOS microelectrode array provides in Advanced Interfacing Techniques for Sensors, pp. 191–
high spatially and temporally resolved images,” Biosensors 217, Springer, Cham, 2017.
& Bioelectronics, vol. 114, pp. 78–88, 2018. [90] Y. Wu, F. F. Hanzaee, D. Jiang, R. H. Bayford, and
[76] V. Rimpiläinen, L. M. Heikkinen, M. Kuosmanen et al., “An A. Demosthenous, “Electrical impedance tomography for bio-
electrical impedance tomography-based approach to medical applications: circuits and systems review,” IEEE Open
monitorin vitrosodium chloride dissolution from pharma- Journal of Circuits and Systems, vol. 2, pp. 380–397, 2021.
ceutical tablets,” The Review of Scientific Instruments, [91] J. S. Park, M. K. Aziz, S. Li et al., “1024-pixel CMOS multimo-
vol. 80, no. 10, article 103706, 2009. dality joint cellular sensor/stimulator array for real-time
[77] B. Lobo, C. Hermosa, A. Abella, and F. Gordo, “Electrical holistic cellular characterization and cell-based drug screen-
impedance tomography,” Annals of Translational Medicine, ing,” IEEE Transactions on Biomedical Circuits and Systems,
vol. 6, no. 2, pp. 26–26, 2018. vol. 12, no. 1, pp. 80–94, 2018.
20 BME Frontiers

[92] D. Jung, J. S. Park, G. V. Junek et al., “A 21952-pixel multi- ance sensing,” Sensors and Actuators B: Chemical, vol. 288,
modal CMOS cellular sensor array with 1568-pixel parallel pp. 707–715, 2019.
recording and 4-point impedance sensing,” in 2019 Sympo- [106] R. Bounik, M. Gusmaroli, P. M. Misun, V. Viswam,
sium on VLSI Circuits, pp. C62–C63, Kyoto, Japan, June A. Hierlemann, and M. M. Modena, “Integration of discrete
2019. sensors and microelectrode arrays into open microfluidic
[93] Y. Liu, E. W. Abel, J. J. F. Belch, and S. Chen, “An impulse hanging-drop networks,” in 2019 IEEE 32nd International
response flow cytometric technique for blood cell Conference on Micro Electro Mechanical Systems (MEMS),
characterisation-instrumentation and preliminary evalua- pp. 441–444, Seoul, Korea (South), 2019.
tion,” in Proceedings of the 20th Annual International Confer- [107] K. Cheung, S. Gawad, and P. Renaud, “Impedance spectros-
ence of the IEEE Engineering in Medicine and Biology Society. copy flow cytometry: on-chip label-free cell differentiation,”
Vol.20 Biomedical Engineering Towards the Year 2000 and Cytometry Part A, vol. 65, pp. 124–132, 2005.
Beyond (Cat. No.98CH36286), pp. 1881–1884, Hong Kong,
[108] V. Srinivasaraghavan, J. Strobl, D. Wang, J. R. Heflin, and
China, 1998.
M. Agah, “A comparative study of nano-scale coatings on gold
[94] R. Bragós, R. Blanco-Enrich, O. Casas, and J. Rosell, “Charac- electrodes for bioimpedance studies of breast cancer cells,” Bio-
terisation of dynamic biologic systems using multisine based medical Microdevices, vol. 16, no. 5, pp. 689–696, 2014.
impedance spectroscopy,” in Conference Record - IEEE
[109] J. J. Jun, N. A. Steinmetz, J. H. Siegle et al., “Fully integrated
Instrumentation and Measurement Technology Conference,
silicon probes for high-density recording of neural activity,”
vol. 1, pp. 44–47, Budapest, Hungary, 2001.
Nature, vol. 551, no. 7679, pp. 232–236, 2017.
[95] T. J. Hamilton, N. M. Nelson, D. Sander, and P. Abshire, “A
[110] D. A. Soscia, D. Lam, A. C. Tooker et al., “A flexible 3-
cell impedance sensor based on a silicon cochlea,” in 2009
dimensional microelectrode array for in vitro brain models,”
IEEE Biomedical Circuits and Systems Conference, pp. 117–
Lab on a Chip, vol. 20, no. 5, pp. 901–911, 2020.
120, Beijing, China, 2009.

Downloaded from https://spj.science.org on August 31, 2023

[96] A. Hedayatipour, S. Aslanzadeh, and N. McFarlane, “CMOS [111] H. Shin, S. Jeong, J. H. Lee, W. Sun, N. Choi, and I. J.
based whole cell impedance sensing: challenges and future Cho, “3D high-density microelectrode array with optical
outlook,” Biosensors & Bioelectronics, vol. 143, article stimulation and drug delivery for investigating neural cir-
111600, 2019. cuit dynamics,” Nature Communications, vol. 12, no. 1,
pp. 1–18, 2021.
[97] G. Nabovati, E. Ghafar-Zadeh, A. Letourneau, and M. Sawan,
“Towards high throughput cell growth screening: a new [112] E. Moore, O. Rawley, T. Wood, and P. Galvin, “Monitoring of
CMOS 8 × 8 biosensor array for life science applications,” cell growth in vitro using biochips packaged with indium tin
IEEE Transactions on Biomedical Circuits and Systems, oxide sensors,” Sensors and Actuators B: Chemical, vol. 139,
vol. 11, no. 2, pp. 380–391, 2017. no. 1, pp. 187–193, 2009.
[98] G. Nabovati, E. Ghafar-Zadeh, A. Letourneau, and M. Sawan, [113] J. Martinez, A. Montalibet, E. McAdams, M. Faivre, and
“Smart cell culture monitoring and drug test platform using R. Ferrigno, “Effect of electrode material on the sensitivity
CMOS capacitive sensor array,” IEEE Transactions on Bio- of interdigitated electrodes used for Electrical Cell-Substrate
medical Engineering, vol. 66, no. 4, pp. 1094–1104, 2019. Impedance Sensing technology,” in 2017 39th Annual Inter-
national Conference of the IEEE Engineering in Medicine
[99] F. Widdershoven, A. Cossettini, C. Laborde et al., “A CMOS
and Biology Society (EMBC), pp. 813–816, Jeju, Korea
pixelated nanocapacitor biosensor platform for high-
(South), 2017.
frequency impedance spectroscopy and imaging,” IEEE
Transactions on Biomedical Circuits and Systems, vol. 12, [114] M. Parviz, P. Toshniwal, H. M. Viola et al., “Real-time bioim-
no. 6, pp. 1369–1382, 2018. pedance sensing of antifibrotic drug action in primary human
cells,” ACS Sensors., vol. 2, no. 10, pp. 1482–1490, 2017.
[100] S. B. Prakash, P. Abshire, M. Urdaneta, and E. Smela, “A
CMOS capacitance sensor for cell adhesion characterization,” [115] D. M. Lewis, N. Mavrogiannis, Z. Gagnon, and S. Gerecht,
in IEEE International Symposium on Circuits and Systems “Microfluidic platform for the real time measurement and
(IEEE, 2005), pp. 3495–3498, Kobe, Japan, 2005. observation of endothelial barrier function under shear
[101] H. Wang, C. C. Weng, and A. Hajimiri, “Phase noise and fun- stress,” Biomicrofluidics, vol. 12, no. 4, article 042202, 2018.
damental sensitivity of oscillator-based reactance sensors,” [116] N. Adly, S. Weidlich, S. Seyock et al., “Printed microelectrode
IEEE Transactions on Microwave Theory and Techniques, arrays on soft materials: from PDMS to hydrogels,” npj flexi-
vol. 61, no. 5, pp. 2215–2229, 2013. ble electronics, vol. 2, no. 1, pp. 1–9, 2018.
[102] F. Cardes, A. Quintero, E. Gutierrez, C. Buffa, A. Wiesbauer, [117] R. Green, “Elastic and conductive hydrogel electrodes,”
and L. Hernandez, “SNDR limits of oscillator-based sensor Nature Biomedical Engineering, vol. 3, no. 1, pp. 9-10, 2019.
readout circuits,” Sensors, vol. 18, no. 2, p. 445, 2018. [118] J. Yoo, H. Kwak, J. Kwon et al., “Long-term intracellular
[103] L. Wang, H. Wang, L. Wang, K. Mitchelson, Z. Yu, and recording of optogenetically-induced electrical activities
J. Cheng, “Analysis of the sensitivity and frequency character- using vertical nanowire multi electrode array,” Scientific
istics of coplanar electrical cell-substrate impedance sensors,” reports, vol. 10, no. 1, pp. 1–10, 2020.
Biosensors & Bioelectronics, vol. 24, no. 1, pp. 14–21, 2008. [119] V. Emmenegger, M. E. J. Obien, F. Franke, and
[104] V. Viswam, M. E. J. Obien, F. Franke, U. Frey, and A. Hierlemann, “Technologies to study action potential prop-
A. Hierlemann, “Optimal electrode size for multi-scale agation with a focus on HD-MEAs,” Frontiers in Cellular
extracellular-potential recording from neuronal assemblies,” Neuroscience, vol. 13, pp. 1–11, 2019.
Frontiers in Neuroscience, vol. 13, 2019. [120] Q. Liu, J. Yu, L. Xiao et al., “Impedance studies of bio-
[105] Y. T. Lai, Y. S. Chu, J. C. Lo, Y. H. Hung, and C. M. Lo, behavior and chemosensitivity of cancer cells by micro-
“Effects of electrode diameter on the detection sensitivity electrode arrays,” Biosensors & Bioelectronics, vol. 24, no. 5,
and frequency characteristics of electric cell-substrate imped- pp. 1305–1310, 2009.
BME Frontiers 21

[121] A. Mucha, M. Schienle, and D. Schmitt-Landsiedel, “A

CMOS integrated impedance-to-frequency converter for
sensing cellular adhesion,” Advances in Radio Science,
vol. 9, pp. 281–287, 2011.
[122] J. Mamouni and L. Yang, “Interdigitated microelectrode-
based microchip for electrical impedance spectroscopic study
of oral cancer cells,” Biomedical Microdevices, vol. 13, no. 6,
pp. 1075–1088, 2011.
[123] M. Kollo, R. Racz, M.-E. Hanna et al., “CHIME: CMOS-
hosted in vivo microelectrodes for massively scalable neuro-
nal recordings,” Frontiers in Neuroscience, vol. 14, 2020.
[124] B. Dutta, E. M. Trautmann, M. Welkenhuysen et al., “The
Neuropixels probe: A CMOS based integrated microsystems
platform for neuroscience and brain-computer interfaces,”
in 2019 IEEE International Electron Devices Meeting
(IEDM)pp. 10.1.1–10.1.4, San Francisco, CA, USA, 2019.
[125] E. Musk and Neuralink, “An integrated brain-machine inter-
face platform with thousands of channels,” Journal of Medical
Internet Research, vol. 21, no. 10, article e16194, 2019.
[126] J. P. Leitzke and H. Zangl, “A review on electrical impedance
tomography spectroscopy,” Sensors, vol. 20, pp. 1–19, 2020.

Downloaded from https://spj.science.org on August 31, 2023

[127] Y. Fan and L. Ying, “Solving electrical impedance tomogra-
phy with deep learning,” Journal of Computational Physics,
vol. 404, article 109119, 2020.
[128] Z. Chen, Y. Yang, J. Jia, and P. Bagnaninchi, “Deep Learning
Based Cell Imaging with Electrical Impedance Tomography,”
in 2020 IEEE International Instrumentation and Measure-
ment Technology Conference (I2MTC), pp. 1–6, Dubrovnik,
Croatia, 2020.
[129] H. Kudo, M. Courdurier, F. Noo, and M. Defrise, “Tiny a
priori knowledge solves the interior problem in computed
tomography,” Physics in Medicine and Biology, vol. 53,
no. 9, pp. 2207–2231, 2008.
[130] D. Flores-Tapia and S. Pistorius, “Electrical impedance
tomography reconstruction using a monotonicity approach
based on a priori knowledge,” in 2010 Annual International
Conference of the IEEE Engineering in Medicine and Biology
Society, EMBC’10, pp. 4996–4999, Buenos Aires, Argentina,
2010.
[131] B. M. G. Rosa and G. Z. Yang, “Bladder volume monitoring
using electrical impedance tomography with simultaneous
multi-tone tissue stimulation and DFT-based impedance cal-
culation inside an FPGA,” IEEE Transactions on Biomedical
Circuits and Systems, vol. 14, no. 4, pp. 775–786, 2020.
[132] M. Darnajou, A. Dupre, C. Dang et al., “High speed EIT with
multifrequency excitation using FPGA and response analysis
using FDM,” IEEE Sensors Journal, vol. 20, no. 15, pp. 8698–
8710, 2020.