Siemens-Healthineers Mi Alpha White Paper

White paper
Automatic landmarking and

parsing of human anatomy
(ALPHA) for innovative and
smart MI applications
Vijay Shah, Ph.D.
siemens-healthineers.com
Table of contents
Acronyms 3
Introduction 4
ALPHA 5
Core idea: mimic human vision 6
How ALPHA learns 6
Increase accuracy and reliability via redundancy 7
Applications on intelligent PET/CT scanner platform (AIDAN) 10
FlowMotion™ AI 11
OncoFreeze™ AI 13
Multiparametric PET AI 16
FAST PET Workflow AI 18
MI Applications on the syngo®.via intelligent reading solution 19
Automated image registration 19
Creation of reference regions for Lesion Scout and Deauville criteria 22
Deauville criteria 23
Spine and rib labeling 24
Auto views 25
Auto ranges 26
Conclusion 27
Acknowledgments 27
References 28
2 White paper · Automatic landmarking and parsing of human anatomy (ALPHA) for innovative and smart MI applications
Acronyms
AC_CT Attenuation correction CT
AI Artificial intelligence
ALPHA Automatic landmarking and parsing of human anatomy
CT Computed tomography
FAST Fully assisting scanner technologies
FDG Fluorodeoxyglucose
HL Hodgkin lymphoma
Ki Metabolic rate
LEAP Learning ensembles of anatomical patterns
MI Molecular imaging
MR Magnetic resonance
MRFDG (18F FDG) metabolic rate
MRGlu Glucose metabolic rate
NHL Non-Hodgkin lymphoma
PACS Picture archiving and communication system
PERCIST PET response criteria in solid tumors
PET Positron emission tomography
ROI Region of interest
SiPM Silicon photomultiplier
SPECT Single photon emission computed tomography
SUV Standardized uptake value
VD Distribution volume
WB Whole-body
White paper · Automatic landmarking and parsing of human anatomy (ALPHA) for innovative and smart MI applications 3
Introduction
Over the past decades, molecular imaging (MI) has grown into a vital tool for
precision medicine with potential to offer tremendous capabilities for the detection
and assessment of the significance of diseases, risk stratification, therapy selection,
and monitoring in oncology, cardiology, and neurology.1 Modern MI systems are
typically hybrid systems—PET/CT or SPECT/CT—that seamlessly combine the richness
in anatomical images (CT) with functional images (PET and SPECT). Furthermore,
technology advances in equipment, such as the introduction of silicon photomultiplier
(SiPM)-based PET detector, have increased spatial resolution and sensitivity that can be
utilized to obtain images of increased quality at lower radiation doses. Improvements in
reconstruction algorithms, such as xSPECT™ technology, further expand the spectrum
of clinical applications by incorporating CT-based tissue segmentation into a zonal
image reconstruction to enable a new degree of image quality2 as well as quantitative
SPECT measurements.
In the era of value-based medicine, it has become essential to assist technologists and
clinicians by reducing manual, laborious, and non-essential work to enable the efficient
processing of high volumes of data in a fashion that is efficient, robust, reproducible,
and less operator dependent. This will also allow them to focus their time and attention
on patients and critical clinical questions. Thus, to increase throughput without
sacrificing quality requires innovative, impactful features using unique intelligent
technologies.
Artificial intelligence (AI)-powered algorithms impact the entire image acquisition,

reconstruction, and interpretation of molecular images. Automatic landmarking and
parsing of human anatomy (ALPHA), a Siemens Healthineers proprietary technology,
supports optimized workflows on both scanners and syngo.via reading solutions.
This white paper introduces ALPHA technology as well as scanner technologies and
syngo.via features driven by ALPHA technology.
ALPHA
ALPHA1-6 can detect and contextualize anatomical structures at various levels
of anatomical abstraction, as shown in Figure 1.
Figure 1
(a)
Organ Segmentation
(eg, a ﬁne segmentation of
kidney boundary on CT)
Bounding Boxes
(eg, bounding box for left kidney)
Slice Ranges
(eg, ﬁrst and last slices containing colon)
Anatomical Orientation
(eg, knee meniscus midsagital plane)
Figure 1. (a) ALPHA, example

of how it works. (b) ALPHA
Anatomical detects various anatomical
Landmarks structures to assist the
(eg, celiac bifurcation, acquisition or reading of
apex of left lung, etc.) medical images. Several
examples from left to right
of the image demonstrate
robustness of ALPHA
technology: automatic
(b) vertebra localization and
labelling performed to provide
planes at C1-C7, T1-T12,
L1-L5, S (abnormal patient
positioning); results for head
scan ranges—the lower edge
should go through the
canthomeatal line (works
even when full anatomy is not
present); detection of bounding
boxes around the breast or
separately for right and left
breast regions on MRI images;
plane alignment for optimal
imaging of the menisci in
MRI images.
Core idea: mimic human vision
ALPHA can be used to recognize and infer anatomical patterns and consists of an
algorithm to detect anatomical landmarks at its core, as shown in Figure 1a. An
anatomical landmark is a discrete labelled point in 2D or 3D medical images that can
be used as a reference point by readers to identify anatomy, eg, apex of left lung or
top of liver dome. The algorithm is designed to mimic the human foveal vision system,
which only focuses on points of interest, for example, landmark points while perceiving
a rough and blurred peripheral context at any given time. Typically, the human visual
system analyzes the complete scene and use redundancy and relationships of
surrounding structures to achieve a reliable recognition of objects or context. ALPHA was
designed in the same way to recognize an abundant number of anatomical landmarks
and structures by exploiting the rich context and high redundancy in a medical imaging.
The underlying technology is a multi-layered, machine-learning algorithm called learning
ensembles of anatomical patterns (LEAP).4-5
How ALPHA learns

ALPHA at its core has an engine to identify landmarks. The algorithm has been trained
using many expert-annotated images across modalities consisting of many landmarks
distributed throughout the body as shown in Figure 2. Using a statistical learning and
feature selection algorithm, the engine learned to formulate a knowledge base to
detect landmarks based on image features that best discriminate the target landmark
appearance from the other structures.
Algorithm learning is based on training using images present in a vast database

without making any explicit assumptions regarding the target anatomy or modality.
This allows ALPHA to be highly scalable to different anatomical structures and to
different imaging modalities (eg, CT or MRI) or appearance (eg, 2D images, 3D images,
topograms). For a new image volume, a runtime engine uses the knowledge base to
identify landmarks. A sample landmark detection in liver is shown in Figure 2.
Figure 2
Annotated
Annotated
database Supervised learning phase
database
ALPHA
ALPHA
landmarking Annotated
landmarking
learning Annotated
database
learning
engine database
engine
Figure 2. The supervised

learning phase of the ALPHA
landmarking engine. The
database consists of image data
annotated by expert readers. The ALPHA
landmarking
algorithm is trained on this data ALPHA
learning
landmarking
engine
to create a knowledge base that learning
engine
is used by the runtime engine
to detect landmarks in a new New image Landmark
image (center of liver anatomical detection
landmark is detected).
Increase accuracy and reliability via redundancy3
In the presence of a visual illusion, the human brain is vulnerable to misinterpretation
and ‘human error;’ conversely, machine learning algorithms similarly are also susceptible
to misinterpretation.7 In order to achieve robustness against diseases or imaging
artifacts, ALPHA uses multiple layers of redundancy3 to increase the reliability of the
results produced by the algorithm. The concept of redundancy has been shown to
work well in safety-critical systems, as failure of a single component can be either
detected or corrected by other working components. ALPHA uses three redundancy
concepts to achieve robustness.
1. R
edundancy of training data based on anatomical landmarks
Research has shown that ensemble learning algorithms, such as bagging6, improves
performance of the system: learning is performed by selecting a few samples for
training from the available training data to train multiple detectors. In contrast, ALPHA
uses a spatial re-alignment scheme that uses all available training data.3 The training
data was aligned according to anatomical landmarks that were annotated by the
experts as shown in Figure 3. Because of inherent anatomical variability, all landmarks
cannot be aligned together. Redundancy was included by replicating the data for
multiple landmarks within the body.3-6
Figure 3
(a)
Redundancy via spatial ensemble through re-alignment
(b)
(c)
Figure 3. Spatial ensemble

learning to exploit redundan-
cies and dependencies for
(d)
improved robustness. Original
training data (a) is spatial
ensemble through re-align-
ment, for example all training
data is duplicated and realigned
Consensus decision via voting to different anatomical land-
marks (b). Finally, the median
value is calculated from all the
training datasets (c) followed
by running the landmark
detection algorithm3-6 in (d).
Final target prediction of the
(e) landmark (e) is performed
using the consensus decision
made via a voting process.
Thus, if n-landmarks are annotated in data, there will be n times the original data
present for the training as new dataset is created by re-alignment of n-landmarks
as shown in Figure 3b. The advantage of spatial alignment can be seen in Figure 3c,
the median image, where images are much sharper in the region of landmark point
and blurry at the region away from the landmark point. The detector algorithm is
run separately on each of these images, creating redundant prediction of landmarks,
which is then appropriately selected based on the voting scheme described later in
section 3.
2. Redundancy from different voxel resolutions

Traditionally for object detection, a coarse-to-fine level search strategy is employed
to improve the speed of target search. ALPHA employs a multi-scale approach to
landmark detection but is optimized to the trade-off between efficiency (time to
detect the landmarks) and robustness (accuracy of landmark location). To introduce
redundancy, the detection performed at finer scales is not dependent on proposal
obtained by detection performed at the coarser level. This allows a finer detector to
search in a range as wide as allowable by the speed requirements of the application.
Such a strategy can capture occasional errors, especially at finer scales. Ambiguity or
contradiction can happen on medical images, for example, when an abnormal region
in liver has a shape like a right lung, as shown for illustrative purposes in Figure 4. In
such scenarios, contextual information helps resolve the ambiguity and redundancy
examinations across different scales.
Figure 4
Figure 4. Liver image with

artificially inserted abnormality
for illustration purpose, and the
zoom of the inserted defect (a), (a)
and coronal thoracic view (b).
On a coarse scale, the artificially
inserted region looks like just
another abnormal region in the
liver, however, a finer scale
shape of the same region
appears similar to the right lung
region and traditional algorithms
may detect candidate lungs
landmarks (eg, apex of right
lung) in this region. However,
such a discrepency can be
detected and ruled out by the
ALPHA through the voting
mechanism. (b)
3. Redundancy via voting on other anatomical landmarks
Due to redundancy design during the detection, multiple candidate anatomical
landmarks are detected independently. In this step, every landmark casts a vote for
every landmark. Thus, each detected landmark serves as a candidate and a voter.
Voting exploits redundancy of spatial relationships between landmarks in a way where
correct landmarks reinforce each other, and incorrect landmarks are voted out. The
cumulative information is used to define landmark locations with both confidence,
and precision. The illustrative example in Figure 5 demonstrates how ALPHA
landmarking achieves robustness through redundancy. This example shows four
candidate landmarks for the “Right Costodiaphragmatic Recess,” which have similar
image features. Correct landmarks are positively reinforced by being voted in (thick
lines), while incorrect landmarks are voted out (thin lines). Both local and distant
anatomy is accounted for during the voting.
The algorithm accuracy is based on the following: performance evaluations using at

least 200 clinically relevant cases, all landmarks are correctly detected, at least 85%
of the cases with no more than 5% of false positives. With the reference hardware[a],
the execution time for detecting landmarks is not more than one second per landmark
on average.
Figure 5
Figure 5. Illustrative
example demonstrating the
voting mechanism of ALPHA
landmarking algorithm
[a] CPU: 4 cores, 64 bit; RAM:>=32 GB
Applications on intelligent
PET/CT scanner platform (AIDAN)
The intelligent imaging platform, AIDAN, for our Biograph™ family of PET/CT scanners,
includes the proprietary ALPHA technology to bring artificial intelligence to PET scanner
operations, enabling more efficient, personalized, and standardize exams. The four
features enabled by the AIDAN platform[b] are shown in Figure 6.
Figure 6
AI AI
ALPHA ALPHA
Multiparametric PET Suite AI OncoFreeze AI
AIDAN
AI AI
ALPHA ALPHA
FlowMotion AI FAST PET Workflow AI

Figure 6. ALPHA-powered
scanner features in the
AIDAN platform.
[b] AIDAN platform scanner software version VG80/VJ30
FlowMotion AI
FlowMotion AI integrates FlowMotion8 continuous bed motion technology and
anatomical algorithms to automatically plan scan ranges using a CT topogram. With
FlowMotion AI9, landmarks associated with the top of the head, beneath the eye orbits,
the clavicle, the aortic arch, the adrenal gland, the mid-thigh, and the tip of the toes are
obtained from the CT topogram and used to establish PET scan ranges routinely used
within clinical workflows. Figure 7 and Figure 8 illustrate the PartialBody and TotalBody
configurations, respectively, which are typically defined on anterior-posterior (AP)
topograms. The PartialBody options represent “eyes-to-thighs” scans, with optional lung
and head-neck ranges. The TotalBody options span from the top of the head to the tip
of the toes, and include configurations for lung, head-neck, and legs ranges.
Additionally, an AP heart and a lateral brain range are also offered for single bed
acquisitions, as shown in Figure 9. The heart and brain ranges are designed to
respectively position the heart and cerebrum in the center of the field of view (FoV).
Figure 7
Figure 7. PartialBody
scan configurations in
FlowMotion AI.
Figure 8
Figure 8. TotalBody
scan configurations in
FlowMotion AI.
Figure 9
Figure 9. Heart (a) and brain

Heart Brain (b) scan configurations
FlowMotion AI enables a PET acquisition that can be personalized for an individual
patient but standardized in clinical protocol strategy. This is achieved by specifying one
or more scan ranges, as is exemplified in Figure 10, and setting corresponding bed
speeds for each range. For example, Figure 10c illustrates a protocol strategy customized
for colorectal or prostate cancer, where the speed is reduced within the pelvic and
abdominal region to improve image statistics in the most clinically relevant area. The
bed speed can also be increased substantially outside the boundary of the region of
interest (ROI) to improve efficiency. Furthermore, respiratory gating strategies can
be employed to a certain range to reduce the impact of patient motion.
Protocols can be stored, including the indication-based scan and reconstruction

parameters according to hospital preference. This protocol can be reloaded to ensure
reproducibility in multi-timepoint or follow-up studies and reduces user dependency.
Figure 10
H&N Lung/liver Colorectal/prostate Lymphoma/melanoma
High High
resolution resolution
Normal
speed
Respiratory
Normal gating
speed
Normal
speed High
resolution
Fast Fast
speed speed
Fast
speed
(a) (b) (c) (d)
Figure 10. FlowMotion scan ranges can be tailored to patient anatomy and the clinical indication.
OncoFreeze AI
Physiological motion of internal organs causes motion-related blurring that affects
quantification and delineation of the smallest detectable lesions and structures more
severely, making compensation of respiratory motion in PET and PET/CT increasingly
important.10 Almost 90% of all oncological PET/CT cases involve lesions in the chest or
abdomen.10 In oncology, motion-related blurring is commonly seen in lung nodules and
tumors, as well as in liver and pancreatic tumors.11 Research shows that displacement
happens in organs and lesions by a range of 5-30 mm, blurring images and reducing
diagnostic confidence.12 Without respiratory gating, 40% of lung lesions may even not
be detected.13 Thoracic lesions tend to be more susceptible to respiratory motion that
can cause errors in definition of target volumes during planning of radiation therapy.
However, image acquisition for PET requires several minutes, so lesion blurring due to
respiratory motion is common. This can lead to an over-estimation of lesion volume, as
well as under-estimation of radionuclide uptake by the lesion. However, challenges in
respiratory motion management limit the use of gated images:
• Traditional motion management using physiological devices is labor intense and error
prone to provide for every patient
• Setup of external physiological devices can add complexity and additional time to
patient exams
• Acquiring enough counts to create a motion-frozen image often requires additional

time to scan a patient if traditional gating approaches are being used
Figure 11
(a)
Deviceless waveform signal Waveform signal from physiological device
Range affected by respiration

Signal
Figure 11. Shows the range of

body to perform respiratory
gating (a) and range of body
Head/Neck Clavicle Upper Pubis Feet
habitus potentially affected by
Position (b) respiratory motion (b).
Siemens Healthineers OncoFreeze,14-17 the existing solution for respiratory motion
correction for oncology, consists of three distinct steps:
1. Respiratory gating: This step involves sorting of the acquired raw data into different
gates based on the point in the respiratory cycle, at which the events were detected
2. Estimation of Motion Information: In this step images reconstructed from the gated
raw data are used to estimate motion between the gates
3. Motion Correction: Information from previous two steps is then applied to the image
reconstruction to generate motion-corrected PET images
OncoFreeze AI benefits from its predecessor, OncoFreeze.14-17 Both solutions support

the generation of motion-frozen images without any additional scan time; however
OncoFreeze AI does not require an external gating device.
OncoFreeze AI is designed to offer motion-frozen PET images in areas affected by

respiratory motion (as shown in Figure 11) by enabling a simple check box on the
scanner UI. This feature for oncological whole-body acquisitions enables deviceless
respiratory motion correction by integrating the Siemens Healthineers deviceless
waveform technology and eliminating the need for external respiratory sensors.
With FlowMotion1 as the foundation for image acquisition, the scanner bed moves
continuously. ALPHA technology identifies anatomy that typically is impacted by
Figure 13
Figure 13. Sample Siemens

Healthineers deviceless
waveform signal.
respiratory motion, and the deviceless waveform technique generates the respiratory
signal without any additional setup or patient interaction. Representative waveforms
from three patients are shown in Figure 13. Finally, using OncoFreeze14-17 for the image
reconstruction, 100% of the acquired PET counts can be used for image reconstruction,
which means no additional scan time for a motion-frozen image. An example image
comparison between static and OncoFreeze image is shown in Figure 14.
OncoFreeze AI, being a deviceless approach for motion correction, also improves
patient experience by eliminating the need of external device and helps to acquire
high-quality images with no additional scan time.
Figure 14
Figure 14. Comparison of static

MIP and OncoFreeze MIP. Data
courtesy of University Medical
Center Groningen, Groningen,
The Netherlands. Injected dose:
5.1 mCi (190 MBq), 18F-FDG;
Scan acquisition: 440 x 440
matrix, PSF + TOF 4i5s,
0.8 mm/sec FlowMotion
(a) (b) continuous bed motion.
Multiparametric PET AI
The implementation of SUV facilitates the utilization of semi-quantitative metrics in
the clinic for PET imaging. However, SUV cannot account for the PET tracer distribution
over time—a dynamic process altered by several factors specific to each organ and ROI.
Parametric analysis of dynamic PET helps to measure in vivo physiological processes
such as 18F fluorodeoxyglucose (18F FDG) metabolism (MRFDG), blood perfusion, oxygen
consumption, cell proliferation, and receptor density. MRFDG can be used to infer the
glucose metabolic rate (MRGlu) with corrections for metabolic differences between
18
F FDG and glucose, but, due to a typically cumbersome workflow to perform MRFDG,
its measurement remains a large, unmet need in the clinic to both stage and monitor
cancer treatments.18
The intense manual effort for measuring MRFDG obtained based on Patlak modeling18
is no longer required with the implementation of Multiparametric PET Suite AI, which
leverages several unique innovations to provide a robust solution: FlowMotion
continuous bed motion technology, whole-body (WB) dynamic imaging, ALPHA
technology, and direct reconstruction (see Figure 15).
Fludeoxyglucose F 18 5-10mCi as an IV injection

Indications and usage • Blood Glucose Abnormalities: In the oncology and
Fludeoxyglucose F 18 Injection is indicated for positron neurology setting, suboptimal imaging may occur in
emission tomography (PET) imaging in the following patients with inadequately regulated blood glucose
settings: levels. In these patients, consider medical therapy
•O ncology: For assessment of abnormal glucose and laboratory testing to assure at least two days
metabolism to assist in the evaluation of malignancy of normoglycemia prior to Fludeoxyglucose F 18
in patients with known or suspected abnormalities Injection administration.
found by other testing modalities, or in patients with • Adverse Reactions: Hypersensitivity reactions with
an existing diagnosis of cancer. pruritus, edema and rash have been reported; have
•C ardiology: For the identification of left ventricular emergency resuscitation equipment and personnel
myocardium with residual glucose metabolism immediately available. Full prescribing information
and reversible loss of systolic function in patients for Fludeoxyglucose F 18 Injection can be found at
with coronary artery disease and left ventricular the conclusion of this publication.
dysfunction, when used together withmyocardial
perfusion imaging. Dosage forms and strengths
•N eurology: For the identification of regions of multiple-dose 30 mL and 50 mL glass vial containing
abnormal glucose metabolism associated with 0.74 to 7.40 GBq/mL (20 to 200 mCi/mL) of Flude
foci of epileptic seizures. oxyglucose F 18 injection and 4.5 mg of sodium
chloride with 0.1 to 0.5% w/w ethanolas a stabilizer
Important safety information (approximately 15 to 50 mL volume) for intravenous
•R adiation Risks: Radiation-emitting products, administration.
including Fludeoxyglucose F 18 Injection, may
increase the risk for cancer, especially in pediatric
patients. Use the smallest dose necessary for imaging Fludeoxyglucose F 18 injection is manufactured by
and ensure safe handling to protect the patient and Siemens’ PETNET Solutions, 810 Innovation Drive,
health care worker. Knoxville, TN 39732
or indications and important safety information for Fludeoxyglucose F 18 injection

F
(18F FDG) see page 16. For full prescribing information see pages 31-33.
Figure 15
(a)
SUV Metabolic rate (Ki) Distribution Volume (DV)
(b)
Figure 15. Multiparametric

workflow showing FlowMotion
acquisition (a) and SUV
as well as derived images—
MRFDG and DV (b).
The Patlak model is a graphical analysis technique based on a compartment model to estimate
the dynamic physiological process of PET using linear regression.18 The acquisition protocol for
parametric PET differs from the standard SUV acquisition protocol. The subject is either injected
on the bed at the same time as the PET acquisition is started (Figure 15), or the initial bolus
concentration in the blood is estimated using non-imaging techniques. The input function
obtained via non-imaging techniques, such as blood draws, can be invasive and requires
significant manual intervention, prohibiting its use for routine clinical practice. Research has
shown that an image derived input function can be substituted for invasive blood draw
techniques.18-19 However, a user is required to manually draw the ROI in a blood pool region,
eg, descending aorta or left ventricle, which can be automated with the use of AI technology.
In this fully automated approach, CT and dynamic PET data are used to generate an
image-derived input function (IDIF). Using ALPHA technology, landmarks are found
in left ventricle (LV) and proximal descending aorta on the CT images. The CT
reconstruction that is used for the PET attenuation correction (AC_CT by default) is
used to find these landmarks. These landmarks are transferred to the PET image to
create the volume of interest (VOI) around the landmark point in the descending aorta
(cylinder: 5 mm radius, 20 mm height), and LV (30 mm diameter sphere). The IDIF can
therefore be automatically generated from both the LV blood pool and proximal
descending aorta using ALPHA as shown in Figure 16.
An internal analysis on 110 WB AC_CTs showed a success rate of >95%. Potential

reasons for failure are incomplete CT scan coverage over the heart and descending
aorta regions, or if the algorithm reports low confidence in the location of the landmark.
If an input function VOI is not found automatically, the user will be required to generate
one manually on the scanner console, export the time activity curve, and reload it into
the scanner user interface.
Figure 16
Automatic derivation of Arterial Input Function (AIF)
Figure 16. Image derived input

function from left ventricle and
AI algorithm identifies anatomical landmarks.1
placement of reference region Region of intererst (ROI) is defined, eg, descending aorta.
in the aorta. AIF is derived using dynamic images and ROI.
FAST PET Workflow AI
Once the scan is complete, technologists may still have multiple tasks on the scanner
console to prepare a patient exam for reading and archiving. Such additional tasks limit
time for a technologist to spend on patient care. Additionally, such repetitive tasks that
could be automated can influence reduction of technologist fatigue.
Radiologists have no need to review blank images (as shown in Figure 17; depicted by
red lines), as it is inefficient to manually scroll through air as well as taxing on the data
storage systems. Technologists therefore often manually review images and either define
ranges to remove blank images or auto-create ranges and subsequently select the images
containing tissue to be sent to picture archiving and communication system (PACS).
Fully Assisting Scanner Technologies (FAST) Workflow Auto PACS-ready ranges use ALPHA
to identify a skin-to-skin range of images and automatically creates ranges with anatomy
present in the images. Figure 17 illustrates an example with the eliminated slices in red
and useful slices shown in white. With the FAST PET Workflow AI feature, we remove the
blank images in coronal, sagittal, and axial views. Additionally, images can be exported
to a predefined storage location at any specific time of the day. Since blank slices are
removed, it also helps to reduce the storage space for PACS, as shown in Figure 18.
Figure 17
Figure 17. Eliminated slices are

shown in red and retained slices
encompassing the anatomy are
shown in white.
Figure 18
#of images: traditional #of images: fast #of blank images removed
CT PET Fused CT PET Fused
Transaxial 325 325 325 324 324 324 3
Sagittal 155 145 155 127 127 127 74
Coronal 155 145 155 70 70 70 245
• Without blank image removal these images would have to be

deleted from the patient browser manually before being exported.
– This results in 54 additional user interactions per patient

examination.
– Also conserves system storage space. In one example 0.51 GB
less than orginal.
Figure 18. Sample reduction in
blank slices and storage space.
MI applications on the syngo.via
intelligent reading solution
AI-powered applications in syngo.via to enable efficient clinical analysis of PET and
SPECT images are shown in Figure 19.
Figure 19
AI AI
ALPHA ALPHA
Automated image registration Spine and rib labeling
AI AI
ALPHA SEG
Auto Ranges Auto Lung 3D
syngo.via
AI AI
SEG ALPHA
RT Image Suite Auto Views
Figure 19. MI syngo.via AI

applications incorporating INDENT
ALPHA technology.
[c] Lesion Scout with Auto ID

is not available for sale in the
USA and is not commercially
+
available in all countries.
Future availability cannot be
guaranteed. Please contact
your local Siemens Healthineers Lesion Scout Auto ID[c]
organization for further details.
Automated image registration
The utility of molecular imaging is expanding in the selection of patients who would
most benefit from targeted therapy, early treatment responders to predict therapeutic
efficacy, and tumor response to survival.20 Clinicians need to visually compare the same
lesion(s) from two or more timepoints but also quantitatively measure SUV changes.
Traditional image registration algorithms have helped clinicians by automatically

spatially aligning studies from multiple timepoints. These conventional algorithms
often have limitations when imaging conditions change from one timepoint to the
other, for example, with different FoVs (see Figure 20), or with patient posture changes
(eg, hands-up versus hands down), or with different patient table setup (eg, a thick
versus a thin cushion). Traditional image registration algorithms using pixel-by-pixel
comparison logic may find a “local minimum” by, for example, matching shoulder with
pelvis or by matching up the two tables instead of the patient anatomy. Clinicians must
perform manual alignment of the image volumes if the algorithm fails. This is a tedious
task which reduces throughput and can cause reader fatigue when reading many cases.
As described previously, ALPHA can automatically recognize anatomical structures in

medical images and is well suited to the task of robust image registration. Depending
on the anatomical scan range, ALPHA detects multiple (> 4) landmarks in each study,
filters them based on an anatomical consistency check and uses the overlapping
landmarks to align studies from different time-points. Because ALPHA registration is
based on recognition of anatomical structures just like the way a human observer
would do it, instead of low-level pixel matching, it is robust to the variations mentioned
above. Figure 21 shows an example with grossly mismatched CT FoVs, where ALPHA
landmark-based registration is still robust for lesion tracking.
Furthermore, as ALPHA’s capability is learned from examples and trained to recognize

landmarks in MRI images as well, it can even align studies from different modalities
(CT/MRI) as shown in Figure 22.
Figure 20
Figure 20. Registration of

two studies with different
FoVs—Head&Neck region and
extended torso region with
no common overlap region.
Baseline image shown to have
lesion (L8M1) in Head&Neck
region (a). Follow up scan
(a) shown in (b), where
registration of baseline and
follow up performed using
“optimal” pixel-based match
registration algorithm resulted
in matching the lesion (L8M1)
at top of the lung with area
in pelvis region (L8M2) on
follow-up scan. While tracking
lesions over time, a reader
will have to correct for misregis-
tration to look at trends for a
(b) particular lesion over time.
Figure 21
Figure 21. Alignment of multiple

time-point studies using ALPHA
registration full length CT
available at all time point (a)
and non-overlapping region (b).
(a) The lesion on Prior2 L5VOI1

(bottom row) is automatically
registered to the same lesion
on follow-up scans L5VOI2
(Prior1, middle row) and
L5VOI3 (current, top row)
when full images CT images are
available for registration.
(b) To demonstrate the

robustness of registration
algorithm based on ALPHA-
landmarks, a new dataset was (a)
created so that CT images on
all three timepoints from (a)
had different FoVs. These
timepoints were again loaded
in syngo.via MM Oncology to
see if registration (performed
with on CT images) was robust
for tracking the lesions.
The PET FoV was not changed.
The lesion on Prior2 L1VOI1
(bottom row) is automatically
correctly registered to lesion
on follow-up scans L1VOI2
(prior1, middle row) and L1VOI3
(current, top row) even when
CT images do not have matching
FoVs. While performing ALPHA-
landmarking the algorithm can
predict the location of land-
marks that are not present in
the FoV, which enables a very
robust registration. (b)
Figure 22
Figure 22. Multi-modality

registration between CT image
(a) and MRI images (b). Various
landmark points are common
between MRI and CT images.
Registration algorithm is
optimized to register those
landmark points between
two images to get the optimal
alignment. Registration is
robust even when the CT
image is acquired with hands-
up, while the MRI image is
acquired with hands down.
Creation of reference regions for Lesion Scout and Deauville criteria
Lesion Scout (formerly known as Multi-Foci Segmentation (MFS) in previous versions
of syngo.via) is designed to offer the reading physician a way to segment uptake and
create findings based on pre-selected criteria. Lesion Scout works on PET or SPECT
data from any quantitative reconstruction that results in data with SUV values and is
agnostic of the tracer used. This enables the reader to customize the segmentation
criteria to varying clinical needs, different radiopharmaceuticals, or local site
segmentation preferences. This feature may also aid in removing inter-user variability
because the parameters used for segmentations can be easily and repeatably applied
by different reading physicians, supporting manual as well as PERCIST-based
measurements.21
PERCIST recommends the use of reference ROIs positioned either in the right lobe of the
liver or the descending aorta, to determine reportable lesions or to quantify changes in
lesion uptake across timepoints. To manually draw and place an ROI of a particular size
in a consistent and appropriate location inside the liver or descending aorta is not a
difficult task for the clinician, but surely a tedious one, especially if PERCIST criteria is
used on all the patients.
ALPHA landmarking supports the automatic placement of reference ROIs in the liver and
in the descending aorta using the CT image. ALPHA detects multiple landmarks in and
around the liver and the descending aorta and uses all of them to infer and confirm the
final placement of the two ROIs. As a result, the reference ROI detectors are very
reliable despite changes in image contrast and highly robust to abnormalities such as
calcifications in the aorta or lesions in the liver. The landmarks are transferred to PET
image and ROIs are centered around these (see Figure 23).
Figure 23
Figure 23. Reference region

identified by ALPHA to calculate
the PERCIST threshold in blood
(a) (b) pool (a) and liver (b).
Deauville criteria
The Deauville scale is an internationally recommended scale used by physicians in
staging and assessment of treatment response in Hodgkin lymphoma (HL) and certain
types of non-Hodgkin lymphomas (NHL).23 This criterion provides a simple visual
assessment of FDG uptake in the hottest lesion relative to blood and liver reference
regions.
A Deauville score is calculated for each lesion as shown in Figure 23, and the score for
the hottest lesion can be assessed across timepoints in the trending tab, as shown in
Figure 24. The Deauville criteria21 are evaluated on a scale from 1 to 5, where 1 is the
best and 5 is the worst:
1 - No uptake
2 - Uptake ≤ mediastinum
3 - Uptake > mediastinum but ≤ liver
4 - Moderately increased uptake compared to liver
5 - Markedly increased uptake compared to liver and/or new lesions (Markedly
increased uptake is taken to be uptake > 2-3 times the SUVmax in a normal liver.)
Figure 24
Figure 24. 24 Deauville

score calculation for a
segmented lesion.
Data courtesy of Le Centre

Hospitalier Universitaire
Vaudois (CHUV), Lausanne,
Switzerland.
Figure 25
Figure 25
Figure 25. Trend of hottest

lesion by Deauville score
baseline has a score of 4, and,
at the follow-up, the score is 5.
Spine and rib labeling
The spine consists of 33 vertebrae (7 cervical, 12 thoracic, 5 lumbar, 5 fused sacrum,
4 fused coccyx), 23 intervertebral disks, the spinal cord, and connecting 24 ribs, making
it a complex anatomy. Hybrid technology (SPECT/CT or PET/CT) is a sensitive tool to
detect skeletal metastasis in known malignancies. However, its high sensitivity and
low specificity may account for false positive diagnosis in cases of trauma, infection,
inflammation, and other benign conditions. When detecting such skeletal metastasis,
careful review is required by a reading physician, necessitating accurate anatomical
correlation with findings on hybrid imaging.22 -25 For the spine and rib labelling,
ALPHA detects the center of each vertebral body and suggests its label (C1-C7 in
cervical, T1-T12 in thoracic region, L1-L5 lumbar region, S - sacrum) along with the
ribs (right 1-12, left 1-12) in CT datasets, as shown in Figure 26.
Figure 26
(a)
Figure 26. FDG PET/CT scan

demonstrating overlay of verte-
bra labels (a) and rib labels in
(b); label (C1-C7 in cervical,
T1-T12 in thoracic region, L1-L5
Lumbar region, S - sacrum) and
ribs (Right 1-12, left 1-12).
Data courtesy of
Le Centre Hospitalier
Universitaire Vaudois (CHUV),
(b) Lausanne, Switzerland.
Auto views
Increasing use of molecular imaging in non-oncology applications, eg, orthopedics,
requires re-orientation of CT and fused (PET/CT or SPECT/CT) datasets based on the
anatomy of interest or when data is prepared for a report. It may be required, for
example, to re-orient the patient anatomy to be the same as a baseline scan.
Additionally, for each patient anatomy of interest is re-oriented in a consistent manner
that can help to standardize the reading report. Auto Views functionality helps to
eliminate manual steps to re-orient images and zoom in on a required region. An ALPHA
gallery shows landmark points and body regions, along with the preview of how image
will be reoriented based on the user selection (Figure 27 (a)). When a user makes a
particular selection, the ALPHA landmarking algorithm runs to detect the landmark and
then re-orients image, as shown with an example in Figure 27 (b).
Figure 27
(a)
Figure 27. ALPHA gallery that

shows three presets Abdomen,
Kidney Right, and Kidney Left
with a preview on how image
will be re-orientated (a). Users
can add any of the regions/
landmarks in a list of their
favorites. Reorientation of hip
region on xSPECT image using
the Hip Acetabulum right
anatomical landmark point (b). (b)
Auto ranges
Radiologists have no need to review blank images, as described in the AIDAN platform
feature FAST PET Workflow AI. However, images could come from non-Siemens
Healthineers scanners, so this feature is also included in syngo.via. Auto Ranges uses
ALPHA to detect the skin-to-skin images and creates ranges automatically only with
anatomy present in the images, like the FAST PET Workflow AI feature on the PET/CT
scanner. The newly created ranges can be transfer to PACS. Figure 28 shows the
syngo.via application of creating parallel ranges that removes extra blank slices.
Figure 28
Figure 28. Parallel ranges

created in syngo.via shows
removal of unnecessary
blank slices.
Conclusion
ALPHA technology reliably provides anatomical context and has been used to
enrich applications on the AIDAN platform and syngo.via reading solutions. Scanner
applications help to improve the patient experience by reducing the scan time and
performing motion management without external devices. Automatic adjustment of
scan parameters and multi-parametric analysis based on the anatomy and disease
condition will open new opportunities for precision medicine. Applications in reading
solutions enable optimal workflow solutions for performing daily clinical analysis, such
as tracking lesion over time, placement of reference region for lesion segmentation
or Deauville score measurement, and re-orienting images in standardize format for
reporting purposes. Additionally, these applications are efficient to use, enabling
reproducible and standardized results. Rather than performing mundane tasks in the
workflow, results only need to be reviewed and verified before making any clinical
conclusions. This will increase efficiency of both technologist and nuclear medicine
readers and will allow them to focus on patients to improve patient experience and
transform care delivery.
Acknowledgments
The author thanks Dr. Bruce Spottiswoode, Dr. Sven Zuehlsdorff, Dr. Anne Smith,
Dr. Carl Von Gall, Dr. Gerardo Hermosillo Valadez, Dr. Zhigang Peng, Dr. Sebastian
Fuerst, Ms. Jessie Reed, Ms. Katherina Swystun, Ms. Colleen Smith, and Ms. Nicole
Trezciak for valuable discussions, contributions, and review of this document.
References
1 J. Czernin, I. Sonni, A. Razmaria, and J. Calais, “The Future of Nuclear Medicine as
an Independent Speciality,” JNM 60(2):3S-12S, Sep 1 2019
2 P. Ghosh, “xSPECT Bone: a clinical overview,” White Paper, Siemens Healthineers,
2018
3 X. S. Zhou, Y. Zhan, Z. Peng, M. Dewan, B. Jian, A. Krishnan, M. Harder, R.

Schwarz, L. Lauer, H. Meyer, S. Grosskopf, U. Feuerlein, and H. Ditt, “Reliability and
redundancy: reducing error cost in medical imaging”, in: B Krishnapuram, S. Yu, and R.
Rao (Eds.): Cost-Sensitive Machine Learning, Chapman and Hall/CRC, 2011
4 X. S. Zhou, Y. Zhan, V. C. Raykar, G. H. Valadez, L. Bogoni, Z. Peng, M. Dewan, M.

Wolf, Y. Shinagawa, S. Park, M. Salganicoff, L. Raghupathi, and P. Devarakota,
“Mining anatomical, physiological and pathological information from medical images.”
SIGKDD Explorations 14(1):25-34, 2012
5 Y. Zhan, M. Dewan, M. Harder, A. Krishnan, X. S. Zhou, “Robust automatic knee MR

slice positioning through redundant and hierarchical anatomy detection,” IEEE Trans.
Med. Imaging 30(12): 2087-2100, 2011
6 Y. Tao, Z. Peng, A. Krishnan, X. S. Zhou, “Robust learning-based parsing and

annotation of medical radiographs.” IEEE Trans. Med. Imaging 30(2): 338-350, 2011
7 W. Watanabe, A. Kitaoka, K. Sakamoto, M. Yasugi, and K. Tanaka, “Illusory Motion

Reproduced by Deep Neural Networks Trained for Prediction,” Front Psychol. 15 March
2018 (https://doi.org/10.3389/fpsyg.2018.00345)
8 “ Technical and Clinical Advances with FlowMotion Technology,” White paper, Siemens
Healthcare GmbH, 2016.
9 D. Pigg and B. Spottiswoode, “FlowMotion AI,” White paper, Siemens Healthineers,
2019
10 P. J. Keall, G. S. Mageras, J. M. Balter, R. S. Emery, K. M. Forster, S. B. Jiang, J. M.

Kapatoes, D. A. Low, J. M. Murphy, B. R. Murray, C. R. Ramsey, M. B. Van Herk, S.
S. Vedam, J. W. Wong, E. Yorke “The management of respiratory motion in radiation
oncology report of AAPM Task Group 76.” Med Phys 2006; 33:3874–3900
11 B
IO-TECH SYSTEMS, INC. Report 2008.
12 Grills, Inga S et al. “Potential for reduced toxicity and dose escalation in the
treatment of inoperable non–small-cell lung cancer: A comparison of intensity-
modulated radiation therapy (IMRT), 3D conformal radiation, and elective nodal
irradiation.” International Journal of Radiation Oncology, Biology, Physics, Volume
57, Issue 3, 875-890.
13 Garcia Vicente AM, et al. (18) F-FDG PET-CT respiratory gating in characterization of
pulmonary lesions: approximation towards clinical indications. Ann Nucl Med. 2010
April 24 (3) 207-14
14 J. Hamill, P. Ghosh, “HD Chest Respiratory Gating for PET,” White Paper, Siemens
Medical Solutions, USA, 2011
15 M. Dawood, F. Gigengack, X. Jiang, KP Schäfers, “A mass conservation-based
optical flow method for cardiac motion correction in 3D-PET.” Med Phys. 2013;40:1-9.
16 I Hong, J Jones, M Casey, “Ultrafast Elastic Motion Correction via Motion Deblurring.”
In; 2014 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/
MIC). IEEE; 2014:1-2
17 S Poesse, I Hong, D Mannweiler, et al. “Respiratory motion compensation in PET/CT:

Evaluation of a fast elastic motion compensation technique based on motion
deblurring”. J Nucl Med .2017;58(suppl 1):1349.
18 A . Smith, B. Spottiswoode, V. Shah, J. Hu, C. von Gall, “FlowMotion

Multiparametric PET Suite-The Patlak Model,” White Paper, Siemens Medical Solutions
USA, Inc. 2019.
19 D. Vriens, L. de Geus-Oei, W. Oyen, E. Visser, “A Curve-Fitting Approach to Estimate

the Arterial Plasma Input Function for the Assessment of Glucose Metabolic Rate and
Response to Treatment,” JNM 2009 Dec 50(12): 1933-9.
20 AR Pantel, DA Mankoff. “Molecular imaging to guide systemic cancer therapy:

Illustrative examples of PET imaging cancer biomarkers.” Cancer Lett. 2017;387:
25-31. doi:10.1016/j.canlet.2016.05.008
21 R. Wahl, H. Jacene, Y. Kasamon, M. Lodge, “From RECIST to PERCIST: Evolving

Considerations for PET Response Criteria in Solid Tumors” 2009 JNM: 50(Suppl 1),
122S-150S. https://dx.doi.org/10.2967/jnumed.108.057307
22 K Schmiedehausen, “NaF PET/CT in Prostate Cancer,” White Paper, Siemens Medical

Solutions USA, Inc., 2014
23 S. Barrington, R. Kluge, “FDG PET for therapy monitoring in Hodgkin and
non-Hodgkin lymphomas,” EJNMMI (2017) 44(S1): S97-S110.
24 V. Vassiliou, D. Andreopoulos, S Frangos , N. Tselis, E. Giannopoulou, S. Lutz.,

“Bone metastases: assessment of therapeutic response through radiological and
nuclear medicine imaging modalities” Clin Oncol (R Coll Radiol) 2011;23:632–645
25 A . Hans Vija, C. Von Gall, P. Ghosh, “Accurate, reproducible, and standardized

quantification,” White paper, Siemens Healthineers, 2018
Clinical Results F 18 Injection, USP
Fludeoxyglucose

Fludeoxyglucose
HIGHLIGHTS OF PRESCRIBING The recommended dose: and reversible loss of systolic function in patients with coronary artery disease and left
INFORMATION • for adults is 5 to 10 mCi (185 to 370 MBq), ventricular dysfunction, when used together with myocardial perfusion imaging.
These highlights do not include all the in all indicated clinical settings (2.1). 1.3 Neurology
information
HIGHLIGHTS OFneeded
PRESCRIBING to use Fludeoxyglu- The• recommended
for pediatric patients dose: is 2.6 mCi in the neu- andFor the identification
reversible of regions
loss of systolic of abnormal
function in patients glucose metabolism
with coronary associated
artery diseasewithand foci
left of
cose F 18 Injection safely and effectively.
INFORMATION • forrology
adults setting
is 5 to 10 (2.2).
mCi (185 to 370 MBq), epileptic dysfunction,
ventricular seizures. when used together with myocardial perfusion imaging.
See highlights
These full prescribing do not information
include all for theFlude- Initiate
in imaging
all indicated withinsettings
clinical 40 minutes
(2.1).following 1.32 Neurology
DOSAGE AND ADMINISTRATION
oxyglucose
information F 18 Injection.
needed to use Fludeoxyglu- drug
• for injection;
pediatric acquire
patients is static emission
2.6 mCi images
in the neu- ForFludeoxyglucose
the identificationF of 18regions
Injection emits radiation.
of abnormal glucoseUse procedures
metabolism to minimize
associated with radiation
foci of
Fludeoxyglucose
cose F 18 Injection safely F 18 Injection, USP
and effectively. 30 to 100
rology minutes
setting (2.2).from time of injection (2). exposure.
epileptic Calculate the final dose from the end of synthesis (EOS) time using proper
seizures.
SeeFor intravenous
full prescribinguse information for Flude- Initiate DOSAGE
imaging within FORMS40AND STRENGTHS
minutes following 2 radioactive
DOSAGE AND decay factors. Assay the final dose in a properly calibrated dose calibrator
ADMINISTRATION
Initial U.S.F Approval:
oxyglucose 18 Injection. 2005 drugMulti-dose
injection;30mL acquire and 50mL
static glass vial
emission conta-
images before administration
Fludeoxyglucose to the patient
F 18 Injection [see Description
emits radiation. (11.2)]. to minimize radiation
Use procedures
Fludeoxyglucose INDICATIONS AND USAGE
F 18 Injection, USP 30ining
to 1000.74 minutesto 7.40from GBq/mL
time of(20 to 200
injection mCi/
(2). 2.1exposure.
Recommended CalculateDose for Adults
the final dose from the end of synthesis (EOS) time using proper
ForFludeoxyglucose
intravenous useF 18 Injection is indicated mL)DOSAGEFludeoxyglucose
FORMS ANDF STRENGTHS 18 Injection and Within the oncology, cardiology and neurology
radioactive decay factors. Assay the final dose in a properly calibrated settings, the recommended dose for
dose calibrator
for positron emission
Initial U.S. Approval: 2005 tomography (PET) ima- 4.5mg of sodium chloride
Multi-dose 30mL and 50mL glass vial conta- with 0.1 to 0.5% adults
before is 5 to 10 mCito(185
administration to 370 MBq)
the patient as an intravenous
[see Description (11.2)].injection.
ging inINDICATIONS
the following AND settings:
USAGE w/w0.74
ining ethanol
to 7.40 as a GBq/mL
stabilizer(20
(approximately
to 200 mCi/15 2.12.2Recommended
RecommendedDose Dosefor forAdults
Pediatric Patients
• Oncology: ForF assessment
Fludeoxyglucose 18 Injectionofisabnormal indicated glu- mL) to 50 mL volume) for Fintravenous
Fludeoxyglucose 18 Injectionadministra-
and Within
Within thethe neurology
oncology, setting, the
cardiology andrecommended
neurology settings,dose for thepediatric patients dose
recommended is 2.6for
mCi,
cose metabolism
for positron to assist in the
emission tomography evaluation
(PET) ima- tion (3).
4.5mg of sodium chloride with 0.1 to 0.5% as an
adults is 5intravenous
to 10 mCi (185 injection.
to 370 TheMBq)optimal
as andose adjustment
intravenous on the basis of body size or
injection.
ging of malignancy
in the followinginsettings:patients with known or w/w ethanol as CONTRAINDICATIONS
a stabilizer (approximately 15 weight has not
2.2 Recommended Dose been
fordetermined [see Use in Special Populations (8.4)].
Pediatric Patients
suspected
• Oncology: For abnormalities
assessment of abnormal found by glu- other None.
to 50 mL volume) for intravenous administra- 2.3Within
Patient
the Preparation
neurology setting, the recommended dose for pediatric patients is 2.6 mCi,
cose testing modalities,
metabolism to assistor in patients
in the with an
evaluation tion (3). WARNINGS AND PRECAUTIONS • To
as an minimize
intravenous the injection.
radiation absorbed
The optimal dose to the
dose bladder, encourage
adjustment on the basis adequate
of bodyhydration.
size or
of existing
malignancydiagnosis of cancer.
in patients with known or • Radiation risks: use smallest dose neces-
CONTRAINDICATIONS Encourage
weight has notthe patient
been to drink[see
determined water Useorinother fluids
Special (as tolerated)
Populations in the 4 hours before
(8.4)].
• Cardiology:
suspected For the identification
abnormalities found by other of left None. sary for imaging (5.1). their PET
2.3 Patient study.
Preparation
ventricular myocardium
testing modalities, or in patients with residual
with an glu- • Blood
WARNINGSglucoseAND adnormalities:
PRECAUTIONS may cause • To• Encourage
minimize the theradiation
patient to void as soon
absorbed dose asto the imaging
bladder, study
encourageis completed
adequate and as often as
hydration.
cose metabolism
existing diagnosis ofand reversible loss of sys-
cancer. suboptimal
• Radiation risks: imaging (5.2). dose neces-
use smallest possible the
Encourage thereafter
patientfor to at least
drink one or
water hour.
other fluids (as tolerated) in the 4 hours before
tolic function
• Cardiology: For in thepatients with coronary
identification of left ar- ADVERSE
sary for imaging (5.1). REACTIONS • Screen
their patients for clinically significant blood glucose abnormalities by obtaining a his-
PET study.
tery disease
ventricular myocardiumand left withventricular
residual glu- dys- Hypersensitivity
• Blood reactions have
glucose adnormalities: may occurred;
cause tory and/or [see Warnings and Precautions (5.2)]. Prior
• Encourage thelaboratory
patient to void testsas soon as the imaging study is completed andtoas
Fludeoxyglu-
often as
cose function,
metabolism whenand used together
reversible losswith myo-
of sys- have emergency
suboptimal imaging resuscitation
(5.2). equipment cose Fthereafter
possible 18 PET imaging in the
for at least oneoncology
hour. and neurology settings, instruct patient to fast
toliccardial perfusion
function imaging.
in patients with coronary ar- and personnel immediately available (6).
ADVERSE REACTIONS for 4patients
• Screen to 6 hours forprior to the
clinically drug’s injection.
significant blood glucose abnormalities by obtaining a his-
• Neurology:
tery disease For andtheleft identification
ventricular of regions
dys- To report SUSPECTED ADVERSE
Hypersensitivity reactions have occurred; • In and/or
tory the cardiology
laboratorysetting,
testsadministration
[see Warningsofand glucose-containing
Precautions (5.2)]. foodPrior
or liquids (e.g., 50 to 75
to Fludeoxyglu-
of abnormal
function, when glucose metabolism
used together with associat-
myo- REACTIONS, contact PETNET Solutions, Inc.
have emergency resuscitation equipment cosegrams)
F 18 prior to Fludeoxyglucose
PET imaging in the oncologyF 18 Injection facilitates
and neurology localization
settings, of cardiac
instruct ischemia.
patient to fast
ed with foci of epileptic seizures (1).
cardial perfusion imaging. andatpersonnel
877-473-8638 or FDA available
immediately at 1-800-FDA-1088
(6). 2.4forRadiation
4 to 6 hours Dosimetry
prior to the drug’s injection.
• Neurology:DOSAGE For the AND ADMINISTRATION
identification of regions To or www.fda.gov/medwatch.
report SUSPECTED ADVERSE Thecardiology
• In the estimated human
setting, absorbed radiation
administration doses (rem/mCi)
of glucose-containing food ortoliquids
a newborn
(e.g., 50(3.4
to 75kg),
Fludeoxyglucose
of abnormal glucose F 18 metabolism
Injection emits radiation.
associat- REACTIONS, USEcontact
IN SPECIFIC PETNET POPULATIONS
Solutions, Inc. 1-year
grams) prioroldto(9.8 kg), 5-year old
Fludeoxyglucose F 18(19 kg), 10-year
Injection oldlocalization
facilitates (32 kg), 15-yearof cardiacoldischemia.
(57 kg), and
Usewith
ed procedures to minimize
foci of epileptic seizures radiation
(1). expo- • Lactation: Temporarily
at 877-473-8638 discontinue breast-
or FDA at 1-800-FDA-1088 adult (70
2.4 Radiation Dosimetry kg) from intravenous administration of Fludeoxyglucose F 18 Injection are
sure. Screen for
DOSAGE AND blood glucose abnormalities.
ADMINISTRATION feeding. A
or www.fda.gov/medwatch. lactating woman should pump
Theshown in Table 1. These estimates were calculated based on to human data(3.4
andkg),
using
2
• In the oncology estimated human absorbed radiation doses (rem/mCi) a newborn
Fludeoxyglucose F 18 and neurology
Injection settings, in-
emits radiation. andUSEdiscard
IN SPECIFIC breastmilk for 9 hours after
POPULATIONS the data published by theoldInternational Commission on Radiological kg), andfor
Protection 4
struct patients 1-year old (9.8 kg), 5-year (19 kg), 10-year old (32 kg), 15-year old (57
Use procedures to to fast for radiation
minimize 4 to 6 hours prior
expo- Fludeoxyglucose
• Lactation: Temporarily F 18discontinue
Injection (8.2).
breast- Fludeoxyglucose F. The dosimetry data show that there are slight F 18variations
Injection inareab-
18
adult (70 kg) from intravenous administration of Fludeoxyglucose
sure. to the for
Screen drug’s
blood injection. Consider medical
glucose abnormalities. • Pediatric
feeding. Use: Safety
A lactating womanand should
effectiveness
pump in sorbed radiation dose for variouswere organs in each of the age groups. 2 These dissimilari-
therapy and laboratory testing to assure shown in Table 1. These estimates calculated based on human data and using
• In the oncology and neurology settings, in- at andpediatric
discard patients
breastmilk have
for not been after
9 hours estab-
leastpatients
struct two days of normoglycemia
to fast for 4 to 6 hours prior prior to lished in the oncology and cardiology set- theties
datain published
absorbed radiation dose are dueCommission
by the International to developmental age variations
on Radiological (e.g.,4 organ
Protection for
Fludeoxyglucose F 18 Injection (8.2). size, location, and
the drug’s administration
to the drug’s injection. Consider medical (5.2). tings (8.4). Fludeoxyglucose 18
F. overall metabolic
The dosimetry datarate
showfor that
eachthere
age group).
are slight Thevariations
identifiedincritical
ab-
• Pediatric Use: Safety and effectiveness in organs (in descending order) across all age groups evaluated are the urinary bladder,
• In theand
therapy cardiology
laboratory setting,
testing administration
to assure at of See 17 forpatientsPATIENThave COUNSELING sorbed radiation dose for various organs in each of the age groups. These dissimilari-
pediatric not been estab-
leastglucose-containing
two days of normoglycemiafood or liquidsprior (e.g.,
to 50 INFORMATION tiesheart, pancreas,
in absorbed spleen,dose
radiation and lungs.
are due to developmental age variations (e.g., organ
lished in the oncology and cardiology set-
thetodrug’s
75 grams) prior to the
administration drug’s injection fa-
(5.2). Revised: 10/2019 size, location, and overall metabolic rate for each age group). The identified critical
tings (8.4).
• In cilitates localization
the cardiology of cardiac
setting, ischemia (2.3).
administration of organs
Table (in descending
1. Estimated order)Radiation
Absorbed across all Doses
age groups evaluated
(rem/mCi) Afterare the urinary bladder,
Intravenous
See 17 for PATIENT COUNSELING
Aseptically withdraw
glucose-containing foodFludeoxyglucose
or liquids (e.g.,F 50 18 heart, pancreas,
Administration ofspleen, and lungs. F 18 Injectiona
Fludeoxyglucose
INFORMATION
Injection
to 75 grams) from its container
prior to the drug’s andinjection
administer fa- Revised: 10/2019 Newborn 1-year old 5-year old 10-year old 15-year old Adult
by intravenous
cilitates localization injection
of cardiac(2). ischemia (2.3). Organ
Table 1. Estimated Absorbed
(3.4 kg) Radiation Doses(19
(9.8 kg) (rem/mCi)
kg) After Intravenous
(32 kg) (57 kg) (70 kg)
Aseptically withdraw Fludeoxyglucose F 18 Administrationb of Fludeoxyglucose F 18 Injectiona
Injection from its container and administer Bladder wall 4.3 1.7 0.93 0.60 0.40 0.32
Newborn 1-year old 5-year old 10-year old 15-year old Adult
by FULL PRESCRIBING
intravenous INFORMATION:
injection (2). CONTENTS* 8.2 Lactation Organ
Heart wall 2.4 1.2 0.70 0.44 0.29 0.22
1 INDICATIONS AND USAGE 8.4 Pediatric Use (3.4 kg) (9.8 kg) (19 kg) (32 kg) (57 kg) (70 kg)
Pancreas 2.2 0.68 0.33 0.25 0.13 0.096
1.1 Oncology 11 DESCRIPTION Bladder wallb 4.3 1.7 0.93 0.60 0.40 0.32
1.2 Cardiology
FULL PRESCRIBING INFORMATION: CONTENTS* 8.211.1 Chemical Characteristics
Lactation Spleen 2.2 0.84 0.46 0.29 0.19 0.14
Heart wall 2.4 1.2 0.70 0.44 0.29 0.22
1.3 Neurology
1 INDICATIONS AND USAGE 8.411.2 Physical
Pediatric Characteristics
Use Lungs 0.96 0.38 0.20 0.13 0.092 0.064
21.1DOSAGE AND ADMINISTRATION Pancreas 2.2 0.68 0.33 0.25 0.13 0.096
Oncology 1112 CLINICAL PHARMACOLOGY
DESCRIPTION
Kidneys 0.81 0.34 0.19 0.13 0.089 0.074
1.22.1Cardiology
Recommended Dose for Adults 12.1 Mechanism of Action
11.1 Chemical Characteristics Spleen 2.2 0.84 0.46 0.29 0.19 0.14
2.2 Recommended
1.3 Neurology Dose for 11.212.2 Pharmacodynamics
Physical Characteristics Ovaries 0.80 0.8 0.19 0.11 0.058 0.053
Lungs 0.96 0.38 0.20 0.13 0.092 0.064
2 DOSAGEPediatric Patients
AND ADMINISTRATION 12.3 Pharmacokinetics
12 CLINICAL PHARMACOLOGY Uterus 0.79 0.35 0.19 0.12 0.076 0.062
2.12.3Recommended
Patient Preparation
Dose for Adults 13
12.1NONCLINICAL
Mechanism of TOXICOLOGY
Action Kidneys 0.81 0.34 0.19 0.13 0.089 0.074
2.22.4Recommended
Radiation Dosimetry
Dose for 12.213.1 Carcinogenesis, Muta-genesis,
Pharmacodynamics LLI wall * 0.69 0.28 0.15 0.097 0.060 0.051
Ovaries 0.80 0.8 0.19 0.11 0.058 0.053
2.5Pediatric
Radiation Safety – Drug Handling
Patients Impairment of Fertility
12.3 Pharmacokinetics Liver 0.69 0.31 0.17 0.11 0.076 0.058
Uterus 0.79 0.35 0.19 0.12 0.076 0.062
2.32.6Patient
DrugPreparation
Preparation and Administration 1314 CLINICAL STUDIES
NONCLINICAL TOXICOLOGY Gallbladder wall 0.69 0.26 0.14 0.093 0.059 0.049
2.42.7Radiation
ImagingDosimetry
Guidelines 13.114.1 Oncology
Carcinogenesis, Muta-genesis, LLI wall * 0.69 0.28 0.15 0.097 0.060 0.051
32.5DOSAGE FORMS
Radiation Safety AND STRENGTHS
– Drug Handling 14.2 Cardiologyof Fertility
Impairment Small intestine 0.68 0.29 0.15 0.096 0.060 0.047
Liver 0.69 0.31 0.17 0.11 0.076 0.058
42.6CONTRAINDICATIONS
Drug Preparation and Administration 14.3 Neurology
14 CLINICAL STUDIES ULI wall ** 0.67 0.27 0.15 0.090 0.057 0.046
52.7WARNINGS AND PRECAUTIONS 16 HOW SUPPLIED/STORAGE AND DRUG Gallbladder wall 0.69 0.26 0.14 0.093 0.059 0.049
Imaging Guidelines 14.1 Oncology
5.1 Radiation RisksSTRENGTHS HANDLING Stomach wall 0.65 0.27 0.14 0.089 0.057 0.047
3 DOSAGE FORMS AND 14.2 Cardiology Small intestine 0.68 0.29 0.15 0.096 0.060 0.047
5.2 Blood Glucose Abnormalities
4 CONTRAINDICATIONS 17 PATIENT
14.3 COUNSELING INFORMATION
Neurology Adrenals 0.65 0.28 0.15 0.095 0.061 0.048
ULI wall ** 0.67 0.27 0.15 0.090 0.057 0.046
5 6WARNINGS
ADVERSEAND REACTIONS
PRECAUTIONS 16*HOW
Sections or subsections omitted
SUPPLIED/STORAGE from the
AND DRUG Testes 0.64 0.27 0.14 0.085 0.052 0.041
75.1DRUG INTERACTIONS
Radiation Risks full prescribing information are not listed.
HANDLING Stomach wall 0.65 0.27 0.14 0.089 0.057 0.047
85.2USE IN SPECIFIC
Blood Glucose POPULATIONS
Abnormalities 17 PATIENT COUNSELING INFORMATION Red marrow 0.62 0.26 0.14 0.089 0.057 0.047
Adrenals 0.65 0.28 0.15 0.095 0.061 0.048
8.1 Pregnancy
6 ADVERSE REACTIONS * Sections or subsections omitted from the Thymus 0.61 0.26 0.14 0.086 0.056 0.044
Testes 0.64 0.27 0.14 0.085 0.052 0.041
7 FULL
DRUGPRESCRIBING
INTERACTIONS INFORMATION full prescribing information are not listed. Thyroid 0.61 0.26 0.13 0.080 0.049 0.039
8 1USE IN SPECIFIC POPULATIONS Red marrow 0.62 0.26 0.14 0.089 0.057 0.047
INDICATIONS AND USAGE
8.1 Fludeoxyglucose
Pregnancy Muscle 0.58 0.25 0.13 0.078 0.049 0.039
F 18 Injection is indicated for positron emission tomography (PET) ima- Thymus 0.61 0.26 0.14 0.086 0.056 0.044
ging in the following
FULL PRESCRIBING INFORMATIONsettings: Bone surface 0.57 0.24 0.12 0.079 0.052 0.041
Thyroid 0.61 0.26 0.13 0.080 0.049 0.039
1 1.1INDICATIONS
Oncology AND USAGE Breast 0.54 0.22 0.11 0.068 0.043 0.034
For assessmentFof Muscle 0.58 0.25 0.13 0.078 0.049 0.039
Fludeoxyglucose 18abnormal
Injection glucose metabolism
is indicated to assist
for positron in thetomography
emission evaluation of(PET)
malignancy
ima- Skin 0.49 0.20 0.10 0.060 0.037 0.030
in in
ging patients with known
the following or suspected abnormalities found by other testing modalities, or in
settings: Bone surface 0.57 0.24 0.12 0.079 0.052 0.041
patients
1.1 Oncology with an existing diagnosis of cancer. Brain 0.29 0.13 0.09 0.078 0.072 0.070
Breast 0.54 0.22 0.11 0.068 0.043 0.034
1.2ForCardiology
assessment of abnormal glucose metabolism to assist in the evaluation of malignancy Other tissues 0.59 0.25 0.13 0.083 0.052 0.042
For the with
identification Skin 0.49 0.20 0.10 0.060 0.037 0.030
in patients known orof left ventricular
suspected myocardium
abnormalities with
found by residual
other testingglucose metabolism
modalities, or in a
MIRDOSE 2 software was used to calculate the radiation absorbed dose.
patients with an existing diagnosis of cancer. Brain
b
The dynamic bladder 0.29 0.13 voiding0.09
model with a uniform frequency of 0.078 0.072
1.5 hours was used. 0.070
1.2 Cardiology * LLI tissues
Other = lower large intestine;
0.59 ** ULI = upper
0.25 large intestine
0.13 0.083 0.052 0.042
For the identification of left ventricular myocardium with residual glucose metabolism a
MIRDOSE 2 software was used to calculate the radiation absorbed dose.
b
The dynamic bladder model with a uniform voiding frequency of 1.5 hours was used.
* LLI = lower large intestine; ** ULI = upper large intestine
Clinical
Fludeoxyglucose F 18 Results
Injection, USP
2.5 Radiation Safety – Drug Handling Clinical Considerations

• Use waterproof gloves, effective radiation shielding, and appropriate safety measures To decrease radiation exposure to the breastfed infant, advise a lactating woman to pump
when handling Fludeoxyglucose F 18 Injection to avoid unnecessary radiation exposure to and discard breastmilk and avoid close (breast) contact with the infant for at least 9 hours
the patient, occupational workers, clinical personnel and other persons. after the administration of Fludeoxyglucose F 18 Injection.
• Radiopharmaceuticals should be used by or under the control of physicians who are quali- 8.4 Pediatric Use
fied by specific training and experience in the safe use and handling of radionuclides, and The safety and effectiveness of Fludeoxyglucose F 18 Injection in pediatric patients with
whose experience and training have been approved by the appropriate governmental epilepsy is established on the basis of studies in adult and pediatric patients. In pediatric
patients with epilepsy, the recommended dose is 2.6 mCi. The optimal dose adjustment
agency authorized to license the use of radionuclides.
on the basis of body size or weight has not been determined. In the oncology or cardiolo-
• Calculate the final dose from the end of synthesis (EOS) time using proper radioactive de-
gy settings, the safety and effectiveness of Fludeoxyglucose F 18 Injection have not been
cay factors. Assay the final dose in a properly calibrated dose calibrator before administra-
established in pediatric patients.
tion to the patient [see Description (11.2)]. 11 DESCRIPTION
• The dose of Fludeoxyglucose F 18 used in a given patient should be minimized consistent 11.1 Chemical Characteristics
with the objectives of the procedure, and the nature of the radiation detection devices Fludeoxyglucose F 18 Injection is a positron emitting radiopharmaceutical that is used for
employed. diagnostic purposes in conjunction with positron emission tomography (PET) imaging.
2.6 Drug Preparation and Administration The active ingredient 2-deoxy-2-[18F]fluoro-D-glucose has the molecular formula of C6H-
• Calculate the necessary volume to administer based on calibration time and dose. 1118FO5 with a molecular weight of 181.26, and has the following chemical structure:
• Aseptically withdraw Fludeoxyglucose F 18 Injection from its container.
• Inspect Fludeoxyglucose F 18 Injection visually for particulate matter and discoloration
before administration, whenever solution and container permit.
• Do not administer the drug if it contains particulate matter or discoloration; dispose of
these unacceptable or unused preparations in a safe manner, in compliance with applica-
ble regulations.
• Use Fludeoxyglucose F 18 Injection within 12 hours from the EOS. Fludeoxyglucose F 18 Injection is provided as a ready to use sterile, pyrogen free, clear,
2.7 Imaging Guidelines colorless solution. Each mL contains between 0.740 to 7.40GBq (20.0 to 200 mCi) of
2-deoxy-2-[18F]fluoro-D-glucose at the EOS, 4.5 mg of sodium chloride and 0.1 to 0.5%
• Initiate imaging within 40 minutes following Fludeoxyglucose F 18 Injection administration.
w/w ethanol as a stabilizer. The pH of the solution is between 4.5 and 7.5. The solution is
• Acquire static emission images 30 to 100 minutes from the time of injection.
packaged in a multiple-dose glass vial and does not contain any preservative.
3 DOSAGE FORMS AND STRENGTHS
11.2 Physical Characteristics
Multiple-dose 30 mL and 50 mL glass vial containing 0.74 to 7.40 GBq/mL (20 to 200 mCi/ Fluorine F 18 has a physical half-life of 109.7 minutes and decays to Oxygen O 16 (stable)
mL) of Fludeoxyglucose F 18 Injection and 4.5 mg of sodium chloride with 0.1 to 0.5% by positron decay. The principal photons useful for imaging are the dual 511 keV “annihila-
w/w ethanol as a stabilizer (approximately 15 to 50 mL volume) for intravenous admini- tion” gamma photons, that are produced and emitted simultaneously in opposite direction
stration. when the positron interacts with an electron (Table 2).
4 CONTRAINDICATIONS
None.
Table 2. Principal Radiation Emission Data for Fluorine F 18
5 WARNINGS AND PRECAUTIONS
5.1 Radiation Risks Radiation/Emission % Per Disintegration Mean Energy
Radiation-emitting products, including Fludeoxyglucose F 18 Injection, may increase the Positron (β+) 96.73 249.8 keV
risk for cancer, especially in pediatric patients. Use the smallest dose necessary for ima-
Gamma (±)* 193.46 511.0 keV
ging and ensure safe handling to protect the patient and health care worker [see Dosage
and Administration (2.5)]. *Produced by positron annihilation
5.2 Blood Glucose Abnormalities From: Kocher, D.C. Radioactive Decay Tables DOE/TIC-I 1026, 89 (1981)
In the oncology and neurology setting, suboptimal imaging may occur in patients with in-
adequately regulated blood glucose levels. In these patients, consider medical therapy The specific gamma ray constant (point source air kerma coefficient) for fluorine F 18 is 5.7
and laboratory testing to assure at least two days of normoglycemia prior to Fludeoxyglu- R/hr/mCi (1.35 x 10-6 Gy/hr/kBq) at 1 cm. The half-value layer (HVL) for the 511 keV photons is
cose F 18 Injection administration. 4 mm lead (Pb). The range of attenuation coefficients for this radionuclide as a function of lead
6 ADVERSE REACTIONS shield thickness is shown in Table 3. For example, the interposition of an 8 mm thickness of Pb,
Hypersensitivity reactions with pruritus, edema and rash have been reported in the with a coefficient of attenuation of 0.25, will decrease the external radiation by 75%.
post-marketing setting. Have emergency resuscitation equipment and personnel imme-
diately available. Table 3. Radiation Attenuation of 511 keV Photons by lead (Pb) shielding
7 DRUG INTERACTIONS
Shield thickness (Pb) mm Coefficient of attenuation
The interactions of Fludeoxyglucose F 18 Injection with other drugs taken by patients un-
dergoing PET imaging has not been studied. 0 0.00
8 USE IN SPECIFIC POPULATIONS 4 0.50
8.1 Pregnancy
Risk Summary 8 0.25
Data from published case series and case reports describe Fludeoxyglucose F 18 Injection 13 0.10
crossing the placenta with uptake by the fetus (see Data). All radiopharmaceuticals have
26 0.01
the potential to cause fetal harm depending on the fetal stage of development and the
magnitude of the radiation dose. However, published studies that describe Fludeoxyglu- 39 0.001
cose F 18 Injection use in pregnant women have not identified a risk of drug-associated 52 0.0001
major birth defects, miscarriage, or adverse maternal or fetal outcomes. If considering
Fludeoxyglucose F 18 Injection administration to a pregnant woman, inform the patient For use in correcting for physical decay of this radionuclide, the fractions remaining at selected
about the potential for adverse pregnancy outcomes based on the radiation dose from intervals after calibration are shown in Table 4.
Fludeoxyglucose F 18 Injection and the gestational timing of exposure. The estimated
background risk of major birth defects and miscarriage for the indicated population is
Table 4. Physical Decay Chart for Fluorine F 18
unknown. All pregnancies have a background risk of birth defect, loss, or other adverse
outcomes. In the U.S. general population, the estimated background risk of major birth Minutes Fraction Remaining
defects and miscarriage in clinically recognized pregnancies are 2-4% and 15-20%, respectively. 0* 1.000
Data
Human Data 15 0.909
Data from published case series and case reports describe Fludeoxyglucose F 18 Injection 30 0.826
crossing the placental barrier and visualization of radioactivity throughout the body of the
60 0.683
fetus. The estimated fetal absorbed radiation dose from the maximum labeled dose (370
MBq) of Fludeoxyglucose F 18 was 10 mGy with first trimester exposure to PET alone and 110 0.500
20 mGy with first trimester exposure to PET/CT scan combination. Long-term adverse radi- 220 0.250
ation effects to a child exposed to Fludeoyxglucose F 18 Injection in utero are unknown. No
adverse fetal effects or radiation-related risks have been identified for diagnostic procedures *calibration time
involving less than 50 mGy, which represents less than 20 mGy fetal doses.
8.2 Lactation 12 CLINICAL PHARMACOLOGY
Risk Summary 12.1 Mechanism of Action
A published case report and case series show the presence of Fludeoxyglucose F 18 Injec- Fludeoxyglucose F 18 is a glucose analog that concentrates in cells that rely upon glucose
tion in human milk following administration. There are no data on the effects of Fludeoxy- as an energy source, or in cells whose dependence on glucose increases under pathophy-
glucose F 18 Injection on the breastfed infant or the effects on milk production. Exposure siological conditions. Fludeoxyglucose F 18 is transported through the cell membrane by
of Fludeoxyglucose F 18 Injection to a breastfed infant can be minimized by temporary facilitative glucose transporter proteins and is phosphorylated within the cell to [18F] FDG-
discontinuation of breastfeeding (see Clinical Considerations). The developmental and he- 6-phosphate by the enzyme hexokinase. Once phosphorylated it cannot exit until it is de-
alth benefits of breastfeeding should be considered along with the mother’s clinical need phosphorylated by glucose-6-phosphatase. Therefore, within a given tissue or pathophy-
for Fludeoxyglucose F 18 Injection, any potential adverse effects on the breastfed child siological process, the retention and clearance of Fludeoxyglucose F 18 reflect a balance
from Fludeoxyglucose F 18 Injection or from the underlying maternal condition. involving glucose transporter, hexokinase and glucose-6-phosphatase activities. F 18 is
used to assess glucose metabolism.
Fludeoxyglucose
In comparison to background activity of the specific organ or tissue type, regions of 14.2 Cardiology
decreased or absent uptake of Fludeoxyglucose F 18 reflect the decrease or absence of The efficacy of Fludeoxyglucose F 18 Injection for cardiac use was demonstrated in ten
glucose metabolism. Regions of increased uptake of Fludeoxyglucose F 18 reflect greater independent, prospective studies of patients with coronary artery disease and chronic
than normal rates of glucose metabolism. left ventricular systolic dysfunction who were scheduled to undergo coronary revascula-
12.2 Pharmacodynamics rization. Before revascularization, patients underwent PET imaging with Fludeoxyglu-
Fludeoxyglucose F 18 Injection is rapidly distributed to all organs of the body after intravenous cose F 18 Injection (74 to 370 MBq, 2 to 10 mCi) and perfusion imaging with other dia-
administration. After background clearance of Fludeoxyglucose F 18 Injection, optimal PET gnostic radiopharmaceuticals. Doses of Fludeoxyglucose F 18 Injection ranged from 74
imaging is generally achieved between 30 to 40 minutes after administration. to 370 MBq (2 to 10 mCi). Segmental, left ventricular, wall-motion assessments of
In cancer, the cells are generally characterized by enhanced glucose metabolism partially asynergic areas made before revascularization were compared in a blinded manner to
due to (1) an increase in activity of glucose transporters, (2) an increased rate of phospho- assessments made after successful revascularization to identify myocardial segments
rylation activity, (3) a reduction of phosphatase activity or, (4) a dynamic alteration in the with functional recovery.
balance among all these processes. However, glucose metabolism of cancer as reflected Left ventricular myocardial segments were predicted to have reversible loss of systolic
by Fludeoxyglucose F 18 accumulation shows considerable variability. Depending on tu- function if they showed Fludeoxyglucose F 18 accumulation and reduced perfusion
mor type, stage, and location, Fludeoxyglucose F 18 accumulation may be increased, nor- (i.e., flow-metabolism mismatch). Conversely, myocardial segments were predicted to
mal, or decreased. Also, inflammatory cells can have the same variability of uptake of Flu- have irreversible loss of systolic function if they showed reductions in both Fludeoxyglu-
deoxyglucose F 18. cose F 18 accumulation and perfusion (i.e., matched defects).
In the heart, under normal aerobic conditions, the myocardium meets the bulk of its ener- Findings of flow-metabolism mismatch in a myocardial segment may suggest that suc-
gy requirements by oxidizing free fatty acids. Most of the exogenous glucose taken up by cessful revascularization will restore myocardial function in that segment. However,
the myocyte is converted into glycogen. However, under ischemic conditions, the oxidati- false-positive tests occur regularly, and the decision to have a patient undergo revascu-
on of free fatty acids decreases, exogenous glucose becomes the preferred myocardial larization should not be based on PET findings alone. Similarly, findings of a matched
substrate, glycolysis is stimulated, and glucose taken up by the myocyte is metabolized defect in a myocardial segment may suggest that myocardial function will not recover in
immediately instead of being converted into glycogen. Under these conditions, phospho- that segment, even if it is successfully revascularized. However, false-negative tests oc-
rylated Fludeoxyglucose F 18 accumulates in the myocyte and can be detected with PET cur regularly, and the decision to recommend against coronary revascularization, or to
imaging. recommend a cardiac transplant, should not be based on PET findings alone. The rever-
In the brain, cells normally rely on aerobic metabolism. In epilepsy, the glucose metabo- sibility of segmental dysfunction as predicted with Fludeoxyglucose F 18 PET imaging
lism varies. Generally, during a seizure, glucose metabolism increases. Interictally, the sei- depends on successful coronary revascularization. Therefore, in patients with a low like-
zure focus tends to be hypometabolic. lihood of successful revascularization, the diagnostic usefulness of PET imaging with
12.3 Pharmacokinetics Fludeoxyglucose F 18 Injection is more limited.
Distribution: In four healthy male volunteers, receiving an intravenous administration of 14.3 Neurology
30 seconds induration, the arterial blood level profile for Fludeoxyglucose F 18 decayed In a prospective, open label trial, Fludeoxyglucose F 18 Injection was evaluated in 86
triexponentially. The effective half-life ranges of the three phases were 0.2 to 0.3 minutes, patients with epilepsy. Each patient received a dose of Fludeoxyglucose F 18 Injection in
10 to 13 minutes with a mean and standard deviation (STD) of 11.6 (±) 1.1 min, and 80 to the range of 185 to 370 MBq (5 to 10 mCi). The mean age was 16.4 years (range: 4
95 minutes with a mean and STD of 88 (±) 4 min. months to 58 years; of these, 42 patients were less than 12 years and 16 patients were
Plasma protein binding of Fludeoxyglucose F 18 has not been studied. less than 2 years old). Patients had a known diagnosis of complex partial epilepsy and
Metabolism: Fludeoxyglucose F 18 is transported into cells and phosphorylated to [18F]- were under evaluation for surgical treatment of their seizure disorder. Seizure foci had
FDG-6-phosphate at a rate proportional to the rate of glucose utilization within that tis- been previously identified on ictal EEGs and sphenoidal EEGs. Fludeoxyglucose F 18 In-
sue. [18F]-FDG-6-phosphate presumably is metabolized to 2-deoxy-2-[18F]fluoro-6-phos- jection PET imaging confirmed previous diagnostic findings in 16% (14/87) of the pati-
pho-D-mannose([18F]FDM-6-phosphate). ents; in 34% (30/87) of the patients, Fludeoxyglucose F 18 Injection PET images provi-
Fludeoxyglucose F 18 Injection may contain several impurities (e.g., 2-deoxy-2-chloro-D- ded new findings. In 32% (27/87), imaging with Fludeoxyglucose F 18 Injection was
glucose (ClDG)). Biodistribution and metabolism of ClDG are presumed to be similar to inconclusive. The impact of these imaging findings on clinical outcomes is not known.
Fludeoxyglucose F 18 and would be expected to result in intracellular formation of 2-de- Several other studies comparing imaging with Fludeoxyglucose F 18 Injection results to
oxy-2-chloro-6-phospho-D-glucose (ClDG-6-phosphate) and 2-deoxy-2-chloro-6-phospho- subsphenoidal EEG, MRI and/or surgical findings supported the concept that the degree
D-mannose (ClDM-6-phosphate). The phosphorylated deoxyglucose compounds are de- of hypometabolism corresponds to areas of confirmed epileptogenic foci. The safety
phosphorylated and the resulting compounds (FDG, FDM, ClDG, and ClDM) presumably and effectiveness of Fludeoxyglucose F 18 Injection to distinguish idiopathic epilepto-
leave cells by passive diffusion. Fludeoxyglucose F 18 and related compounds are cleared genic foci from tumors or other brain lesions that may cause seizures have not been
from non-cardiac tissues within 3 to 24 hours after administration. Clearance from the established.
cardiac tissue may require more than 96 hours. Fludeoxyglucose F 18 that is not involved 16 HOW SUPPLIED/STORAGE AND DRUG HANDLING
in glucose metabolism in any tissue is then excreted in the urine. Fludeoxyglucose F 18 Injection is supplied in a multi-dose, capped 30 mL and 50 mL
Elimination: Fludeoxyglucose F 18 is cleared from most tissues within 24 hours and can glass vial containing between 0.740 to 7.40 GBq/mL (20 to 200 mCi/mL), of no carrier
be eliminated from the body unchanged in the urine. Within 33 minutes, a mean of 3.9% added 2-deoxy-2-[18F-fluoro-D-glucose, at end of synthesis, in approximately 15 to 50
of the administrated radioactive dose was measured in the urine. The amount of radiation mL. The contents of each vial are sterile, pyrogen-free and preservative-free.
exposure of the urinary bladder at two hours post-administration suggests that 20.6% NDC 40028-511-30; 40028-511-50
(mean) of the radioactive dose was present in the bladder. Receipt, transfer, handling, possession, or use of this product is subject to the radioac-
Special Populations: tive material regulations and licensing requirements of the U.S. Nuclear Regulatory
The pharmacokinetics of Fludeoxyglucose F 18 Injection have not been studied in renal- Commission, Agreement States or Licensing States as appropriate.
ly-impaired, hepatically impaired or pediatric patients. Fludeoxyglucose F 18 is eliminated Store the Fludeoxyglucose F 18 Injection vial upright in a lead shielded container at
through the renal system. Avoid excessive radiation exposure to this organ system and 25°C (77°F); excursions permitted to 15-30°C (59-86°F).
adjacent tissues. Store and dispose of Fludeoxyglucose F 18 Injection in accordance with the regulations
The effects of fasting, varying blood sugar levels, conditions of glucose intolerance, and and a general license, or its equivalent, of an Agreement State or a Licensing State.
diabetes mellitus on Fludeoxyglucose F 18 distribution in humans have not been ascerta- The expiration date and time are provided on the container label. Use Fludeoxyglucose
ined [see Warnings and Precautions (5.2)]. F 18 Injection within 12 hours from the EOS time.
13 NONCLINICAL TOXICOLOGY 17 PATIENT COUNSELING INFORMATION
13.1 Carcinogenesis, Mutagenesis, Impairment of Fertility Instruct patients in procedures that increase renal clearance of radioactivity. Encourage
Animal studies have not been performed to evaluate the Fludeoxyglucose F 18 Injection patients to:
carcinogenic potential, mutagenic potential or effects on fertility. • drink water or other fluids (as tolerated) in the 4 hours before their PET study.
14 CLINICAL STUDIES • void as soon as the imaging study is completed and as often as possible thereafter for at
14.1 Oncology least one hour.
The efficacy of Fludeoxyglucose F 18 Injection in positron emission tomography cancer Pregnancy: Advise pregnant women of the risk of fetal exposure to radiation with Flude-
imaging was demonstrated in 16 independent studies. These studies prospectively oxyglucose F 18 Injection [see Use in Specific Populations (8.1)].
evaluated the use of Fludeoxyglucose F 18 in patients with suspected or known malignan- Lactation: Advise lactating women that exposure to Fludeoxyglucose F 18 Injection
cies, including non-small cell lung cancer, colo-rectal, pancreatic, breast, thyroid, melano-
through breast milk can be minimized by pumping and discarding breast milk and
ma, Hodgkin‘s and non-Hodgkin‘s lymphoma, and various types of metastatic cancers to
avoiding close (breast) contact with the infant for 9 hours after Fludeoxyglucose F 18
lung, liver, bone, and axillary nodes. All these studies had at least 50 patients and used
Injection [see Use in Specific Populations (8.2)].
pathology as a standard of truth. The Fludeoxyglucose F 18 Injection doses in the studies
ranged from 200 MBq to 740 MBq with a median and mean dose of 370 MBq.
Manufactured and distributed by:
In the studies, the diagnostic performance of Fludeoxyglucose F 18 Injection varied with
PETNET Solutions, Inc.
the type of cancer, size of cancer, and other clinical conditions. False negative and false
810 Innovation Drive
positive scans were observed. Negative Fludeoxyglucose F 18 Injection PET scans do not
Knoxville, TN 37932
exclude the diagnosis of cancer. Positive Fludeoxyglucose F 18 Injection PET scans can not
replace pathology to establish a diagnosis of cancer. Non-malignant conditions such as
fungal infections, inflammatory processes and benign tumors have patterns of increased
glucose metabolism that may give rise to false-positive scans. The efficacy of Fludeoxyglu-
cose F 18 Injection PET imaging in cancer screening was not studied.
Trademarks and service marks used in this material [a]
CPU: 4 cores, 64 bit; RAM:>=32 GB
are property of Siemens Healthcare GmbH. All other [b]
AIDAN platform scanner software version VG80/VJ30
company, brand, product, and service names may [c]
Lesion Scout with Auto ID is not available for sale
be trademarks or registered trademarks of their in the USA and is not commercially available in all
respective holders. countries. Future availability cannot be guaranteed.
Please contact your local Siemens Healthineers
Please contact your local Siemens Healthineers sales organization for further details.
representative for the most current information or
contact one of the addresses listed below. Note:
Original images always lose a certain amount of detail
when reproduced. All photographs © 2021 Siemens
Healthcare GmbH. All rights reserved.
“Siemens Healthineers” is considered a brand name.

Its use is not intended to represent the legal entity to
which this product is registered. Please contact your
local Siemens organization for further details.
All photographs © 2021 Siemens

Healthcare GmbH. All rights reserved.
Siemens Healthineers Headquarters Published by

Siemens Healthcare GmbH Siemens Medical Solutions USA, Inc.
Henkestr. 127 Molecular Imaging
91052 Erlangen, Germany 2501 North Barrington Road
Phone: +49 9131 84-0 Hoffman Estates, IL 60192
siemens-healthineers.com USA
Phone: +1 847 304-7700
siemens-healthineers.com/mi
MI-5250.NT.JV PDF only · © Siemens Healthcare GmbH, 10.2021

Siemens-Healthineers Mi Alpha White Paper

Uploaded by

Copyright:

Available Formats

Siemens-Healthineers Mi Alpha White Paper

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Siemens-Healthineers Mi Alpha White Paper

Uploaded by

Copyright:

Available Formats

White paper

Automatic landmarking and

ALPHA Automatic landmarking and parsing of human anatomy

FAST Fully assisting scanner technologies

LEAP Learning ensembles of anatomical patterns

MRFDG (18F FDG) metabolic rate

MRGlu Glucose metabolic rate

NHL Non-Hodgkin lymphoma

PACS Picture archiving and communication system

PERCIST PET response criteria in solid tumors

PET Positron emission tomography

ROI Region of interest

SiPM Silicon photomultiplier

SPECT Single photon emission computed tomography

SUV Standardized uptake value

Artificial intelligence (AI)-powered algorithms impact the entire image acquisition,

Figure 1. (a) ALPHA, example

How ALPHA learns

Algorithm learning is based on training using images present in a vast database

Figure 2. The supervised

Redundancy via spatial ensemble through re-alignment

Figure 3. Spatial ensemble

2. Redundancy from different voxel resolutions

Figure 4. Liver image with

The algorithm accuracy is based on the following: performance evaluations using at

[a] CPU: 4 cores, 64 bit; RAM:>=32 GB

Multiparametric PET Suite AI OncoFreeze AI

FlowMotion AI FAST PET Workflow AI

[b] AIDAN platform scanner software version VG80/VJ30

Figure 9. Heart (a) and brain

Protocols can be stored, including the indication-based scan and reconstruction

H&N Lung/liver Colorectal/prostate Lymphoma/melanoma

(a) (b) (c) (d)

• Acquiring enough counts to create a motion-frozen image often requires additional

Deviceless waveform signal Waveform signal from physiological device

Range affected by respiration

Figure 11. Shows the range of

OncoFreeze AI benefits from its predecessor, OncoFreeze.14-17 Both solutions support

OncoFreeze AI is designed to offer motion-frozen PET images in areas affected by

Figure 13. Sample Siemens

Figure 14. Comparison of static

Fludeoxyglucose F 18 5-10mCi as an IV injection

 or indications and important safety information for Fludeoxyglucose F 18 injection

SUV Metabolic rate (Ki) Distribution Volume (DV)

Figure 15. Multiparametric

An internal analysis on 110 WB AC_CTs showed a success rate of >95%. Potential

Automatic derivation of Arterial Input Function (AIF)

Figure 16. Image derived input

Figure 17. Eliminated slices are

CT PET Fused CT PET Fused

Transaxial 325 325 325 324 324 324 3

Sagittal 155 145 155 127 127 127 74

Coronal 155 145 155 70 70 70 245

• Without blank image removal these images would have to be

– This results in 54 additional user interactions per patient

Automated image registration Spine and rib labeling

Auto Ranges Auto Lung 3D

RT Image Suite Auto Views

Figure 19. MI syngo.via AI

or indications and important safety information for Fludeoxyglucose F 18 injection

• Without blank image removal these images would have to be

– This results in 54 additional user interactions per patient

3 X. S. Zhou, Y. Zhan, Z. Peng, M. Dewan, B. Jian, A. Krishnan, M. Harder, R.

4 X. S. Zhou, Y. Zhan, V. C. Raykar, G. H. Valadez, L. Bogoni, Z. Peng, M. Dewan, M.

5 Y. Zhan, M. Dewan, M. Harder, A. Krishnan, X. S. Zhou, “Robust automatic knee MR

6 Y. Tao, Z. Peng, A. Krishnan, X. S. Zhou, “Robust learning-based parsing and

7 W. Watanabe, A. Kitaoka, K. Sakamoto, M. Yasugi, and K. Tanaka, “Illusory Motion

10 P. J. Keall, G. S. Mageras, J. M. Balter, R. S. Emery, K. M. Forster, S. B. Jiang, J. M.

17 S Poesse, I Hong, D Mannweiler, et al. “Respiratory motion compensation in PET/CT:

18 A . Smith, B. Spottiswoode, V. Shah, J. Hu, C. von Gall, “FlowMotion

19 D. Vriens, L. de Geus-Oei, W. Oyen, E. Visser, “A Curve-Fitting Approach to Estimate

20 AR Pantel, DA Mankoff. “Molecular imaging to guide systemic cancer therapy:

21 R. Wahl, H. Jacene, Y. Kasamon, M. Lodge, “From RECIST to PERCIST: Evolving

22 K Schmiedehausen, “NaF PET/CT in Prostate Cancer,” White Paper, Siemens Medical

24 V. Vassiliou, D. Andreopoulos, S Frangos , N. Tselis, E. Giannopoulou, S. Lutz.,

25 A . Hans Vija, C. Von Gall, P. Ghosh, “Accurate, reproducible, and standardized