C191-E044

Technical A New Peak Integration Algorithm for

Report LabSolutions

Yusuke Osaka1, Shinji Kanazawa1, Hiroaki Kozawa1, Etsuho Kamata1

A b s tra c t:
The time required to analyze large quantities of data obtained from the increasingly quick analyses and increasingly prevalent simultaneous
multicomponent analyses in recent years has become a challenging issue. In addition, the strengthening of regulations associated with data
integrity has given rise to demands for automated and simplified manual methods for integrating the areas of problematic chromatogram
peaks. Herein, we introduce a new peak integration algorithm for LabSolutions, i-PeakFinder, which is designed to resolve these issues.

Keywords: Data processing, peak integration, algorithm

1 . I n tro d u c tio n der and main peaks using traditional peak integration methods, i-Peak-
Finder can automatically detect shoulder peaks while maintaining con-
i-PeakFinder, a new peak integration algorithm for LabSolutions, is a sistent peak detection sensitivity throughout the entire chromatogram.
completely automated integration function that can detect peaks with In general, it is difficult to automatically detect considerably small shoul-
high levels of accuracy without requiring special parameter adjust- der peaks, as shown in Fig. 2; however, i-PeakFinder can even detect
ments. In addition, this algorithm has adjustable parameters that such peaks automatically based on the threshold judgment.
allow the integration function to be applied to a wide range of com-
plex chromatogram patterns. Furthermore, the adjustable parameters
allow to output highly accurate peak integration results from large
quantities of data, even for batch analysis.

2 . F e a tu re s o f i-Pe a k Fin d e r
With Shimadzu's traditional peak integration methods or competitor
methods, some complex chromatograms require time programming in
addition to the typical parameter adjustments for peak integration. Highly accurate detection of shoulder peaks
However, i-PeakFinder can perform peak integration of these chro-
matograms just with simple parameter adjustments. This peak inte-
gration function has the following features:
• Highly accurate detection of shoulder peaks
• Simple adjustment of peak-baseline processing
• Improved peak-baseline positioning that results in improved reproducibility
• Performing accurate peak integration even with variabilities caused
by baseline drift
Peak detection at low S/N levels Automatic determination of
Shimadzu also emphasizes on compatibility, so LabSolutions can also
baseline undulation and drift
be used with the traditional peak integration methods (Chromatopac
mode). Switching between the traditional peak integration and
i-PeakFinder methods during analysis is easy, allowing the user to
select an appropriate peak integration method for the circumstances.
This includes selecting a traditional method for compatibility with
past data. Fig. 1 shows examples of using the completely automated
integration function to analyze typical peaks.

2-1. Highly Accurate Detection of Automatic tracking of peak Automatic tracking of noise
Shoulder Peaks width variation intensity changes

i-PeakFinder can accurately detect shoulder peaks. In cases wherein Fig. 1 i-PeakFinder's Completely Automated Integration Function
manual peak integration is required to distinguish and detect the shoul-

1 Analytical & Measuring Instruments Division 1

 mAU mV
2.0
4

Main peak

Impurity
[Detection Threshold] = 10

1.5 3

2
1.0

1
0.5
0

0.0
12.5 15.0 17.5 min
17.5 20.0 min [Baseline Type] Not configured
Normal shoulder peaks can be detected with no special parameter adjustment mV

mAU 4

Impurity
Main peak
2.0
[Detection Threshold] = 10
3
[Detection Threshold of
1.5 Unseparated Small Peak] = 0.3
2

1.0 1

0
0.5

12.5 15.0 17.5 min

0.0
[Baseline Type] Vertical Division
mV
17.5 20.0 min
Even very small shoulder peaks can be detected with additional parameter adjustments. 4

Main peak

Impurity
3
Fig. 2 Examples of Shoulder Peak Detection

2
2-2. Simple Adjustment of Peak-baseline
Processing 1

0
Accurate detection of impurity peaks is essential for pharmaceutical
quality control and other applications. Impurity peaks are often fused at
12.5 15.0 17.5 min
the base of a major peak, and quantitative results obtained via area nor-
[Baseline Type] Base to Base
malization can vary depending on the method used for peak-baseline
processing. The method used for peak-baseline processing also differs Note: Other than peak integration range and peak-baseline processing type, all settings are at default.

based on the samples and testing objectives. With traditional methods,

performing a specific type of peak-baseline processing requires the user Fig. 4 Peak-baseline Processing Example
to include time programming or perform peak integration manually.
Table 1 Quantitative Results Obtained via Area Normalization with
However, i-PeakFinder has adjustable parameters in its basic settings Different Methods of Peak-baseline Type
for performing a specific type of peak-baseline processing, allowing
Not configured Vertical Division Base to Base
the user to easily implement the optimum peak-baseline type in each
Main peak 99.681 99.448 99.680
situation. Fig. 3 shows a list of peak-baseline type in the settings
Impurity 0.160 0.338 0.160
window, Fig. 4 shows the result of peak-baseline type setting of an
impurity peak that is fused at the base of a major peak, and Table 1
summarizes the quantitative results obtained via area normalization
with different peak-baseline type settings. Peak-baseline processing
2-3. Improved Peak-Baseline Positioning
appropriate for a particular situation can be simply performed by for Enhanced Reproducibility
changing a few basic parameters.
When a target component is separated with peak tailing or leading in
some cases, using a traditional peak integration method would result
in variability in the calculated peak areas depending on which point
along the trailing and leading edge was defined as the peak base.
This affected the reproducibility of area results.
With i-PeakFinder, the user can adjust the height at which the start
and end of a peak-baseline needs to be defined so that peak-baseline
processing is consistent even with tailing and leading peaks.
The reproducibility of results obtained using the traditional method and
i- PeakFinder is summarized in Table 2 for the chromatogram shown in Fig.
5. In the example below, the peak area results obtained using i-PeakFinder
exhibit superior reproducibility compared to the traditional method.
Fig. 3 Peak-baseline Type Settings
2
 50
mV
3-2. Detailed Peak Detection Settings
40 For complex chromatograms, adjustments of the detection threshold,
peak integration range, and peak-baseline type may not be enough to
30 obtain the desired peak integration results. i-PeakFinder is compatible
20
with a wide variety of chromatograms and allows the user to configure
more detailed peak detection conditions. Some of these detailed settings
10 are described below:

0
(1) To Detect Peaks Not Affected by Noise [Minimum Half Width]
1.00 1.05 1.10 min Smoothing is sometimes performed on chromatograms obtained via
Fig. 5 Chromatogram with Peak Tailing (N=5)
LCMS. If the noise frequency is close to the peak frequency, then it be-
comes difficult to determine peaks automatically and single peaks may
be recognized as multiple peaks. In this situation, configuring the mini-
Table 2 Comparison of Area Reproducibility Between the Traditional
mum half width setting ignores noise smaller than that value and en-
Method and i-PeakFinder sures that peaks with a FWHM value above the set minimum value are
i-PeakFinder Traditional method detected among the peaks with a wide wave profile due to smoothing.
%RSD 0.106 0.275
Fig. 8 shows the difference made by increasing the minimum FWHM
value. This feature is useful when noise can be observed in a peak.
Peak-baseline Type: Base to Base Width: 1 s
Processing
Peak-Baseline Height* 2 Slope: 2000 uV/min
* Described in more detail later in the article

3 . S e tti n g P a ra me te rs
i-PeakFinder not only performs highly accurate peak integration with de-
14.00 14.50 15.00 14.00 14.50 15.00
fault settings but also includes more detailed settings for adjusting the peak [Minimum Half Width] Not configured [Minimum Half Width] = 2
detection conditions (peak width and peak-baseline start and end heights)
and peak-unifying conditions that determine how to combine multiple
peaks. The most important of these settings will be explained here.

3-1. Basic Peak Detection Parameters

The basic parameters that allow the user to adjust peak detection condi- 14.00 14.50 15.00 14.00 14.50 15.00
tions are the peak-baseline type mentioned above, peak detection [Minimum Half Width] = 5 [Minimum Half Width] = 10
threshold, and peak integration range. Using the detection threshold set- Fig. 8 Example Configuration of Minimum FWHM
ting, the peaks below a certain threshold are not detected based on the
estimated noise level calculated using a proprietary algorithm. Decreas- (2) To Ensure Peak Area Accuracy and Linearity [Peak Baseline Height]
ing the peak detection threshold value allows the detection of smaller With tailing peaks and chromatograms with a large amount of baseline
peaks. The peak integration range specifies the time range during which noise, the peak start and end points can vary depending on the data,
peaks will be detected. Fig. 6 shows the change in results produced by which can reduce area accuracy. By using the peak-baseline height set-
changing the peak detection threshold from the default setting of 5 to ting, the peak start and end points are recognized as the product of a
2,000. These intuitive controls allow the user to detect or not to detect value entered for peak-baseline height and noise intensity calculated
small peaks with a simple adjustment. Fig. 7 shows an example of adjust- using a proprietary algorithm. Consequently, the larger the peak-base-
ing the peak integration range. Without adjusting the peak integration line height value is, the higher the peak-baseline is positioned.
range, all peaks fall within the range and the peak-baseline is affected by
negative peaks. Setting the peak integration range to exclude negative Using this setting provides good reproducibility in the determination of
peaks allows the user to configure an appropriate baseline. peak- baseline start and end points. This improves peak area reproducibil-
ity, as mentioned earlier in Table 2, and provides linearity across results
mAU mAU
obtained for a target component at different concentrations. Fig. 9
30 30
shows an example configuration of peak-baseline height where the base-
25 25
line length of the tailing peaks is adjusted by specifying the peak-baseline
20 20
height.
15 15

10 10

5 5

0 0
1.0 2.0 min 1.0 2.0 min
[Detection Threshold] = 5 (default value) [Detection Threshold] = 2000

Fig. 6 Peak Detection Threshold

15.0 min 15.0 min
[Peak Baseline Height] Not configured

5.0 10.0 min 5.0 10.0 min

[Peak Integration Range] Not configured [Peak Integration Range] ≥ 4.1 (min) 15.0 min 15.0 min

Fig. 7 Peak Integration Range [Peak Baseline Height] = 50

Fig. 9 Example Configuration of Peak-Baseline Height for Tailing Peaks 3
(3) To Avoid Recognizing Long-Period Undulations as Peaks
[Maximum Half Width]
The maximum half width, which is the opposite of the minimum half
width, is a parameter that can be specified to ignore larger peaks as
baseline undulations. For example, Fig. 10 shows baseline drift ap-
pearing as a large baseline bulge that could be recognized as a peak.
This undulation can be removed by specifying a maximum FWHM.

0.95 1.00 min 0.95 1.00 min

[Unify Peaks by Shoulder Ratio] Not configured [Unify Peaks by Shoulder Ratio] = 0.01 (%)

Fig. 12 Example of Setting a Threshold for Recognizing Shoulder Peaks

4. Cus tomizable Default Valu es

0.0 5.0 min 0.0 5.0 min
[Maximum Half Width] Not configured [Maximum Half Width] = 20 (sec) The i-PeakFinder parameters can be configured as the default values
Fig. 10 Example Configuration of Maximum FWHM of methods. For example, if a user decides to use vertical division as
the peak-baseline type, specifying vertical division as the default set-
(4) To Unify Fused Peaks into a Single Peak ting can omit the effort involved in adjusting the parameter for each
[Unify Peaks by Separation Width] new analysis. LabSolutions is an integrated workstation capable of
The minimum half width setting is used to avoid mistakenly recogniz- controlling a range of different instrument models. However, the
ing noise as peaks, while unifying peaks by separation factor is used HPLC and GC systems occasionally require different default peak in-
to combine fused peaks into a single peak. Fig. 11 shows three fused tegration settings. Accordingly, default settings can be specified for
peaks. Configuring this setting will unify the two side peaks into the each type of instrument (Fig. 13).
larger peak. Note that this setting is only effective for baseline inter-
vals containing fused peak.

8.1 9.0 min 8.1 9.0 min

[Unify Peaks by Separation Width] Not configured [Unify Peaks by Separation Width] = 2

Fig. 11 Example of Unifying Unresolved Peaks

Fig. 13 The Default Settings Window
(5) To Decide Whether to Recognize Shoulder Peaks
[Unify Peaks by Shoulder Ratio]
When impurities were fused as shoulder peaks at the base of a main 5. Summary
component peak, the traditional peak integration method required
time programming or manual peak integration to detect the shoulder i-PeakFinder, a new peak integration algorithm developed by Shi-
peak. i-PeakFinder allows easy detection of shoulder peaks and also madzu, can perform either automated peak integration or peak inte-
allows the user to decide whether to recognize (or not) impurity peaks gration with simple settings adjustments, even for cases in which the
based on a threshold ratio of the main component peak height to the traditional peak integration method requires time programming or
shoulder peak tangential height. Fig. 12 shows an example of shoul- manual peak integration for individual data sets. i-PeakFinder also
der peak detection implemented without time programming or allows the user to easily configure the peak-baseline type and per-
manual peak integration and an example of using the threshold value form accurate peak integration of small peaks.
to unify the shoulder peak with the main component peak. Configur- Furthermore, i-PeakFinder can quickly and easily process large quanti-
ing a threshold value can also be used as a condition for deciding ties of data without errors and improve the efficacy and reliability of
whether to recognize impurity peaks. analytical operations.

First Edition: July, 2017

© Shimadzu Corporation, 2017