GC Image Users' Guide

Comparative Analysis and Visualization

GC Image has a Compare Images tool for comparative analysis and visualization; a Comparison Report for comparing blobs, blob sets, and areas from multiple chromatograms; and the Image Investigator for advanced analysis of multiple chromatograms. Comparisons are useful for process monitoring, sample classification or identification, correlative determinations, and other important tasks. GCxGC is a powerful technology for chemical analysis, but methods for comparative visualization must address challenges posed by GCxGC data: inconsistency and complexity.

Compare Images

The Compare Images tool allows comparisons between two GCxGC chromatograms: the dataset currently selected for analysis, referred to as the analyzed image, is compared to another image, referred to as the reference image . Prior to comparison, each image is processed separately to correct acquisition artifacts (e.g., background removal) and to detect, identify, and quantify chemical peaks. These operations are described in previous chapters.

To initiate a comparison, select the View -> Compare Images from the menu bar or click the Compare images button. The Compare Images tool can be invoked for the entire image or for a rectangular subimage that encloses selected graphics (in graphics mode) or selected blobs (in blob mode). The Compare Images tool is started with the current image used as the analyzed image.

Next, select the reference image to be compared to the analyzed image, as shown in Figure 1.

Figure 1: The Compare Images tool.

As the reference image is loaded, two data processing steps automatically are performed on it to remove incidental differences with the analyzed image.

• Registration performs a transformation of the reference image so that the retention times of identified peaks in the reference image are better matched to the retention times of the corresponding peaks in the analyzed image.
• Scaling normalizes the response (i.e., total peak intensity) for quantitative standard(s) in the reference image to the response for the same standard(s) in the analyzed image.

These two steps are critical for suppressing incidental variations and emphasizing real chemical differences.

Visual Comparisons

The Compare Images tool has a pull-down menu to choose one of several visual comparison methods:

• Grayscale Difference — Pixel-by-pixel subtraction of the analyzed image minus the reference image. The range of pixel values are mapped linearly to a grayscale from black to white (0 to 255), with dark values (0-127) for negative differences, gray (value 128) for zero difference, and bright values (128-255) for positive differences.
• Difference + Peaks — Same as Grayscale Difference, except that an overlay with green points marking the locations of peaks in the analyzed image and red points marking the locations of peaks in the reference image. This view is useful to visualize the quality of the registration between the images.
• Colorized Difference — This visualization uses Hue-Saturation-Value (HSV) color space, with hue indicating which image has the greater value (green for analyzed, red for reference), saturation indicating the magnitude of the difference, and brightness indicating the value of the larger of the two images.
• Grayscale Fuzzy Difference — This difference is non-zero only if the pixel in one image is outside the range of pixel values in a small neighborhood in the other image. The image is presented using a grayscale, as described above. The size of the window is set by sliders.
• Colorized Fuzzy Difference — This method uses the neighborhood-based fuzzy difference and HSV colorization.
• Grayscale Ratio — This method uses division rather than subtraction. The value is computed as the pixel-by-pixel division of the analyzed image by the reference image.
• Colorized Ratio — This method uses HSV colorization for the ratio.
• Grayscale Fuzzy Ratio — This method uses a fuzzy ratio, which is zero unless the pixel value is outside the range of pixel values in a small neighborhood in the other image.
• Colorized Fuzzy Ratio — This method combines HSV colorization with the fuzzy ratio.
• Addition — This method uses the Red-Green-Blue color space, with the analyzed image pixels values in the green band and the reference image pixel values in the red band.
• Flicker — This method cycles between the two images, alternately displaying the analyzed image and the reference image. The delay time for each image is set by a slider. With this the GUI has buttons for start, stop, and step.

Several controls are available for all visual comparison methods.

• There is a selector to show all pixels in the selected image or to show only some pixels: pixels in all blobs, pixels in included blobs, pixels in selected blobs, or pixels in included graphic objects.
• The image can be scaled by a pull-down menu.
• The intensity mapping can be parameterized with a slider to perform logarithmic, linear, or exponential scaling of the difference. The image must be recomputed (with the Recompute Image button) to use the new parameter for intensity mapping.

Other Comparison Features

Additional GUI components indicate range of comparative values, the position of the cursor, the values of both images at the cursor location, and the tool status.

Clicking the Blob Table button brings up a table with data about the blobs in both images (shown in Figure 2).

Figure 2: The comparison Blob Table.

The table has columns for:

• compound name,
• blob IDs for both images,
• blob group names for both images,
• blob inclusion flags for both images,
• internal standard blob IDs for both images,
• peak retention times for both dimensions for both images plus the difference (analyzed minus reference) and percentage (analyzed divided by reference times 100),
• peak value for both images plus difference and percentage,
• blob area for both images plus difference and percentage, and
• blob volume for both images plus difference and percentage.

The table is sortable by any column. Selecting rows in the table activates graphical highlights in the comparison image with a green for blobs in the analyzed image and red for blobs in the reference image.

The difference image can be saved by clicking the Save Image button of the Compare Images tool.

Multi-Run Comparison Report

GC Project allows comparing blobs, blob sets, and areas from multiple images with the Comparison Report. The comparison is name-based and performed on all named blobs, blob sets, and areas. To create a report,

1. Open a project in GC Project.
2. Select Menu: Reports -> Comparison Report.

   

3. Add images to the Selected list.
4. Select the objects to be compared.
5. Click “Configure Features” button to specify the analyte features or select “Use Configuration in Image” to use the features specified in the first image.
6. Select "Set Filter" to set a filter rule for the comparison. (Refer to Filter Report for more information about the filter rules.) Click OK to apply the filter rule.
7. Click Compare to enter the report file name and the format, then save the report.

The Comparison Report (e.g., as shown in Figure 3) has several components:

• The Image List is a table that lists the images reported (those selected by the user), with an identifying number for each image given by the ordinal position of the image in the list. The Image List is always present in the report.
• The Blobs Report contains (A) a Statistics Summary table, (B) a table for each reported compound, and (C) a trend graph for each blob table field with a numeric value. The list of reported compounds is the set of blob names used in any reported image.
  1. The Statistics Summary table has one line for each reported compound with: compound name, number of chromatograms containing the named blob, and the average and standard deviation for each blob table field that has a numeric value. (For example, Volume has a numeric value, but Group Name does not.) If any image contains duplicate blob names, only one of the named blobs will be reported and the report will contain a warning.
  2. The table for each reported compound has one line for each image with: the image name, compound name, and each blob table field. The last two rows of the table give the average and standard deviation for each blob table field that has numeric values.
  3. The trend graphs (one for each numeric blob table field for each reported compound) plot values for each image in the order specified by the user. The graph between plotted points is determined by linear interpolation.
• The Areas Report is identical to the Blobs Report except that only the blob table fields applicable to area objects are reported.
• The Blob Sets Report is similar to the Blobs Report. The Statistics Summary table has a row for each reported blob set with: blob set name, number of chromatograms with the named blob set, and the average and standard deviation for each blob set table field.
• The Sample Report has a section for each sample run that includes a table of metadata for the sample run and an image of the processed chromatogram.

Figure 3: An example Comparison Report.

Image Investigator™

Image Investigator™ allows users to analyze multiple chromatograms. Users can compare chromatograms, blobs, areas, and blob sets from a set of chromatograms, and examine statistical characteristics and trends. Image Investigator provides the same results as Comparison Report, but with an interactive graphic user interface. The comparison is name based and performed on all named blobs, blob sets, and areas. In later releases, Image Investigator will be enhanced with additional capabilities for clustering, classification, and fingerprinting.

Image Investigator can be either run as a stand-alone interface or invoked from GC Project. To invoke Image Investigator from GC Project:

1. Open a project in GC Project.
2. Go to the menu: Actions -> Investigate Images.
3. Add images to the Selected List.
4. Click OK when done.

Then, the Image Investigator will present a Load Options dialog for the selected images, as shown in Figure 4.

Figure 4: The Load Options dialog.

The Load Options dialog allows the user to assign optional class labels to chromatograms. If no label is specified, all chromatograms will be assigned the same default label. The user also specifies which feature option is used for the comparative analysis:

• As is:

  The existing blobs, areas, and blob sets of the selected chromatograms are used for the analysis.

• Apply an existing feature template:

  The user can specify a feature template to apply to the selected chromatograms. The existing blobs, areas, and blob sets of selected chromatograms are ignored in preference for those in the feature template.

• Auto generate a feature template:

  The software can automatically generate a feature template using a generic framework for automated non-targeted cross-sample analysis. The framework supports automated comprehensive feature generation, alignment, and matching across many samples. The framework uses advanced peak-region features which provide a high degree of selectivity and implicit cross-sample feature matching.

  The user specifies a filename for the automatically generated feature template and the output folder where the intermediate results and the final template will be saved. The user also can specify an optional configuration file with customized data processing settings (e.g., for blob detection). After specifying all fields, Click OK and a processing monitor will pop up and display the processing progress as shown in Figure 5. When the processing is finished, click OK to load the chromatograms and apply the auto-generated feature template.

  All chromatograms must be processed with blobs detected before being used to automatically generate a feature template. The processing is usually time and memory consuming. If out-of-memory errors occur, the user could either use a smaller set of chromatograms or reprocess the raw data files with filters to reduce data size. If the processing is interrupted, the processing can be resumed by loading the same set of chromatograms, specifying the same output location, and choosing No when the software asks for overwriting.

  Figure 5: The processing monitor for auto-generated feature templates.

After loading all chromatograms and applying the feature template, Image Investigator presents four perspectives for analysis:

• Images Perspective

  The Images perspective,shown in Figure 6, allows users to compare chromatograms, blobs, areas, and blob sets from a set of samples. The left side pane of the Images tab contains the list of chromatograms being compared. The user can cycle though the images by simply clicking on the chromatogram name or using the Up and Down arrow keys. Once a chromatogram is selected, its image is shown on the top-right pane. The bottom-right pane shows its Image Attributes, Blob Table, Area Table, and Blob Set Table.

  Figure 6: The Images Perspective.

• Compounds Perspective

  The Compounds perspective, shown in Figure 7, allows users to examine the changing trends of certain statistics of a compound (Blob or Area) or a compound set across chromatograms. The left side pane of the Compounds tab contains the list of available compounds and compound sets. The top-right pane shows the trend graphs (one for each statistic for the selected compound). The data points are plotted in the order of the chromatograms specified by the user. The bottom-right pane shows the data table for the selected compound. The bottom rows of the table show the calculated statistics including mean, standard deviation, relative standard deviation (RSD), and Fisher ratio.

  Figure 7: The Compounds Perspective.

• Statistics Perspective

  The Statistics perspective, shown in Figure 8, allows users to examine the distribution of a certain statistic for compounds or compound sets across chromatograms. The left side pane of the Statistics tab contains the list of available statistics. The top-right pane shows the distribution graph (one line per chromatogram) and a distribution summary graph that shows the means and standard deviations. The bottom-right pane shows the data table of the selected statistics. The bottom rows of the table show the calculated statistics including mean, standard deviation, relative standard deviation (RSD), and Fisher ratio.

  Figure 8: The Statistics Perspective.

• Summary Perspective

  The Summary perspective, shown in Figure 9, provides a summary of all the computed means, standard deviation,s relative standard deviations (RSDs), and Fisher ratios for all statistics of each compound category. There are the three compound categories: blobs, areas, and compound sets. Once a category is selected in the left side pane, the corresponding Statistical Summary table will be shown in the right side pane.

  Figure 9: The Summary Perspective.

  Below the summary table, there is a filter tool panel that allows filtering features based on various statistics. If a feature template is specified when loading chromatograms, a filtered feature template can be exported based on the current filter. Also, the user can export the table for for analysis with external programs (e.g., spreadsheet software).

Contents Previous: Chemical Identification Next: Configurations

GC Image™ Users' Guide © 2001–2010 by GC Image, LLC, and the University of Nebraska.