Introduction:
WS Recognizer recognizes any type of targets (e.g. red blood cells) and uses kernel regression to calculate their spatial distribution across the entire slide. The average time for processing one slide is about 20~30 minutes, depending on the size of the tissue section and whether you use multiple-core CPU to boost the performance.
The following is the tutorial to reproduce the results.
Video tutorial:
STEP1: Load WSI data
Select File][Open] to load the WSI data and use the navigation window at the bottom to browse the slide. The navigation window shows the mask for WSI analysis. The data outside the mask will be ignored.
STEP2: Train classifier
1. In the center window, do several LEFT mouse clicks to sample the target stains.
2. Similarly, do several RIGHT mouse clicks on non-target stains. You may need to sample as diverse as possible so that the classifier will know what are the background stains.
3. In the recognition window, press "test" to test the classifier. If the result is not good, try adding more samples by left clicks and right clicks to fine tune the classifier. You may start over by press "clean" button to restart the training.
TIP: You may train the classified using an incremental approach fine-tune the result.
TIP: You can save/load the trained classifier for later use.
STEP3: Whole slide image analysis
Click on the "Run whole slide" button to do the whole slide image analysis. The program will make use of the multi-core feature to run the analysis.
TIP: You can increase the number of threads used in analysis in the "Options view"
TIP: You can abort the analysis anytime by click the button again.STEP4: Results
The results are shown in a heap map. You may choose to present "target density" or "target size". You can also modify the contrast.
TIP: The results can be saved as a figure, an MATLAB file, or a TXT file for quantitative analysis.
Text file will have multiple rows, with each row storing the information for each recognized target. The information includes the following:
1) X coordinates of the target in microns
2) Y coordinates of the target in microns
3) The span of the target in microns. The span is calculated by 0.5*(X_span+Y_span) of the target
4) Target area size in micron square.
5) Average intensity of the target. The intensity is calculated by (R+G+B)/3.0
TMA (Tissue microarray) slides will show results for each tissue core. The number of the targets recognized in each core will be summed and reported (see figure).
Publications Using WS Recognizer
1. Wu, Yijen L., Li Liu, Fang-Cheng Yeh, Bedda L. Rosario, and Chien Ho. "MRI Investigation of New Approach to Improve the Recovery of Myocardial Ischemia Reperfusion Injury by Treatment with Intralipid®." World Journal of Cardiovascular Diseases 6, no. 10 (2016): 352.
2. Yeh, Fang‐Cheng, Li Liu, T. Kevin Hitchens, and Yijen L. Wu. "Mapping immune cell infiltration using restricted diffusion MRI." Magnetic resonance in medicine (2016).
3. Yeh, Fang-Cheng, Qing Ye, T. Kevin Hitchens, Yijen L. Wu, Anil V. Parwani, and Chien Ho. "Mapping stain distribution in pathology slides using whole slide imaging." Journal of pathology informatics 5, no. 1 (2014): 1.
4. Yeh, Fang-Cheng, Anil V. Parwani, Liron Pantanowitz, and Chien Ho. "Automated grading of renal cell carcinoma using whole slide imaging." Journal of pathology informatics 5 (2014).
