Home
About DataPath
The Whole Slide Image Processing and Machine Learning Performance Assessment Tool is a software program written in Python for analyzing whole slide images (WSIs), assisting pathologists in the assessment of machine learning (ML) results, and evaluating ML performance.
The toolbox includes modules to: - Generate image patches for AI/ML models - Extract and visualize annotations from WSIs - Generate Aperio ImageScope annotation files for pathologist validation - Normalize stain color across images - Register slides across scanners - Split datasets with category/subtype-aware stratification - Assess model performance and metrics
The Whole Slide Image Processing and Performance Assessment Tool code has been used in the following publications:
Kahaki, Seyed, et al. “Weakly supervised deep learning for predicting the response to hormonal treatment of women with atypical endometrial hyperplasia: a feasibility study.” Medical Imaging 2023: Digital and Computational Pathology. Vol. 12471. SPIE, 2023.
Kahaki, Seyed, et al. “Supervised deep learning model for ROI detection of atypical endometrial hyperplasia and endometrial cancer on histopathology whole slide images for predicting hormonal treatment response.” Medical Imaging 2024: Digital and Computational Pathology.
Kahaki, Seyed, et al. “End-to-End Deep Learning Method for Predicting Hormonal Treatment Response in Women with Atypical Endometrial Hyperplasia or Endometrial Cancer.” Journal of Medical Imaging, Journal of Medical Imaging, Under Review
Mariia Sidulova, et al. “Contextual unsupervised deep clustering of digital pathology dataset”, Submitted to ISBI 2024
Modules
There are several modules in this package including:
WSI handler: includes functions and classes for general WSI analysis such as read whole slide images, tissue segmentation, and normalization.
Annotation Extraction: this module includes several functions for processing annotations such as annotation extraction.
Patch Extraction: which assist pathologist and developers in extracting image patches from whole slide images region of interest.
Color Normalization: Normalizes stain color using Macenko, Vahadane, Reinhard, and histogram matching methods.
Tissue Registration: Registers tissue regions between WSIs scanned from different scanners using ORB-based alignment.
Stratification: Splits datasets into train/val/test while preserving class balance across categories and subtypes.
To see a demo of the functions in this toolbox, please refer to the Jupyter Notebooks files in the root folder of this package.