Segmenting renal whole slide images virtually without training data.

Area of Science:

• Digital pathology
• Computational histopathology
• Medical image analysis

Background:

• Automated systems are essential for processing large digital pathology datasets.
• Variability in slide properties complicates the development of supervised learning models.
• Glomeruli segmentation is critical for renal histopathology analysis.

Purpose of the Study:

• To develop and evaluate a two-stage pipeline for automated glomeruli detection and segmentation.
• To address the challenge of varying image properties without requiring pre-existing training data.
• To adapt and combine unsupervised segmentation methods for precise glomeruli identification.

Main Methods:

• A two-stage pipeline involving weakly supervised patch-based detection and precise segmentation.
• Application of a kernel two-sample test for model adaptation and optimization.
• Adaptation and combination of unsupervised segmentation methods, including level-set and polygon-fitting approaches.

Main Results:

• The best performing polygon-fitting method achieved 51% glomeruli segmentation accuracy (DSC > 0.8).
• The detection stage resulted in 42% false positives.
• The performance is considered good given the challenging application scenario and minimal training data requirement.

Conclusions:

• The proposed pipeline offers a viable solution for glomeruli segmentation in digital pathology, particularly when training data is limited.
• Further strategies are discussed to enhance segmentation performance.
• The approach demonstrates potential for improving automated analysis in renal histopathology.