WEAKLY SUPERVISED SEGMENTATION AND CLASSIFICATION OF ALPHA-SYNUCLEIN AGGREGATES IN BRIGHTFIELD MIDBRAIN IMAGES

Area of Science:

• Neuroscience
• Computational Pathology
• Biomedical Image Analysis

Background:

• Parkinson's disease (PD) involves alpha-synuclein aggregate accumulation.
• Accurate diagnostics rely on identifying Lewy bodies and neurites.
• Automated analysis of histopathology images can enhance understanding of PD pathology.

Purpose of the Study:

• To develop an automated image processing pipeline for segmenting and classifying alpha-synuclein aggregates.
• To differentiate between Lewy bodies and neurites in Parkinson's disease and incidental Lewy Body Disease (iLBD) cases.
• To create a framework for large-scale characterization of aggregate distribution and heterogeneity.

Main Methods:

• Developed a weakly supervised segmentation pipeline for whole-slide images (WSIs) of midbrain tissue.
• Utilized a ResNet50 classifier for aggregate classification.
• Ensured robustness to immunohistochemical labelling variability.

Main Results:

• Achieved a balanced accuracy of 80% in differentiating major aggregate morphologies (Lewy bodies and neurites).
• Successfully segmented and classified alpha-synuclein aggregates in PD and iLBD cases.
• Demonstrated the framework's capability for analyzing brightfield immunohistochemical tissue.

Conclusions:

• The automated pipeline enables precise characterization of alpha-synuclein aggregate morphology and spatial distribution.
• This approach facilitates large-scale studies on PD and iLBD pathology.
• Provides a foundation for investigating the relationship between aggregates and surrounding cells like microglia and astrocytes.