Deep learning-based model for diagnosing Alzheimer's disease and tauopathies

Area of Science:

• Neuropathology
• Computational Pathology
• Artificial Intelligence in Medicine

Background:

• Tauopathies, including Alzheimer's disease (AD), progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), and Pick's disease (PiD), are challenging to differentiate neuropathologically.
• Accurate diagnosis relies on identifying specific tau lesion types in brain tissue.

Purpose of the Study:

• To develop and validate a deep learning-based model for the differential diagnosis of major tauopathies.
• To utilize tau-immunostained digital slide images for automated lesion detection and classification.

Main Methods:

• Trained YOLOv3 object detection algorithm to identify five tau lesion types in CP13-immunostained slides.
• Constructed random forest classifiers using quantitative tau lesion burdens from multiple brain regions.
• Augmented data and validated the model on independent datasets stained with CP13 and AT8.

Main Results:

• The random forest classifier achieved an average test accuracy of 0.97, correctly diagnosing 29 out of 30 cases.
• Validation studies demonstrated diagnostic accuracy >92% (without augmentation) and >95% (with augmentation).
• The model trained on CP13 also performed well on AT8-stained slides.

Conclusions:

• A deep learning model can accurately differentiate between major tauopathies using tau-immunostained digital slides.
• The developed diagnostic tool shows potential for assisting neuropathologists in medical decision-making.
• The model's applicability across different staining protocols (CP13 and AT8) enhances its utility for broader research and clinical use.