Comparison of multiple quantitative strategies for neuropathologic image analyses.

Area of Science:

• Neuropathology
• Digital Pathology
• Artificial Intelligence

Background:

• Traditional semiquantitative (SQ) scoring for neuropathology has limitations including assessor variability and incomplete pathological spectrum capture.
• Digital pathology offers advanced techniques like positive pixel quantitation and AI for more objective assessment.
• A comprehensive comparison of these digital methods against traditional SQ scoring is lacking.

Purpose of the Study:

• To compare traditional semiquantitative (SQ) scoring with computer-driven percent area-stained measures and AI-driven cellular density quantitation for tau pathology.
• To evaluate the performance of these methods in identifying neuropathological changes, particularly in chronic traumatic encephalopathy (CTE).

Main Methods:

• Utilized 1412 cases from Boston University brain banks.
• Compared human-driven SQ scoring with computer-driven percent area-stained measures and AI-driven cellular density quantitation of tau pathology in the dorsolateral frontal cortex.
• Conducted a subanalysis in CTE cases to examine correlations with clinical and neuropathologic variables.

Main Results:

• General agreement was observed between all measures across the entire dataset.
• In CTE cases, all methods predicted neuropathology, but the positive pixel method showed increased variability due to background, noncellular elements, and artifacts.
• AI-driven methods demonstrated superior ability in identifying pathological changes associated with sparse pathology.

Conclusions:

• Significant differences exist among neuropathologic assessment techniques.
• AI-driven methods show promise for more accurate and sensitive detection of sparse pathology.
• Careful consideration of analysis method selection is crucial for reliable neuropathologic assessment.