New Paper for Deep Learning for Quantifying Neuron Loss in Alzheimer’s Disease

Alzheimer’s disease involves the interplay of multiple pathological processes, ultimately leading to neuronal loss and brain degeneration. In our recent collaborative study with the Lab of Neuropathology, we addressed a key challenge: quantifying neuron loss that can be attributed to individual pathologies across thousands of histological images, while accounting for both biological and technical variability.

To achieve this, our colleagues Geethika Arekatla and David Wouters trained a deep learning model on over 100,000 annotated objects. The resulting tool performs two core functions: it detects all objects in an image and classifies them as neurons or non-neurons with 95% accuracy. Independent benchmarking against expert human annotations demonstrated strong concordance, supporting its potential for routine use in neuropathology workflows.

The paper is published in the Alzheimer’s & Dementia® journal, and the The model is open-source and available here.