Machine learning-based delimitation of middle temporal gyrus layers and histopathological features. (A) Representative image of the middle temporal gyrus (MTG) subjected to duplex immunohistochemistry for Aβ (brown) and IBA1 (blue). A deep learning convolutional neural network was trained to delineate layer 1 (L1, orange), layer 2 (L2, cyan), layer 3 (L3, yellow), layer 4 (L4, purple), and layers 5 and 6 (L5/6, dark blue) of the MTG. (B) Higher-resolution view of the region outlined in A, illustrating MTG histopathology across the layers. (C) Higher-resolution view of the region outlined in B, highlighting Aβ plaques and IBA1-positive cells. (D) Machine learning pipeline outputs, identifying IBA1-positive cells (red), Aβ plaques (green), and the co-localization of IBA1-positive cell inclusions within Aβ plaques (yellow). Images and datasets were obtained from the Seattle Alzheimer’s Disease Brain Cell Atlas public database https://portal.brain-map.org/explore/seattle-alzheimers-disease. Scale bars: 1 mm (A), 150 μm (B), 20 μm (C-D).