Researchers unveiled ALLocate, a low-cost AI microscope plugin designed to automate acute myeloid leukemia detection in bone marrow aspirate smears. The system couples a 3D-printed scanning stage (reported around $150) with deep learning models that first locate diagnostically adequate regions and then detect individual nucleated blasts. In evaluation, the region classifier reportedly achieved an AUROC of 0.992 for identifying adequate diagnostic regions. The cell detection model reported precision of 0.84, recall of 0.85, and mAP50 of 0.90, supporting end-to-end slide-level diagnosis without requiring high-cost whole-slide scanners. For settings with limited pathology infrastructure, the work targets speed and standardization—key constraints in hematologic diagnostics.
Get the Daily Brief