A research team developed ALLocate, a low-cost AI microscope plugin aimed at automating acute myeloid leukemia detection from bone marrow aspirate smears. The system pairs a 3D-printed scanning stage (roughly $150 hardware) with deep learning models that first identify diagnostically adequate regions and then detect and classify nucleated cells. In evaluation, the region classifier achieved an AUROC of 0.992, while the cell detector reported precision 0.84, recall 0.85, and mAP50 0.90. The reported performance suggests feasibility for standardizing slide triage where whole-slide scanners are unavailable. If validated in broader clinical workflows, the platform could help reduce interpretive variability and speed diagnostic turnaround, particularly for high-volume hematology labs operating under resource constraints.
Get the Daily Brief