A team at Memorial Sloan Kettering Cancer Center built ALLocate, an AI plugin for standard lab microscopes. It lets those instruments scan bone marrow slides for acute leukemia without a human at the eyepiece. The approach, described in Nature Communications, aims to bring automated screening to labs that cannot afford whole-slide scanners.
ALLocate attaches directly to conventional microscopes and analyzes slides on the instrument, eliminating the need for costly digitization hardware. It was trained on more than 11,000 annotated regions and 130,000 annotated cells, then validated across independent cohorts including 165 physical bone marrow smear slides collected with Memorial Sloan Kettering and UCSF hematopathology services.
In testing, the system achieved an area under the receiver operating characteristic curve above 0.99 for identifying regions of interest, a mean average precision of 0.90 for cell detection, and 88% accuracy for slide-level diagnosis on glass slides.
Acute leukemia diagnosis typically requires specialist review of bone marrow smears, a bottleneck in low-resource settings. The authors say the low-cost format could bridge the gap between AI innovation and practical deployment where access to expert pathologists is limited.
The three-level evaluation design, covering region identification, cell detection and end-to-end slide diagnosis, is intended to show the system works robustly at every step rather than only on curated images.
