PRIMARY EVIDENCE · PEER-REVIEWED
Primary evidence
Chalise S, Roshal M, Gao Q, et al. Deep Learning Enables Automated Detection of Mature B Cell Neoplasms by Flow Cytometry. Modern Pathology. Published online 11 August 2026. DOI: 10.1016/j.modpat.2026.101062.
What it actually demonstrates
The study evaluated 3,070 clinical specimens: 1,369 peripheral blood, 316 bone marrow and 1,385 tissue samples. Reported case-level accuracy was 97.5%, sensitivity 95.6% and specificity 98.9%. These are study-cohort results, not performance guarantees for a new laboratory.
Evidence boundary
Evidence boundary: these results describe the study cohort and architecture. They should not be interpreted as prospective multicentre validation or as proof of equivalent performance in another laboratory, panel or workflow.
Primary record →Clinical-scale datasets matter because flow-cytometry AI has often been demonstrated on small or highly curated cohorts. Larger routine datasets provide a better test of biological diversity, technical variability and disease prevalence.
Strong classification performance is encouraging, but implementation requires more: prospective performance, subgroup analysis, failure-mode review, traceable preprocessing and a clear definition of how the model interacts with expert interpretation.
The most useful near-term role may be decision support and triage rather than autonomous diagnosis — particularly where the system can highlight abnormal populations and make its evidence inspectable.
CFCM view
Clinical value depends on reproducibility, transparent limitations and evidence in the intended workflow. Technical novelty is the beginning of validation, not its endpoint.
CFCM is an independent publication. References to manufacturers, instruments, reagents, software or therapies do not constitute endorsement. Technical content should be interpreted in the context of local validation, applicable regulation and manufacturer instructions for use.