Abstract
Treatment-free remission (TFR) has become a therapeutic objective for selected chronic-phase chronic myeloid leukemia (CP CML). However, no standardized biomarker is yet in clinical use. In this multi-center study, we explored the potential of bone marrow (BM) cytomorphology given its global accessibility and integral role in clinical diagnostics. We included diagnostic BM aspirate samples of 113 CP CML patients from seven clinical sites having attempted first TKI discontinuation. We digitized cytomorphological slides into 100x-magnified high-resolution images and analyzed these with deep learning-based image analysis. We profiled the BM cytomorphological fingerprint of CP CML patients and recapitulated the known granulocytic predominance and reduction of lymphoid, monocytic and erythroid cells in comparison to an extensive cohort of 942 control BM samples. We discovered neutrophil abundance and granulocytic maturation to associate with sustained TFR. We confirmed these visually and demonstrated their independent impact over known clinical factors. Our study underlines the potential of computational BM cytomorphology to identify novel clinical biomarkers and suggests that granulocytic expansion and maturation at diagnosis could reflect intrinsic disease pathology influencing TFR maintenance.