Bone Marrow Biopsy Without the Needle? AI Risk Prediction for Myelofibrosis

In a robust, multi-center validation, the FIBOM-AI model predicted grade 2–3 myelofibrosis using only the humble CBC and age as inputs. For patients where biopsy is technically difficult or poorly tolerated—common scenarios in older adults and those with comorbidities—such a tool could meaningfully alter the diagnostic workup. Notably, the investigators offer both “overall” and “confident” prediction modes, allowing clinicians to choose higher sensitivity (to rule out fibrosis) or higher specificity (to rule in) depending on the clinical scenario. The model’s external and prospective real-world testing strengthens the case. Whether medicolegal and guideline authorities will accept AI as a biopsy surrogate remains open, and real practice will still wrestle with model limitations not apparent in trial cohorts, but the era of reflexive bone marrow sampling may be drawing to a close.