Prediction of bone marrow fibrosis from complete blood count in myeloproliferative neoplasms (FIBOM-AI): a multicentre machine learning model development and validation study - PubMed
3 hours ago
- #Myeloproliferative Neoplasms
- #Machine Learning
- #Non-invasive Diagnosis
- Developed FIBOM-AI, a machine learning model predicting grade 2-3 bone marrow fibrosis in myeloproliferative neoplasms using only complete blood count (CBC) data and age.
- The model was trained and validated on retrospective data from 1995 patients across 13 French hospitals and prospectively tested on 493 patients from 5 centers.
- The Extreme Gradient Boosting algorithm achieved high performance, with AUCs of 0.96 (training), 0.90 (testing), and 0.92 (validation), and prospective accuracy of 85.2% for overall predictions and 98.6% for confident predictions.
- Designed to serve as a non-invasive triage tool to prioritize bone marrow biopsies or provide risk assessment when biopsies are not immediately feasible.