Machine learning-based quantitative analysis of bone marrow fibrosis in patients with MF

This study demonstrates that quantitative analysis using CIF may provide an objective assessment of fibrosis severity and heterogeneity within BMT samples in patients with MF. Additionally, CIF allows an objective comparison between sequential samples from individual patients and accurate comparisons within trial cohorts. 

There was a moderate correlation between CIF score and the conventional manual fibrosis grading for all samples (Spearman’s rho = 0.39); however, discrepancies were identified, with 38% of samples graded as MF-2 falling within the interquartile range of CIF distribution observed in MF-3, and 48% of MF-3 samples falling within the interquartile range observed in MF-2.  

Importantly, several samples graded as MF-2 manually had average CIF scores similar to those graded as MF-0 or MF-1. This suggests the need for more standardized, objective measures in clinical trials to improve the accuracy of therapeutic evaluations. 

While CIF score improvements did not directly correlate with secondary clinical outcomes, there was a trend toward association with best overall response, indicating potential utility as a surrogate marker for clinical efficacy in MF treatment trials.