iwMDS 2025 | Leveraging AI to enrich prognostication in MDS and MPNs

The works that we’ve done in the past years indicate that leveraging artificial intelligence to analyze these large databases of patients whose baseline characteristics and outcomes are well characterized and they have long follow-up in the real world enhances prognostication. It provides improved prediction of their outcomes, and that’s because the training made by the artificial intelligence is non-parametric. It doesn’t rely on predefined statistical ideas, statistical models, and that makes it fit better the reality of our patients. So thanks to that, we obtain better risk stratification models for chronic myeloid malignancies. These models are not only good in our training cohorts but they are very well generalizable so they are reproducible in other cohorts and they keep beating other risk scoring methods. Furthermore, these models, thanks to the AI-driven integration, do actually outperform some models that are molecularly enriched. So molecularly enriched models have the problem that they are difficult to implement, particularly in low and medium resource settings because they need NGS and that’s somehow limiting their applicability. And if you want to broaden the risk stratification to a wide community of patients living in different social backgrounds, leveraging AI to enrich prognostication based on readily applicable parameters is very important.

This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.