Intervention-Aware Clinical World Model for Post-Op Outcome Forecasting in Cardiology
The model updates recurrence risk through the messy 90-day post-ablation recovery window, not just from baseline data.
The paper proposes a structured latent patient state initialized from baseline imaging, then revised as procedures, covariates, elapsed time, physiological data, and irregular post-op events arrive. It is tested on atrial fibrillation ablation data from DECAAF-II, where it reports AUROC 0.756 and AUPRC 0.777 for recurrence prediction. The model also forecasts scar extent with a 2.971 percentage-point MAE without needing follow-up MRI intensities at inference. ArXiv · AI/CL/LG's note
The paper proposes a structured latent patient state initialized from baseline imaging, then revised as procedures, covariates, elapsed time, physiological data, and irregular post-op events arrive. It is tested on atrial fibrillation ablation data from DECAAF-II, where it reports AUROC 0.756 and AUPRC 0.777 for recurrence prediction. The model also forecasts scar extent with a 2.971 percentage-point MAE without needing follow-up MRI intensities at inference. ArXiv · AI/CL/LG's note
score 4