NOAH: Learning the Full Patient Journey. A Longitudinal Multimodal Time-Aware Model for Representation and Forecasting
NOAH is presented as a generative transformer for forecasting full multimodal patient trajectories, not just labeling fixed clinical records.
The paper says the model was built from more than 559 million clinical events across 431,000 hospital visits and 299,000 patients in the MIMIC dataset family. It handles images, time-series and numeric signals, categorical events, and structured and unstructured clinical records. Its design uses bidirectional time integration and a variational latent space to model irregular timing and uncertainty in patient courses. The authors report strong probing performance across clinical outcomes, ICD chapters, comorbidities, and time-to-event prediction. ArXiv · AI/CL/LG's note
The paper says the model was built from more than 559 million clinical events across 431,000 hospital visits and 299,000 patients in the MIMIC dataset family. It handles images, time-series and numeric signals, categorical events, and structured and unstructured clinical records. Its design uses bidirectional time integration and a variational latent space to model irregular timing and uncertainty in patient courses. The authors report strong probing performance across clinical outcomes, ICD chapters, comorbidities, and time-to-event prediction. ArXiv · AI/CL/LG's note
score 4