Admissible diffusion improves treatment outcome prediction in clinical data
Admissible Diffusion for Multimodal Interventional Trajectories
Machine Learning
Summary
Predicting how medical treatments will affect a patient over time is tricky, especially with noisy or incomplete data. The authors developed ADMIT, a system that uses advanced math to create treatment scenarios while respecting medical knowledge and rules. This helps generate more accurate medical treatment paths, even when data is complex or irregular. Their tests show that combining different data types and enforcing realistic constraints improves prediction and keeps outcomes plausible.
What this means in practice
- •For clinical data scientists: Generate patient treatment scenarios that respect medical constraints to improve accuracy in outcome prediction.
- •For healthcare ai developers: Incorporate physiological prior knowledge in AI treatment models to produce plausible intervention paths under varying therapies.
Authors
Xing Han, Shravan Chaudhari, Jiarui Shao, Paul Pu Liang, Suchi Saria
Abstract
Generating a plausible clinical trajectory does not establish what would happen under a different treatment. We present ADMIT, a framework combining irregular multimodal representations, treatment-conditioned latent diffusion and explicit constraints on generated states or actions. We formulate its interventional target through sequential g-computation and distinguish causal assumptions from constraint satisfaction. Its admissibility mechanism translates physiological prior knowledge into explicit constraints on generated states and proposed actions. Treatment-exposure dynamics condition latent transitions, while state projection or action gating applies the constraints during rollout so that they influence subsequent trajectory generation. In our preliminary experiments, multimodal inputs improved supervised hidden-state recovery and reduced treatment-contrast error. In a simulated dosing-schedule experiment with leak-free history encoding, ADMIT predicted most of the tumor-volume change caused by redistributing a fixed total dose. An exposure input improved these predictions around a temporary dose reduction whether or not the assumed clearance rate was correct, but reduced the predicted size of a dose effect, and a deterministic recurrent baseline matched ADMIT's average predictions. Exposure projection reduced constraint violations, although enforcement remained incomplete. Semi-synthetic experiments using eICU context illustrated treatment-response generation under fixed and adaptive policies. Observational examples further characterize model treatment sensitivity. ADMIT provides a framework for testing whether complementary observations and physiological restrictions improve intervention trajectories, with representation recovery, effect accuracy and rule enforcement assessed separately.