Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data-Driven Priors for Uncertainty-Aware Deterioration Risk Prediction with Multimodal Data

L. Julián Lechuga López, Tim G. J. Rudner, Farah E. Shamout

TL;DR

A predictive uncertainty framework that leverages multimodal clinical data for in-hospital risk prediction to improve model performance and reliability and the promise of data-driven priors in advancing robust, uncertainty-aware AI tools for high-stakes clinical applications is proposed.

Abstract

Safe predictions are a crucial requirement for integrating predictive models into clinical decision support systems. One approach for ensuring trustworthiness is to enable models' ability to express their uncertainty about individual predictions. However, current machine learning models frequently lack reliable uncertainty estimation, hindering real-world deployment. This is further observed in multimodal settings, where the goal is to enable effective information fusion. In this work, we propose $\texttt{MedCertAIn}$, a predictive uncertainty framework that leverages multimodal clinical data for in-hospital risk prediction to improve model performance and reliability. We design data-driven priors over neural network parameters using a hybrid strategy that considers cross-modal similarity in self-supervised latent representations and modality-specific data corruptions. We train and evaluate the models with such priors using clinical time-series and chest X-ray images from the publicly-available datasets MIMIC-IV and MIMIC-CXR. Our results show that $\texttt{MedCertAIn}$ significantly improves predictive performance and uncertainty quantification compared to state-of-the-art deterministic baselines and alternative Bayesian methods. These findings highlight the promise of data-driven priors in advancing robust, uncertainty-aware AI tools for high-stakes clinical applications.

Data-Driven Priors for Uncertainty-Aware Deterioration Risk Prediction with Multimodal Data

TL;DR

A predictive uncertainty framework that leverages multimodal clinical data for in-hospital risk prediction to improve model performance and reliability and the promise of data-driven priors in advancing robust, uncertainty-aware AI tools for high-stakes clinical applications is proposed.

Abstract

Safe predictions are a crucial requirement for integrating predictive models into clinical decision support systems. One approach for ensuring trustworthiness is to enable models' ability to express their uncertainty about individual predictions. However, current machine learning models frequently lack reliable uncertainty estimation, hindering real-world deployment. This is further observed in multimodal settings, where the goal is to enable effective information fusion. In this work, we propose , a predictive uncertainty framework that leverages multimodal clinical data for in-hospital risk prediction to improve model performance and reliability. We design data-driven priors over neural network parameters using a hybrid strategy that considers cross-modal similarity in self-supervised latent representations and modality-specific data corruptions. We train and evaluate the models with such priors using clinical time-series and chest X-ray images from the publicly-available datasets MIMIC-IV and MIMIC-CXR. Our results show that significantly improves predictive performance and uncertainty quantification compared to state-of-the-art deterministic baselines and alternative Bayesian methods. These findings highlight the promise of data-driven priors in advancing robust, uncertainty-aware AI tools for high-stakes clinical applications.
Paper Structure (16 sections, 14 equations, 2 figures, 4 tables)

This paper contains 16 sections, 14 equations, 2 figures, 4 tables.

Table of Contents

  1. Analysis of MedCertAIn Across Patient Subpopulations
  2. Discussion
  3. Conclusion
  4. Extended Results
  5. Comparison with Hard Example Mining
  6. Comparison Across Informative Multimodal Data-Driven Priors
  7. Derivation of Variational Objective
  8. A Family of Data-Driven Priors
  9. Data-Driven, Uncertainty-Aware Priors for Fine-Tuning Pre-trained Models
  10. Variational Inference with Uncertainty-Aware Priors
  11. Variational Objective.
  12. Experimental Details
  13. Dataset Training Details
  14. Training details of ConVIRT
  15. Stochastic Hyperparameter Space
  16. ...and 1 more sections

Figures (2)

  • Figure 4: Analysis of MedCertAIn Across Patient Subpopulations. Percentage improvement from MedCertAIn over the deterministic baseline, MedFuse. Our stochastic framework significantly improves selective prediction showing its adaptability over different patient subpopulation groups.
  • Figure B.1: Comparison Across Informative Multimodal Data-Driven Priors.a) Difference across baselines in standard metrics is almost negligible. b) The trends in selective prediction metrics show that MedCertAIn remains the most performing model compared to other ablations, showing that the combination of different context-set priors enhances uncertainty-ware predictions.