Bayesian Integration of Nonlinear Incomplete Clinical Data

TL;DR

BIONIC is a Bayesian framework that effectively integrates heterogeneous multimodal clinical data with missingness, improving predictive performance and interpretability in partially observed settings.

Contribution

It introduces a unified probabilistic model that combines pretrained embeddings and structured variables, explicitly handling missing data in multimodal clinical datasets.

Findings

Strong discriminative performance on biomedical datasets

Robustness under incomplete data scenarios

Provides interpretability through latent structure

Abstract

Multimodal clinical data are characterized by high dimensionality, heterogeneous representations, and structured missingness, posing significant challenges for predictive modeling, data integration, and interpretability. We propose BIONIC (Bayesian Integration of Nonlinear Incomplete Clinical data), a unified probabilistic framework that integrates heterogeneous multimodal data under missingness through a joint generative-discriminative latent architecture. BIONIC uses pretrained embeddings for complex modalities such as medical images and clinical text, while incorporating structured clinical variables directly within a Bayesian multimodal formulation. The proposed framework enables robust learning in partially observed and semi-supervised settings by explicitly modeling modality-level and variable-level missingness, as well as missing labels. We evaluate BIONIC on three multimodal clinical and biomedical datasets, demonstrating strong and consistent discriminative performance compared to representative multimodal baselines, particularly under incomplete data scenarios. Beyond predictive accuracy, BIONIC provides intrinsic interpretability through its latent structure, enabling population-level analysis of modality relevance and supporting clinically meaningful insight.