Bayesian inquiry improves accuracy in multimodal diagnostic inference
Auditable Hierarchical Inference with Multimodal Evidence and Bayesian Active Inquiry
Machine Learning
Summary
Most systems that try to identify medical conditions using multiple signals assume all information is available right away, but real clinical checks happen step by step. The authors created a fake test setup that separates guessing the diagnosis from asking additional questions. Their system uses a smart questioning method to narrow down possibilities more accurately than fixed question orders, although it sometimes asks more questions. These results show a repeatable way to test such systems but do not prove they work in real medical settings.
What this means in practice
- •For clinical decision support teams: Create diagnostic workflows that actively select questions to increase accuracy of syndrome family identification from multiple medical data types.
- •For medical device developers: Integrate Bayesian cost-sensitive active inquiry methods into multimodal medical devices for improved diagnostic confidence and question efficiency.$Commercial implications: Enables development of smarter diagnostic tools that adapt questioning strategies for better accuracy, attractive to healthcare technology companies.
Tested on simulated data.
Authors
Yucheng Liu, Everett X. Wang
Abstract
Most multimodal classifiers assume that all useful evidence is available upfront, whereas clinical inquiry is sequential. We present a synthetic evaluation framework separating objective multimodal inference from active inquiry. An objective model maps synthetic tongue, bilateral multipressure pulse, and acoustic observations to eight syndrome families. A provenance-bearing Knowledge Core and training-only answer model support Bayesian narrowing over 33 fine terms using 25 typed questions, while seven safety questions remain outside the diagnostic posterior. Weak shared symptom factors limit fine-term influence on objective phenotypes, and an objective-only probe tests leakage. In a preregistered 600-case synthetic evaluation, objective family accuracy was 87.0%. Cost-sensitive expected information gain (cost-EIG) raised family accuracy to 95.5% and exact fine-term accuracy to 42.5%, versus 92.7% and 33.0% for an unoptimized fixed-order baseline. Selective family accuracy reached 96.9% at 95.3% coverage. However, cost-EIG asked more questions on average than fixed order (9.795 versus 9.352). Missing-audio routing retained 94.0% family accuracy, and a workflow invariant enforced abstention when all objective modalities were absent. Permuting fine-term answer likelihoods reduced family accuracy to 78.7%. Five of six frozen checks passed, but the question-efficiency hypothesis failed. Comparisons are descriptive because only aggregate, not paired case-level, outputs were retained. These results establish a reproducible synthetic workflow, not clinical diagnostic or treatment validity.