Interactive multimodal disease risk evaluation using a self-service non-invasive device.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Diagnostic Technologies

Background:

• Non-communicable diseases (NCDs) like diabetes, fatty liver, and kidney disease pose significant global health challenges.
• Current diagnostic methods for NCDs are often slow, resource-intensive, and unsuitable for widespread screening.

Purpose of the Study:

• To develop an interactive, multimodal system for early-stage NCD risk evaluation.
• To assess the efficacy of a novel, non-invasive disease detection approach.

Main Methods:

• An interactive, self-service device was developed to capture facial images, tongue images, and pulse signals.
• A representation learning framework was utilized to analyze cross-modal dependencies for disease prediction.
• The system was validated on a cohort of 2,423 participants across nine common NCDs using 10 repeated train-test splits.

Main Results:

• The multimodal system achieved a mean diagnostic accuracy of 92.64% across repeated evaluations.
• The interactive multimodal approach outperformed single-modality baselines by an average of 3.39%.
• Analysis confirmed that complementary signals from different modalities significantly enhance predictive capabilities.

Conclusions:

• Combining non-invasive sensing with an interactive multimodal framework enables accessible and scalable disease risk evaluation.
• The developed system (iMES) is a deployable solution for decentralized NCD prevention in community and low-resource settings.