Presentation summary

Functional dyspepsia (FD) is a chronic disorder with symptoms like fullness and pain without structural issues,affecting 8-46% of Koreans. It significantly impacts quality of life, and treatments are often incomplete, leadingto integrative approaches like acupuncture.

FD in Korean medicine involves syndrome differentiation using questionnaires for patterns like Pi-Wei Xu-Han. However, standardization in diagnosis is needed, as traditional methods are subjective. Pulse waves are key biosignals for hemodynamic changes.

While modern analysis technologies exist, aligning them with traditional interpretations is challenging. AI is being explored for better clinical insights. Research includes gastric and brain signals, with pulse waves offering insights for syndrome differentiation. Advances in non-invasive measurement enhance biosignal acquisition.

This study involved FD patients from Kang-dong Kyung-Hee University, using biosignals like pulse waveformsand fNIRS. Data preprocessing ensured quality for analysis. Feature selection and Random Forest modelswere used to evaluate performance, with LOOCV assessing metrics like accuracy and F1-score. Models combining biosignals and questionnaires performed best, highlighting the value of integrated data.

The study demonstrates the utility of a model combining questionnaires and biosignals for predicting syndromes infunctional dyspepsia. Integrating both questionnaires and biosignals (Model 1 accuracy: 0.450; F1: 0.429) outperforms using either questionnaires alone (Model 3 accuracy: 0.552; F1: 0.497) or biosignals alone (Model2 accuracy: 0.396; F1: 0.383) in terms of accuracy and F1-score. It suggests broader applications and future