Intelligent Recognition of Phlegm-Dampness Syndrome in Hypertension Based on Multimodal Feature Fusion
- VernacularTitle:基于多模态特征融合的高血压痰湿证智能辨识
- Author:
Yingfei LIU
1
;
Wei SHI
;
Quan LIU
;
Ying YANG
;
Xin GAO
Author Information
- Publication Type:Journal Article
- Keywords: Hypertension; Phlegm-dampness syndrome; Attention mechanism; Multimodal fusion
- From: World Science and Technology-Modernization of Traditional Chinese Medicine 2025;27(8):2183-2191
- CountryChina
- Language:Chinese
- Abstract: Objective To construct a diagnostic model for phlegm-dampness syndrome in hypertension based on multimodal feature fusion.Methods Clinical text information(physiological/lifestyle data,symptoms,pulse characteristics)and tongue image data were collected from 261 hypertension patients.For clinical text data,statistical analyses,including ANOVA,Mann-Whitney U test,and Chi-square test,were performed to identify significant features(P<0.05),which were incorporated into a clinical text-based diagnosis model(CTDM)using a multilayer perceptron algorithm.For tongue images,a tongue image-based diagnosis model(TIDM)was constructed based on channel attention mechanisms and residual networks.A multimodal diagnostic model(MDM)was built by fusing clinical text and tongue image features using a feature concatenation method.The diagnostic performance of each model was evaluated using five-fold cross-validation with the area under the receiver operating characteristic curve(AUC),accuracy,specificity,and sensitivity.Results Seven clinical text features,including physiological/lifestyle factors(disease duration,body mass index),symptoms(chest tightness,loss of appetite,excessive phlegm),and pulse characteristics(slippery pulse,damp pulse),were identified as risk factors for phlegm-dampness syndrome in hypertension.The AUC of the CTDM was 0.831±0.021,the AUC of the TIDM was 0.878±0.035,and the MDM achieved an AUC of 0.972±0.015.Conclusion The multimodal diagnostic model that integrates clinical text and tongue image features demonstrates high diagnostic accuracy and provides valuable guidance for AI-assisted diagnosis of phlegm-dampness syndrome in hypertension.
