Machine learning-assisted radial pulse wave analysis as a novel diagnostic tool for menstrual disorders: a cross-sectional study

Background Menstrual disorders, prevalent among young women, correlate with endocrine abnormalities and cardiovascular risks. Current diagnosis relies on subjective assessment, lacking objective tools. While Traditional Chinese Medicine (TCM) pulse diagnosis reflects systemic status, its subjectivity limits clinical use. We developed an automated method integrating machine learning with radial pulse wave analysis. Methods A cross-sectional study involving 246 participants divided into a menstrual disorders group and a normal group. Pulse wave data and TCM constitution information were collected from the two groups. Between-group comparisons were performed using the Mann–Whitney U test, with Benjamini–Hochberg correction for multiple comparisons. All preprocessing, feature selection, and hyperparameter tuning were performed only within the training set using nested cross-validation. Six prespecified models were compared: LR-ElasticNet, LDA-Shrinkage, LR, SVM, DT, and XGBoost. Performance was evaluated using accuracy, recall, specificity, precision, F1 score, and AUC, each with 95% confidence intervals. Results Significant differences in pulse wave parameters and constitution types were found between the menstrual disorders group and normal. The menstrual disorders group predominantly showed a Yin deficiency constitution, with increased bilateral w, w/t, h3/h1 and left-side h5, but decreased bilateral h4/h1, left-side h3 and ad ( P < 0.05, FDR-adjusted q < 0.05). LDA-Shrinkage was selected as the best model, achieving AUC=0.977 (95% CI 0.919–1.000), accuracy=0.986, F1 = 0.989, precision=0.978, recall=1.000, specificity=0.966, with the most favorable calibration and interpretability among the six models. SHAP analysis identified Rw/t and Lh4/h1 as the most influential candidate features, proposing them as core observational parameters for future research. Conclusion Radial pulse wave analysis combined with machine learning showed promising exploratory performance for menstrual disorder screening. The observed TCM constitutional variations provide mechanistic insights into disease heterogeneity, supporting personalized treatment strategies. These findings advance the objective standardization of TCM pulse diagnosis while offering a non-invasive screening tool for clinical practice.

Authors

Institutions

Publication Details

Journal
Frontiers in Digital Health
Published
2026-11-16
DOI
https://doi.org/10.3389/fdgth.2026.1964508
Primary Topic
Traditional Chinese Medicine Studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine learning-assisted radial pulse wave analysis as a novel diagnostic tool for menstrual disorders: a cross-sectional study

Huangbo Lin, Shaoyang Men, Shuang Cheng, Ying Chen
1 links
Frontiers in Digital Health
Traditional Chinese Medicine Studies
article

Machine learning-assisted radial pulse wave analysis as a novel diagnostic tool for menstrual disorders: a cross-sectional study

Huangbo Lin, Shaoyang Men, Shuang Cheng, Ying Chen
article en

Abstract

Background Menstrual disorders, prevalent among young women, correlate with endocrine abnormalities and cardiovascular risks. Current diagnosis relies on subjective assessment, lacking objective tools. While Traditional Chinese Medicine (TCM) pulse diagnosis reflects systemic status, its subjectivity limits clinical use. We developed an automated method integrating machine learning with radial pulse wave analysis. Methods A cross-sectional study involving 246 participants divided into a menstrual disorders group and a normal group. Pulse wave data and TCM constitution information were collected from the two groups. Between-group comparisons were performed using the Mann–Whitney U test, with Benjamini–Hochberg correction for multiple comparisons. All preprocessing, feature selection, and hyperparameter tuning were performed only within the training set using nested cross-validation. Six prespecified models were compared: LR-ElasticNet, LDA-Shrinkage, LR, SVM, DT, and XGBoost. Performance was evaluated using accuracy, recall, specificity, precision, F1 score, and AUC, each with 95% confidence intervals. Results Significant differences in pulse wave parameters and constitution types were found between the menstrual disorders group and normal. The menstrual disorders group predominantly showed a Yin deficiency constitution, with increased bilateral w, w/t, h3/h1 and left-side h5, but decreased bilateral h4/h1, left-side h3 and ad ( P < 0.05, FDR-adjusted q < 0.05). LDA-Shrinkage was selected as the best model, achieving AUC=0.977 (95% CI 0.919–1.000), accuracy=0.986, F1 = 0.989, precision=0.978, recall=1.000, specificity=0.966, with the most favorable calibration and interpretability among the six models. SHAP analysis identified Rw/t and Lh4/h1 as the most influential candidate features, proposing them as core observational parameters for future research. Conclusion Radial pulse wave analysis combined with machine learning showed promising exploratory performance for menstrual disorder screening. The observed TCM constitutional variations provide mechanistic insights into disease heterogeneity, supporting personalized treatment strategies. These findings advance the objective standardization of TCM pulse diagnosis while offering a non-invasive screening tool for clinical practice.

Frontiers in Digital HealthVol. 8
Guangzhou University of Chinese Medicine (CN)
Good health and well-being
Traditional Chinese Medicine Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon