EUVIMEDEUVIMED entdecken
Crossref

Machine Learning‐Based Clustering of Clinical Phenotypes in Adults With Bicuspid Aortic Valve: A Large Single‐Center Study

Xinyue Zhang, Zhuo Chen, Zhaokai Xu, Tingting Liu, Lixin Zhang, Yihua He, Xiaoyan Gu

Echocardiography · 2026-11 · 43 (11)

DOI: 10.1111/echo.70674 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Echocardiography, online erschienen am 06.11.2026. Verfasst von Xinyue Zhang und 6 weiteren Autor:innen. Beteiligte Einrichtung: Echocardiography Medical Center, Beijing Key Laboratory of Maternal-Fetus Medicine in Fetal Heart Disease(BZ0308), Beijing Anzhen Hospital, Capital M…. Hauptthemen laut MEDLINE: Machine Learning, Bicuspid Aortic Valve Disease, Echocardiography, Aortic Valve. Die Arbeit nennt 21 Literaturangaben. Gefördert von: Beijing Anzhen Hospital High Level Research Funding.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

Bicuspid aortic valve (BAV) is clinically heterogeneous, and the relationships among valve morphology, valvular dysfunction, and aortic involvement remain incompletely understood. We aimed to identify clinical phenotypes of adults with BAV by clustering multidimensional clinical data, primarily echocardiographic parameters.

Methods

We retrospectively included adults with echocardiographically confirmed BAV at Beijing Anzhen Hospital between January 2016 and December 2024. Data from the first echocardiographic examination and contemporaneous clinical records were extracted, and nine continuous and three categorical variables were entered into a two‐step clustering model. Continuous and categorical variables were compared across clusters using the Kruskal‐Wallis H test and chi‐square test, respectively, followed by pairwise comparisons. The stability of the four‐cluster solution was assessed using 100 bootstrap resamples; agreement was quantified with the adjusted Rand index (ARI) and Jaccard coefficients. Clustering was also repeated after separately excluding sex and valve dysfunction to assess the influence of these key variables on cluster structure.

Results

A total of 6977 patients with echocardiographically diagnosed BAV were included. The median age was 53.0 years (interquartile range, 40.0–61.0), and 70.0% were men. Two‐step clustering identified four clinical phenotypes that differed significantly in valve morphology, valvular dysfunction, aortic dimensions, left‐heart remodeling, cardiac function, and surgical interventions (all p < 0.001): Cluster 1, a predominantly male, regurgitation‐predominant phenotype with left ventricular dilatation; Cluster 2, a predominantly male, stenosis‐predominant phenotype with concentric hypertrophy; Cluster 3, a male phenotype with no or mild valve dysfunction; and Cluster 4, a female phenotype characterized mainly by no or mild valve dysfunction with a tendency toward stenosis. The proportion undergoing subsequent surgery after the baseline echocardiographic assessment was highest in Cluster 2 (40%) and lowest in Cluster 3 (14%). Left ventricular end‐systolic diameter, left ventricular end‐diastolic diameter, aortic root diameter, valve dysfunction, and sex were the principal drivers of clustering. Across bootstrap resamples, the mean ARI for the four‐cluster solution was 0.696. In sensitivity analyses, exclusion of either sex or valve dysfunction reduced the optimal solution from four clusters to two.

Conclusions

Machine learning‐based two‐step clustering identified four baseline clinical phenotypes among adults with BAV, revealing multidimensional combinations of valve morphology, valvular function, aortic dimensions, and left‐heart remodeling. This classification is intended primarily to characterize the clinical heterogeneity of adult BAV; its longitudinal clinical value requires confirmation in independent prospective cohorts.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Machine LearningBicuspid Aortic Valve DiseaseEchocardiographyAortic ValveHeart Valve DiseasesHumansPhenotypeFemaleMaleRetrospective StudiesClustering AlgorithmsMiddle AgedAdultCluster Analysis

Stichwörter der Autor:innen

Bicuspid aortic valveclinical phenotypeechocardiographymachine learningtwo‐step clustering

Autor:innen und Institutionen

  • Echocardiography Medical Center, Beijing Key Laboratory of Maternal-Fetus Medicine in Fetal Heart Disease(BZ0308), Beijing Anzhen Hospital, Capital Medical University, Beijing, People's Republic of China.
  • College of Computer Science and Artificial Intelligence, Fudan University, Shanghai, People's Republic of China.

Förderung

  • Beijing Anzhen Hospital High Level Research Funding

Zitieren

APA

Zhang, X., Chen, Z., Xu, Z., Liu, T., Zhang, L., He, Y., & Gu, X. (2026). Machine Learning‐Based Clustering of Clinical Phenotypes in Adults With Bicuspid Aortic Valve: A Large Single‐Center Study. Echocardiography, 43(11). https://doi.org/10.1111/echo.70674

Vancouver

Zhang X, Chen Z, Xu Z, Liu T, Zhang L, He Y, et al. Machine Learning‐Based Clustering of Clinical Phenotypes in Adults With Bicuspid Aortic Valve: A Large Single‐Center Study. Echocardiography. 2026;43(11). doi:10.1111/echo.70674

BibTeX
@article{Zhang2026,
  author = {Zhang, Xinyue and Chen, Zhuo and Xu, Zhaokai and Liu, Tingting and Zhang, Lixin and He, Yihua and Gu, Xiaoyan},
  title = {Machine Learning‐Based Clustering of Clinical Phenotypes in Adults With Bicuspid Aortic Valve: A Large Single‐Center Study},
  journal = {Echocardiography},
  year = {2026},
  volume = {43},
  number = {11},
  doi = {10.1111/echo.70674}
}
Kostenlos starten →