Machine learning in a real-world PFO study: analysis of data from multi-centers in China

• DOI: 10.1186/s12911-022-02048-5
• Publication Date: 2022-11-25T08:02:55Z
• Authors: Dongling Luo; Ziyang Yang; Gangcheng Zhang; Qunshan Shen; Hongwei Zhang; Junxing Lai; Hui Hu; Jianxin He; Shulin Wu; Caojin Zhang
• Keywords: Male; Adult; China; Research; Computer applications to medicine. Medical informatics; R858-859.7; Patent foreman ovale; 3. Good health; Machine Learning; Stroke; 03 medical and health sciences; 0302 clinical medicine; Ischemic Attack, Transient; Device closure; Machine learning; Humans; Cluster Analysis; Female; Recurrent stroke; Transient ischemic attack

Abstract Purpose The association of patent foreman ovale (PFO) and cryptogenic stroke has been studied for years. Although device closure overall decreases the risk recurrent stroke, treatment effects varied across different studies. In this study, we aimed to detect sub-clusters in post-closure PFO patients identify potential predictors adverse outcomes. Methods We analyzed with embolic undetermined sources from 7 centers China. Machine learning Cox regression analysis were used. Results Using unsupervised hierarchical clustering on principal components, two main clusters identified a total 196 included. average age was 42.7 (12.37) years 64.80% (127/196) female. During median follow-up 739 days, 12 (6.9%) events happened, including 6 (3.45%) 5 (2.87%) transient ischemic attack (TIA) one death (0.6%). Compared cluster 1 ( n = 77, 39.20%), 2 119, 60.71%) more likely be male, had higher systolic diastolic blood pressure, body mass index, lower high-density lipoprotein cholesterol increased proportion presence atrial septal aneurysm. random forest survival (RFS) analysis, eight top ranking features selected used prediction model construction. As result, RFS outperformed traditional (C-index: 0.87 vs. 0.54). Conclusions There patients. Traditional cardiovascular profiles remain future recurrence or TIA. However, whether maximizing management these factors would provide extra benefits warrants further investigations.