[Development and validation of a risk prediction model for cryptogenic stroke secondary to patent foramen ovale in young people based on multimodal ultrasound].

<b>Objective:</b> To construct a risk prediction model for cryptogenic stroke (CS) secondary to right-to-left shunt (RLS) caused by patent foramen ovale (PFO) in young people using combined transthoracic echocardiography with contrast (cTTE) and transesophageal echocardiography (TEE), and to explore its predictive efficacy. <b>Methods:</b> A retrospective analysis was conducted on clinical data from young patients diagnosed with RLS due to PFO by cTTE and TEE at Xiangya Third Hospital of Central South University between January 2016 and November 2024. Patients were divided into stroke and non-stroke groups based on whether they had CS. cTTE and TEE parameters were compared between the two groups. Multivariate logistic regression was used to identify the relevant factors of CS occurrence, and a predictive model was constructed with a nomogram. Internal validation was performed using the bootstrap method with 1 000 resamplings. Model performance and clinical utility were evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis. <b>Results:</b> A total of 82 patients were included, including 38 males and 44 females, aged [<i>M</i>(<i>Q</i>₁, <i>Q</i>₃)] 20 (17, 22) years. There were 37 cases in the stroke group and 45 cases in the non-stroke group. No statistically significant differences in age or gender were observed between the two groups (all <i>P</i>>0.05). Multivariate logistic regression analysis revealed that a larger left atrial opening diameter of the PFO during the Valsalva maneuver on cTEE (<i>OR</i>=3.880, <i>95%CI</i>: 1.005-14.982) and a longer PFO tunnel length (<i>OR</i>=1.311, <i>95%CI</i>: 1.042-1.649) were risk factors for CS. Conversely, a smaller PFO-inferior vena cava angle (<i>OR</i>=0.821, <i>95%CI</i>: 0.726-0.928) was identified as a protective factor against CS. Based on these findings, a predictive model was developed and visualized as a nomogram. The area under the curve (AUC) of the modeling group was 0.928 (95%<i>CI</i>: 0.873-0.983), and the AUC of the internal validation group was 0.932(95%<i>CI</i>: 0.870-0.976). The calibration curve and clinical decision curve showed good consistency and benefit of the model. <b>Conclusion:</b> The risk prediction model constructed using multimodal ultrasound in young PFO with RLS patients has a certain predictive value for the risk of CS occurrence.