A machine learning predictive model based on conventional two-dimensional echocardiography and serum biomarkers for early detection of ascending aorta dilation in BAV patients

Objective: To compare echocardiographic parameters and lipid biomarker levels between patients with bicuspid aortic valve (BAV) associated with ascending aorta dilation and those without dilation using machine learning models, and to identify early diagnostic features for BAV-associated dilation. Methods: A retrospective cohort of 51 BAV patients (25 with ascending aorta dilation and 26 without) from the Ningxia Medical University Affiliated Hospital (January 2024–November 2024) was enrolled. Routine echocardiography and lipid profile measurements were performed. Lasso regression was applied to select key features, construct a predictive model, and develop a nomogram for individualized risk assessment. Results: BAV dilation group exhibited significantly higher AAoV and AAoMPG than the non-dilation group (P < 0.05 for both). HDL levels were also elevated in the dilation group (P < 0.05). Lasso regression identified five critical features: age, HDL, ApoB, LVMI, and AAoV, with the model formula: Y = −0.02678133 + (0.10741491 × age) + (0.25612267 × HDL) + (−0.12903915 × ApoB) + (0.14253473 × LVMI) + (0.49063628 × AAoV). The model achieved an AUC of 0.825 (95% CI: 0.694–0.933). Conclusion: This study successfully identified key features using Lasso regression and developed an efficient predictive model for BAV-associated ascending aorta dilation. The model demonstrates strong diagnostic performance, interpretability, and clinical applicability. The nomogram provides a practical tool for individualized risk prediction, aiding early diagnosis, intervention decisions, and patient follow-up.