基于多维特征的颅内动脉瘤风险预测与模型评估OA

Objective To accurately predict the risk of intracranial aneurysm rupture,facilitating early diagnosis and personalized treatment for patients.Methods By integrating morphological features(e.g.,maximum diameter,neck width,lobulation,volume)and hemodynamic characteristics(e.g.,wall shear stress,oscillatory shear index,pressure distribution)of aneurysms,feature data were extracted using Mimics and Ansys-Fluent software.A risk prediction model was developed based on the PyTorch framework,incorporating both traditional machine learning algorithms and deep learning models for intracranial aneurysm rupture risk prediction.Results The study demonstrated that machine learning models constructed using parallel tree boosting algorithms and random forests outperformed deep learning methods(represented by multilayer perceptrons)in predicting rupture risks,proving more suitable for intracranial aneurysm risk assessment.Conclusion The proposed aneurysm rupture risk prediction model can assist physicians in formulating more precise diagnostic and therapeutic plans,thereby improving treatment efficacy and patient safety.