Optimizing Coil Selection for Cerebral Aneurysm Treatment Using PyRadiomics and Machine Learning Models

This study presents an innovative method to increase the accuracy of coil selection for treating cerebral aneurysms, leveraging advanced image analysis and machine learning models. We examined 273 cases of saccular cerebral aneurysms treated at The Jikei University School of Medicine. The focus was on using comprehensive feature extraction from 3D medical images to predict the optimal size and length of the initial coil for endovascular coil embolization. Five machine learning regression models were developed and assessed using a 5-fold cross-validation technique. The models demonstrated high accuracy in predicting coil dimensions, with notable improvements observed when incorporating radiological texture features alongside morphological data. The research highlights the potential of integrating advanced image analysis techniques with machine learning to refine treatment strategies in cerebrovascular interventions, reduce the subjectivity in manual image analysis and improve clinical outcomes.