Prediction of Neurological Functional Recovery after Carotid Endarterectomy Using Machine Learning and Carotid...

Area of Science:

• Medical imaging analysis
• Machine learning in healthcare
• Cerebrovascular disease research

Background:

• Symptomatic carotid artery stenosis (SCAS) poses a risk for stroke.
• Carotid endarterectomy (CEA) is a common treatment for SCAS.
• Predicting neurological functional recovery post-CEA is crucial for patient management.

Purpose of the Study:

• To develop and validate machine learning models using carotid CTA radiomics.
• To predict neurological functional recovery in SCAS patients after CEA.
• To assess the performance of different machine learning algorithms and feature sets.

Main Methods:

• Retrospective analysis of 244 SCAS patients undergoing CEA.
• Classification of patients into good (n=164) and poor (n=80) recovery groups based on modified Rankin Scale.
• Extraction of radiomic features from carotid plaque on CTA images.
• Development of logistic regression, SVM, KNN, LightGBM, and MLP models using radiomic, clinical, and combined features.

Main Results:

• The SVM radiomics model achieved 90% accuracy, 91% sensitivity, and 88% specificity.
• LR and LightGBM radiomics models showed high predictive ability (88% and 86% accuracy).
• Combining radiomic and clinical features consistently improved model performance across all algorithms.

Conclusions:

• Machine learning models, especially SVM, LR, and LightGBM, effectively predict neurological recovery post-CEA.
• Combined radiomic and clinical features enhance predictive capabilities.
• These models can aid in postoperative risk stratification and personalized perioperative care.