Classifying Ruptured Middle Cerebral Artery Aneurysms With a Machine Learning Based, Radiomics-Morphological Model: A...

Area of Science:

• Medical Imaging
• Radiology
• Neurosurgery

Background:

• Aneurysm rupture is a critical event in cerebrovascular disease.
• Predictive models for aneurysm rupture often rely on morphological features.
• Multicenter studies on radiomics and morphological features for specific aneurysm locations, like middle cerebral artery (MCA) aneurysms, are limited.

Purpose of the Study:

• To identify robust radiomics features associated with middle cerebral artery (MCA) aneurysm rupture.
• To evaluate the added value of combining morphological and radiomics features for classifying ruptured MCA aneurysms.
• To develop and validate predictive models for MCA aneurysm rupture.

Main Methods:

• A multicenter study included 668 MCA aneurysms (423 ruptured) from 632 patients across five hospitals.
• Radiomics and morphological features were extracted from computed tomography angiography (CTA) images.
• Support vector machine (SVM) was used to develop radiomics (R-model), morphological (M-model), radiomics-morphological (RM-model), and clinical-radiomics-morphological (CRM-model) models, validated internally and externally.

Main Results:

• Seven radiomics features and four morphological predictors of MCA aneurysm rupture were identified.
• The radiomics-morphological model (RM-model) achieved an AUC of 0.848 (training), 0.865 (internal validation), and 0.721 (external validation).
• The combined clinical-radiomics-morphological model (CRM-model) demonstrated superior performance in internal datasets, with AUCs of 0.856 (training), 0.882 (internal validation), and 0.738 (external validation).

Conclusions:

• Specific radiomics features are robust predictors of MCA aneurysm rupture.
• The radiomics-morphological model shows significant potential for classifying ruptured MCA aneurysms.
• Integrating radiomics features into conventional models may enhance the accuracy of ruptured MCA aneurysm classification.