Prediction of Aneurysm Stability Using a Machine Learning Model Based on PyRadiomics-Derived Morphological Features.

Area of Science:

• Neurosurgery
• Medical Imaging
• Artificial Intelligence

Background:

• Intracranial aneurysm stability assessment is crucial for treatment planning, particularly for small aneurysms.
• Radiomics offers a quantitative approach to analyze medical images for disease characterization.

Purpose of the Study:

• To evaluate the feasibility of using machine learning with radiomics-derived morphological features to predict intracranial aneurysm stability.
• To identify key morphological determinants of aneurysm stability.

Main Methods:

• Morphological features were extracted from 420 intracranial aneurysms (4-8 mm diameter) using PyRadiomics.
• Machine learning models were developed incorporating morphological and clinical data to predict aneurysm stability.
• Lasso regression identified significant features, and model performance was assessed using the area under the curve (AUC).

Main Results:

• Twelve morphological features were extracted; Flatness was the most significant predictor of aneurysm stability.
• The final prediction model, including features like Surface Area, Spherical Disproportion, Flatness, and clinical factors, achieved an AUC of 0.853.
• Unstable aneurysms showed lower compactness and sphericity, and higher Spherical Disproportion in hypertensive patients.

Conclusions:

• Radiomics-derived morphological features are valuable for intracranial aneurysm stratification and stability prediction.
• Flatness is a critical morphological determinant for predicting aneurysm stability.
• The developed machine learning model effectively predicts aneurysm stability, with morphology playing a key role, especially in hypertensive patients.