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

作者：Qinglin Liu, Peng Jiang, Yuhua Jiang, Huijian Ge, ShaoLin Li, Hengwei Jin, YouXiang Li · 发表于：Stroke · 年份：2019 · DOI：10.1161/strokeaha.119.025777 · 被引用次数：164 · 研究领域：Intracranial Aneurysms: Treatment and Complications、Aortic aneurysm repair treatments、Cerebrovascular and Carotid Artery Diseases

Background and Purpose- Discrimination of the stability of intracranial aneurysms is critical for determining the treatment strategy, especially in small aneurysms. This study aims to evaluate the feasibility of applying machine learning for predicting aneurysm stability with radiomics-derived morphological features. Methods- Morphological features of 719 aneurysms were extracted from PyRadiomics, of which 420 aneurysms with Maximum3DDiameter ranging from 4 mm to 8 mm were enrolled for analysis. The stability of these aneurysms and other clinical characteristics were reviewed from the medical records. Based on the morphologies with/without clinical features, machine learning models were constructed and compared to define the morphological determinants and screen the optimal model for predicting aneurysm stability. The effect of clinical characteristics on the morphology of unstable aneurysms was analyzed. Results- Twelve morphological features were automatically extracted from PyRadiomics implemented in Python for each aneurysm. Lasso regression defined Flatness as the most important morphological feature to predict aneurysm stability, followed by SphericalDisproportion, Maximum2DDiameterSlice, and SurfaceArea. SurfaceArea (odds ratio [OR], 0.697; 95% CI, 0.476-0.998), SphericalDisproportion (OR, 1.730; 95% CI, 1.143-2.658), Flatness (OR, 0.584; 95% CI, 0.374-0.894), Hyperlipemia (OR, 2.410; 95% CI, 1.029-5.721), Multiplicity (OR, 0.182; 95% CI, 0.082-0.380), Location at midd...