Interpretable machine learning model to predict rupture of small intracranial aneurysms and facilitate clinical decision

Abstract

Estimating the rupture risk of small intracranial aneurysms (IAs) to determine whether to treat is di cult but crucial. We aimed to construct and external validation a convenient machine learning (ML) model for assessing the rupture risk of small IAs.1004 patients with small IAs recruited from two hospitals were included in our retrospective research. The patients at hospital 1 were strati ed into training (70%) and internal validation set (30%) randomly, and the patients at hospital 2 were used for external validation. We selected predictive features using the least absolute shrinkage and selection operator (LASSO) method, and constructed ve ML models applying diverse algorithms including random forest classi er (RFC), categorical boosting (CatBoost), support vector machine (SVM) with linear kernel, light gradient boosting machine (LightGBM) and extreme gradient boosting (XGBoost). The Shapley Additive Explanations (SHAP) analysis provided interpretation for the best ML model.The training, internal and external validation cohorts included 658, 282, and 64 IAs, respectively. The best performance was presented by SVM as AUC of 0.817 in the internal [95% con dence interval (CI), 0.769-0.866] and 0.893 in the external (95% CI, 0.808-0.979) validation cohorts, overperformed than the PHASES score signi cantly (all P < 0.001). SHAP analysis showed maximum size, location and irregular shape were the top three important features to predict rupture. Our SVM model based on readily accessible features presented satisfying ability of discrimination in predicting the rupture IAs with small size. Morphological parameters made important contributions to prediction result.

Three-Dimensional Morphological Change of Intracranial Aneurysms Before and Around Rupture

“…Many morphology-based prediction or classification models for aneurysm instability compare the morphology of ruptured aneurysms with that of unruptured aneurysms to identify predictors for aneurysm rupture status or instability. [19][20][21][22][23][24] The underlying assumption is that the morphology of ruptured aneurysms resembles that of unruptured aneurysms. Our data suggest that the morphology of ruptured aneurysms is specific for the ruptured aneurysm, or at best represents the morphology shortly before rupture, while postrupture morphology is likely to differ significantly from the morphology of the baseline scan that would ideally be used to make predictions about future instability.…”

Nomograms for assessing the rupture risk of anterior choroid artery aneurysms based on clinical, morphological, and hemodynamic features

“…Machine learning has played a pivotal role in the extensive identification and assessment of rupture risk in IAs. Through a comprehensive analysis of various algorithms, Ou et al ( 14 ) and Xiong et al ( 15 ) have substantiated that the predictive accuracy of machine learning significantly surpasses that of logistic regression models and scoring systems. Conversely, a multicenter study conducted in China revealed that traditional logistic regression is not inferior to machine learning algorithms in multidimensional models for predicting the rupture status of unruptured IAs ( 42 ).…”

Role of artificial intelligence and machine learning in the diagnosis of cerebrovascular disease

“…Furthermore, ML may also be particularly useful for smaller-sized aneurysms (≤7 mm in diameter), for which the risk of rupture or growth is often harder for clinicians to predict (Ahn et al, 2021;Lee et al, 2021;Xiong et al, 2022). This has been demonstrated in a recent systematic review using meta-regression that found no significant predictors for small aneurysm growth or size (Lee et al, 2021).…”

“…The authors noted that SVM outperformed the PHASES score in predicting aneurysm rupture with an AUC of 0.817 and 0.893 in the internal and external validity cohorts, respectively. Through the use of ML, the authors concluded that maximum size, location, and irregular shape of the IAs were the major predictors of aneurysmal rupture (Xiong et al, 2022).…”