Details

Autor(en) / Beteiligte

• Zhu, Haoyu
• Liu, Lian
• Liang, Shikai
• Ma, Chao
• Chang, Yuzhou
• Zhang, Longhui
• Fu, Xiguang
• Song, Yuqi
• Zhang, Jiarui
• Zhang, Yupeng
• Jiang, Chuhan

Titel

Rupture risk assessment in cerebral arteriovenous malformations: an ensemble model using hemodynamic and morphological features

Ist Teil von

• Journal of neurointerventional surgery, 2024-08, p.jnis-2024-022208

Erscheinungsjahr

2024

Beschreibungen/Notizen

• Background Cerebral arteriovenous malformation (AVM) is a cerebrovascular disorder posing a risk for intracranial hemorrhage. However, there are few reliable quantitative indices to predict hemorrhage risk accurately. This study aimed to identify potential biomarkers for hemorrhage risk by quantitatively analyzing the hemodynamic and morphological features within the AVM nidus. Methods This study included three datasets comprising consecutive patients with untreated AVMs between January 2008 to December 2023. Training and test datasets were used to train and evaluate the model. An independent validation dataset of patients receiving conservative treatment was used to evaluate the model performance in predicting subsequent hemorrhage during follow-up. Hemodynamic and morphological features were quantitatively extracted based on digital subtraction angiography (DSA). Individual models using various machine learning algorithms and an ensemble model were constructed on the training dataset. Model performance was assessed using the confusion matrix-related metrics. Results This study included 844 patients with AVMs, distributed across the training (n=597), test (n=149), and validation (n=98) datasets. Five hemodynamic and 14 morphological features were quantitatively extracted for each patient. The ensemble model, constructed based on five individual machine-learning models, achieved an area under the curve of 0.880 (0.824–0.937) on the test dataset and 0.864 (0.769–0.959) on the independent validation dataset. Conclusion Quantitative hemodynamic and morphological features extracted from DSA data serve as potential indicators for assessing the rupture risk of AVM. The ensemble model effectively integrated multidimensional features, demonstrating favorable performance in predicting subsequent rupture of AVM.