Establishment of a predictive model for hematoma enlargement after intracerebral hemorrhage based on machine learning and CT radiomics
10.19845/j.cnki.zfysjjbzz.2025.0102
- VernacularTitle:基于CT放射组学的脑出血后血肿扩大机器学习预测模型研究
- Author:
Ying LIU
1
;
Jie HE
2
Author Information
1. 国家电网公司北京电力医院神经内科，北京 100073
2. 浙江大学医学院附属邵逸夫医院放射科，浙江 杭州 310016
- Publication Type:Journal Article
- Keywords:
Spontaneous intracerebral hemorrhage;
Hematoma expansion;
Machine learning;
Radiomics;
Predictive model
- From:
Journal of Apoplexy and Nervous Diseases
2025;42(6):540-544
- CountryChina
- Language:Chinese
-
Abstract:
Objective To establish a model for accurate prediction of hematoma expansion（HE）following spontaneous intracerebral hemorrhage（sICH）based on CT radiomics and different machine learning algorithms. Methods A retrospective analysis was performed for the patients with sICH who were admitted to Sir Run Run Shaw Hospital and Beijing Electric Power Hospital in 2019-2022，and they were established as the training cohort and the external validation cohort，respectively. Radiomic features of hematomas were extracted from CT images and were combined with related clinical risk factors for HE to establish predictive models using four different machine learning algorithms. The receiver operating characteristic（ROC）curve was used to assess the predictive performance of each model，and the area under the ROC curve（AUC）and corresponding confusion matrix metrics were calculated. Results Compared with the other three models，the support vector machine（SVM）model showed the best predictive performance，achieving an AUC of 0.844 in the training cohort and 0.994 in the external validation cohort. In addition，the SVM model achieved the highest F1 scores（0.891 and 0.989），accuracy rates（83.8% and 98.4%），precision rates（94.5% and 99.9%），and specificities（82.4% and 99.9%）. Conclusion The radiomics model based on SVM provides a noninvasive tool for accurately predicting the risk of HE，which can help clinicians to identify high-risk sICH patients in the early stage and adjust treatment strategies in a timely manner.
- Full text:202507171400129077基于CT放射组学的脑出血后血肿扩大机器学习预测模型研究.pdf
