CT影像组学模型预测原发性脑出血血肿扩大及不良结局的临床研究OA

Objective To verify the predictive value of radiomics features extracted from non-contrast CT for patients with spontaneous intracerebral hemorrhage of hematoma expansion and poor functional outcomes,and to compare it with imaging signs and clinical factors.Methods A total of 754 radiomics features extracted from 1 732 sets of scan data of patients admitted to the Second Affiliated Hospital of Jiangxi University of Traditional Chinese Medicine from January 2020 to January 2025 were analyzed retrospectively.The features were processed,and a correlation-based feature selection method was adopted to evaluate the predictive performance of the selected radiomics features.The area under curve(AUC)was used to assess the predictive efficacy of the model.Results The optimal radiomics model achieved an AUC of 0.693 for predicting hematoma expansion and 0.783 for predicting poor functional outcomes.The model using only imaging signs showed a significant decrease in sensitivity,and the combined model of radiomics features and imaging signs did not show advantages over the radiomics model alone.The model using only clinical factors had similar predictive performance to the radiomics model but with lower sensitivity for predicting hematoma expansion.Incorporating clinical factors into radiomics features improved model performance.Conclusions Radiomics features are superior to imaging signs and comparable to clinical factors in predicting hematoma expansion and poor functional outcomes in patients with intracerebral hemorrhage.In addition,the combination of radiomics features and clinical factors can further improve predictive efficacy.