Predictive performance for prognosis of aneurysmal subarachnoid hemorrhage with ventricular hemorrhage by imaging combined with clinical and laboratory quantitative index model

Abstract

Objective: To explore the predictive performance of image quantitative index model, clinical-laboratory index model and image-clinical multi-dimensional fusion model in predicting the prognosis of patients with aneurysmal subarachnoid hemorrhage (aSAH) with intraventricular hemorrhage (IVH). Methods: A total of 349 patients with aSAH and IVH, including 122 males and 227 females, aged 22 to 85 (59±11) years underwent CT scan in the General Hospital of Eastern Theater Command from January 2010 to December 2019 were used as dataset 1 to construct a prognostic model. A prognostic model was constructed for data set 1, and the functional recovery of patients 12 months after discharge was evaluated using the modified Rankin Scale (mRS). According to the results, those patients were divided into two groups: good outcome group (n=267) and poor outcome group (n=82). In addition, 63 aSAH patients with IVH, including 27 males and 36 females, aged 32 to 87 (61±12) years who were admitted to the General Hospital of Eastern Theater Command from January 2020 to December 2021 were collected as dataset 2 for independent verification of the model, including 30 patients with poor prognosis. Clinical information (age and gender), laboratory indicators (blood routine and blood biochemistry), and imaging quantitative indicators (such as volume, density, shape of each ventricle hemorrhage area outlined and extracted on head CT scan etc.) were recorded for all patients (dataset 1 and 2). The clinical, laboratory and imaging quantitative indicators of dataset 1 were screened by using L1 regularization and multiple logistic regression method was used to construct the clinical-laboratory index model, image quantitative index model and image-clinical multi-dimensional fusion model, according to the weight coefficient of features in the clinical-laboratory index model and image quantitative index model, screen out the main features. The model was trained and internally validated by 5-fold cross-validation. The model was validated independently in dataset 2. Results: The AUC (area under the ROC curve) of clinical-laboratory index model, image quantitative index model and multidimensional fusion model constructed based on dataset 1 were 0.75 (95%CI: 0.69-0.81), 0.68 (95%CI: 0.61-0.74) and 0.86 (95%CI: 0.82-0.91). The Delong test showed that there were statistically significant differences between the performance of the multi-dimensional fusion model and the clinical-laboratory index model or image quantitative index model (all P<0.05). The AUC of clinical-laboratory index model, image quantitative index model and multidimensional fusion model of dataset 2 were 0.79 (95%CI: 0.68-0.91), 0.70 (95%CI: 0.57-0.83) and 0.81 (95%CI: 0.70-0.92). In addition, in the clinical-laboratory index model and imaging quantitative index model constructed based on data 1, age, Hunt-Hess grade on admission, Neutrophil/Lymphocyte (N/L) (the weight coefficients in the clinical-laboratory index model were 1.00, -0.59 and 0.44) and the standard deviation of third ventricle hemorrhage density, minimum hemorrhage density of the fourth ventricle, and left ventricle hemorrhage sphericity (the weight coefficients in the image quantitative index model were -1.00, 0.85 and -0.84) were the main features of the screening. Conclusions: Quantitative imaging indicators of ventricular hemorrhage (standard deviation of third ventricular hemorrhage density, minimum density of fourth ventricular hemorrhage, and left ventricular sphericity) are helpful to predict the poor prognosis of patients with aSAH with ventricular hemorrhage. Dimensional fusion model has greater value in predicting poor prognosis of patients.