Combinations of clinical factors and radiomics for the hemorrhage-free survival of brainstem cavernous malformation patients.

OBJECTIVES

This study aimed to develop a MRI-based prognostic nomogram including radiomics and clinical information for brainstem cavernous malformation patients (BSCMs).

METHODS

114 BSCMs were randomly divided into a training cohort and a validation cohort. Clinical and radiomics nomograms were constructed. Radiomics features were selected using three algorithms: univariate analysis, Pearson correlation, and the elastic net algorithm. A Cox regression model was employed to build the radiomics nomogram. The concordance index (C-index), time-independent receiver operating characteristic (ROC) analysis, and decision curve analysis (DCA) were used to evaluate the clinical utility of the nomogram.

RESULTS

The radiomics signature score was calculated using 11 radiomics features related to hemorrhage-free survival (HFS) from the training cohort. Patients were stratified into high-risk and low-risk groups based on the radiomics signature, with the low-risk group demonstrating significantly better HFS. Additionally, three clinical factors-number of hemorrhages, lesion size, and modified Rankin Scale score-along with the radiomics score were used to develop the radiomics nomogram. Calibration plots indicated good agreement between predicted and actual survival probabilities. The C-index for the training and validation cohorts was 0.784 and 0.787, respectively, in predicting HFS. The area under the curve was 72.51 and 76.41 for 3-year survival, and 67.62 and 72.57 for 5-year survival in the training and validation cohorts, respectively. The DCA curve demonstrated that the radiomics nomogram had superior clinical utility compared to the clinical model.

CONCLUSIONS

The radiomics nomogram showed great potential as a sensitive prognostic tool in predicting hemorrhage-free survival in BSCMs.