RESUMO
PURPOSE: To explore imaging biomarkers predictive of intratumoral haemorrhage for lesions intended for elective stereotactic biopsy. METHOD: This study included a retrospective cohort of 143 patients with 175 intracranial lesions intended for stereotactic biopsy. All the lesions were randomly split into a training dataset ( n =121) and a test dataset ( n =54) at a ratio of 7:3. Thirty-four lesions were defined as "hemorrhage-prone tumors" as haemorrhage occurred between initial diagnostic MRI acquisition and the scheduled biopsy procedure. Radiomics features were extracted from the contrast-enhanced T1 Weighted Imaging and T2 Weighted Imaging images. Features informative of haemorrhage were then selected by the LASSO algorithm, and an Support Vector Machine model was built with selected features. The Support Vector Machine model was further simplified by discarding features with low importance and calculating them using a "permutation importance" method. The model's performance was evaluated with confusion matrix-derived metrics and area under curve (AUC) values on the independent test dataset. RESULTS: Nine radiomics features were selected as haemorrhage-related features of intracranial tumours by the LASSO algorithm. The simplified model's sensitivity, specificity, accuracy, and AUC reached 0.909, 0.930, 0.926, and 0.949 (95% CI: 0.865-1.000) on the test dataset in the discrimination of "hemorrhage-prone tumors". The permutation method rated feature "T2_gradient_firstorder_10Percentile" as the most important, the absence of which decreased the model's accuracy by 10.9%. CONCLUSION: Radiomics features extracted on contrast-enhanced T1 Weighted Imaging and T2 Weighted Imaging sequences were predictive of future haemorrhage of intracranial tumours with favourable accuracy. This model may assist in the arrangement of biopsy procedures and the selection of target lesions in patients with multiple lesions.