Table 3.
Summary of the subtype-specific classification performance of the radiomics model.
| Molecular subgroups | Cohorts | AUC | BAL_ACC | SEN | SPE |
|---|---|---|---|---|---|
| IDHwt | Training | 0.8121 (0.7559-0.8682) | 0.7782 (0.6989-0.8401) | 0.7895 (0.6444-0.8947) | 0.7668 (0.6528-0.8912) |
| Testing | 0.6557 (0.5084-0.8029) | 0.6806 (0.4394-0.8181) | 0.7778 (0.2778-100.00) | 0.5833 (0.2292-0.9792) | |
| IDHmt-noncodel | Training | 0.7384 (0.6739-0.8030) | 0.7052 (0.5462-0.7510) | 0.8256 (0.5462-0.9186) | 0.5847 (0.4699-0.8306) |
| Testing | 0.6830 (0.5478-0.8183) | 0.7232 (0.5758-0.8333) | 0.7083 (0.5000-0.9177) | 0.7381 (0.3810-0.8810) | |
| IDHmt-codel | Training | 0.7905 (0.7351-0.8459) | 0.7595 (0.6989-0.8104) | 0.7103 (0.5888-0.8598) | 0.8086 (0.6420-0.8827) |
| Testing | 0.7579 (0.6359-0.8799) | 0.7500 (0.6212-0.8636) | 0.6667 (0.4583-0.9594) | 0.8333 (0.4524-0.9762) |
BAL_ACC, SEN, and SPE are short for balanced accuracy, sensitivity, and specificity, respectively. The 95% confidence interval for each index is shown.