Table 1.
Overview of Relevant Prior Work Towards OS Prediction Using Baseline CT Images in Patients on IO, in a mNSCLC Setting.
| Reference | Total no. of Patients | Accuracy (%) in Test Cohort (mutually exclusive training and test sets sampled from same cohort) |
AUC in Test Cohort (mutually exclusive training and test sets sampled from same cohort) |
Concordance Index (robustness to training/test splits) | Generalization Accuracy (%) (independent cohort) |
|---|---|---|---|---|---|
| Jazieh et al. (5) | 133 | N/A (Ntest = 69) | N/A | 0.69 (N/A) | N/A |
| Gong et al. (6) | 224 | 50.79 (Ntest = 63) | 0.52 (0.39–0.65) | 0.51 (0.43–0.6) | 52.94 (Ntest = 68) |
| He et al. (7) | 236 | 83.3 (Ntest = 48) | N/A | 0.75 (0.59–0.90) | N/A |
| Liu et al. (8) | 46 | N/A (Ntest = 1) | 0.64 (0.48–0.79) | 0.86 (0.74–0.97) | N/A |
| Trebeschi et al. (9) | 123 | N/A (Ntest = 42) | 0.76 (N/A) | N/A | N/A |
| Work Presented* | 917 | 66.8 (Ntest = 199) | 0.71 (0.65–0.78) | 0.66 (0.65–0.67) | 73.5 (Ntest = 144) |
Values not reported are indicated with N/A. Ntest indicates the number of “held-out” samples used to test a model (these samples are not seen by the model during the training phase).
*Results reported are for Model-C+R. Evaluation metrics can be found in Supplementary Table 5 .
AUC, area under the curve; CT, computed tomography; IO, immunotherapy; OS, overall survival; mNSCLC, metastatic non-small cell lung cancer; NA, not applicable.