Table 2.
Class distribution in training, validation, and test sets for each classification task.
| Target Population | Classification Task | Training 1 | Validation 1 | Test 2 |
|---|---|---|---|---|
| Only patients with grade 2 and 3 gliomas (n = 100) |
IDH-mutant (n = 78) vs. IDH-wildtype (n = 22) |
n = 72 vs. n = 16 |
n = 2 vs. n = 2 |
n = 4 vs. n = 4 |
| 1p/19q-codeleted (n = 50) vs. 1p/19q-uncodeleted (n = 50) |
n = 41 vs. n = 41 |
n = 3 vs. n = 3 |
n = 6 vs. n = 6 |
|
| Patients with grade 2, 3, and 4 gliomas (n = 146) |
IDH-mutant (n = 84) vs. IDH-wildtype (n = 62) |
n = 71 vs. n = 49 |
n = 4 vs. n = 4 |
n = 9 vs. n = 9 |
| IDH-mutant and 1p/19q-codeleted (n = 50) vs. IDH-mutant and 1p/19q-uncodeleted (n = 34) vs. IDH-wildtype (n = 62) |
n = 40 vs. n = 24 vs. n = 52 |
n = 4 vs. n = 4 vs. n = 4 |
n = 6 vs. n = 6 vs. n = 6 |
1 The reported numbers of patients in training and validation tests refer to one split within an inner loop.2 The reported numbers of patients in test sets refer to one split within the outer loop.