Table 1.
Participant Demographic Information, and Basic Characteristics of the Datasets.
| Parameter | Dataset |
||
|---|---|---|---|
| Training | Validation | Testing | |
| No. of instancesa | 1530 | 210 | 435 |
| No. of scans | 102 | 14 | 29 |
| No. of patients | 72 | 10 | 16 |
| No. of females (scan based, %) | 31 (30.4%) | 2 (14.3%) | 16 (55.2%) |
| Age (scan based, year) | 52.7 | 46.1 | 50.4 |
| No. of grade IV (%) | 63 (61.8%) | 9 (64.3%) | 22 (75.9%) |
| Time interval after radiation (day) b | 257 (95–448) | 183 (89–375) | 287 (190–542) |
| No. of first progression (%) | 66 (64.7%) | 5 (35.7%) | 11 (37.9%) |
| No. of progression at the slice level (%) | 547 (35.8%) | 59 (28.1%) | 136 (31.3%) |
| No. of progression at the scan level (%) | 61 (59.8%) | 7 (50.0%) | 18 (62.1%) |
| Enrollment time of first scan (YYYY/MM) c | 2010/04 | 2014/10 | 2015/07 |
| Enrollment time of last scan (YYYY/MM) c | 2014/10 | 2015/07 | 2018/02 |
Note. a One instance indicates a set of APTw, T1w, T2w, FLAIR, and GdT1w MR images acquired at the same slice level. b Time interval after radiation indicates the timing of research MRI acquisition with regard to the end of radiation, described as median with 25th percentile and 75th percentile.
c
To simulate the practical use of this algorithm in a simulated prospective study, the whole dataset was split according to the chronological order of the MRI scan date.