Table 1.
Performance of clinical models and deep radiomic models of BCR prediction by univariate cox proportional hazards regression.
| PC (PUTH) | VC (BJFH) | VC2 (PUPH) | p | |
|---|---|---|---|---|
| n = 368 | n = 34 | n = 83 | ||
| No. of BCR event (%) | 99 (27.1) | 16 (47.1) | 31 (37.3) | 0.016 |
| Age (Median) | 70 | 68.5 | 68.0 | 0.275 |
| PSA (Median) | 10.7 | 9.89 | 13.4 | 0.883 |
| GG-NB | 0.125 | |||
| 1 | 138 | 13 | 29 | |
| 2 | 114 | 10 | 23 | |
| 3 | 63 | 6 | 14 | |
| 4 | 89 | 3 | 11 | |
| 5 | 81 | 2 | 6 | |
| cT | 0.436 | |||
| 2 | 221 | 15 | 44 | |
| 3 | 264 | 19 | 39 | |
| 4 | 9 | |||
| pT | 0.687 | |||
| 1 | 2 | 1 | ||
| 2 | 303 | 22 | 57 | |
| 3 | 165 | 12 | 22 | |
| 4 | 15 | 3 | ||
| GG-RP | 0.173 | |||
| 1 | 82 | 4 | 13 | |
| 2 | 127 | 10 | 28 | |
| 3 | 93 | 8 | 18 | |
| 4 | 65 | 9 | 8 | |
| 5 | 118 | 3 | 16 | |
| SM | 0.172 | |||
| Positive | 170 | 12 | 38 | |
| Negative | 315 | 21 | 45 | |
| EPE | 0.639 | |||
| Positive | 171 | 11 | 25 | |
| Negative | 314 | 23 | 58 | |
| SVI | 0.940 | |||
| Positive | 67 | 4 | 11 | |
| Negative | 418 | 30 | 72 | |
| CAPRA | 0.248 | |||
| Low | 53 | 9 | 10 | |
| Intermediate | 168 | 12 | 34 | |
| High | 148 | 13 | 39 |
Note: BCR: biochemical recurrence; PC, primary cohort; VC, validation cohort; PUTH, Peking University Third Hospital; BJFH, Beijing Friendship Hospital; PUPH, Peking University People’s Hospital; PSA, prostate specific antigen; GG-NB, Gleason grade group of needle biopsy; cT. clinical T stage; pT, pathological T stage; GG-RP, Gleason grade group of radical prostatectomy; SM, surgical margin; EPE, extracapsular extension; SVI, seminal vesicle invasion; CAPRA, Cancer of Prostate Risk Assessment Score.