Table 4.
Analysis, validation, and results for csPCa prediction in the selected studies based on deep radiomics as the feature extraction method.
| Reference, Year | Analysis | Validation | Sequence for the Best Model | Best Radiomic Model [CI, 95%] a | PI-RADS Cut-Off | PI-RADS Model [CI, 95%] a | ||||
|---|---|---|---|---|---|---|---|---|---|---|
| AUC | Sensitivity | Specificity | AUC | Sensitivity | Specificity | |||||
| Hamm et al. 2023 [27] | All | Hold-out set//External (PROSTATEx) | T2 + ADC + DWI (high-b value) | 0.89 [0.85, 0.93]//0.87 [0.81, 0.93] | 0.77 [0.69, 0.85]//0.90 [0.83, 0.97] | 0.89 [0.84, 0.95]//0.85 [0.80, 0.90] | NA | NA | NA | NA |
| Index | 0.78//NR | 0.98 [0.95, 1.00] b//NR | NR | |||||||
| Hong et al. 2023 [28] | Index | Hold-out set//External (1 set) | ADC | NR//0.63 | 0.72//0.84 | 0.74//0.48 | NA | NA | NA | NA |
| Zhu et al. 2022 [30] | All | Hold-out set | T2 + ADC | NR | 0.96 [0.89, 0.99] | NR | ≥3 | NR | 0.94 [0.87, 0.98] | NR |
| Sextant | NR | 0.96 [0.90, 0.99] | 0.92 [0.89, 0.93] | NR | 0.93 [0.87, 0.97] | 0.92 [0.90, 0.94] | ||||
| Index | NR | 0.99 [0.92, 0.99] | 0.65 [0.53, 0.76] | NR | 0.99 [0.92, 0.99] | 0.66 [0.54, 0.77] | ||||
| Jiang et al. 2022 [31] | All | Hold-out set//External (PROSTATEx) | T2 + ADC + DWI (b1500) | 0.85 [0.81, 0.88]//0.86 [0.81, 0.91] | 0.93//0.87 | 0.5//0.66 | ≥3 | 0.92 [0.89, 0.95]//0.86 [0.80, 0.90] | 0.94//0.77 | 0.79//0.87 |
| Liu et al. 2021 [32] | All | Hold-out set | T2 + ADC | 0.85 [0.79, 0.91] | 0.90 [0.83, 0.96] | 0.70 [0.59, 0.82] | ≥4 | 0.73 [0.65, 0.80] | 0.83 [0.75, 0.92] | 0.47 [0.35, 0.59] |
| Index | 0.73 [0.59, 0.88] | 0.90 [0.83, 0.96] | 0.47 [0.21, 0.72] | 0.65 [0.52, 0.78] | 0.83 [0.75, 0.91] | 0.27 [0.04, 0.72] | ||||
| Zhong et al. 2019 [37] | All | Hold-out set | T2 + ADC | 0.73 [0.58, 0.88] | 0.64 | 0.8 | ≥4 | 0.71 [0.58, 0.87] | 0.86 | 0.48 |
All = all lesions, AUC = area under the curve, CI = confidence interval, csPCa = clinically significant prostate cancer, CV = cross-validation, Index = index lesion, NA = not applicable, NR = not reported, PI-RADS = Prostate Imaging Reporting and Data System. a Data are expressed in the corresponding metric and the CI, 95% for each validation method separated by//. If the CI is not included it means that it was not reported in the study. b At 2 false-positive rate.