Table 1.
Summary of papers on AI for PDAC detection. The performance for the validation and test sets is reported with respective 95% Confidence Interval or standard deviation when it was provided.
| Authors (Year) | Data | Approach | Model | Metric | Validation Performance | Test Performance | Dev. Cohort | Test Cohort |
|---|---|---|---|---|---|---|---|---|
| Alves et al. (2022) [35] | CT | DL | 3D nnU-Net |
AUC | 0.991 (0.970–1.0) |
** 0.889 (0.833–0.946) |
242 | ** 361 |
| Wang et al. (2021) [38] | CT | DL | 2D U-Net | SEN, SPE | 0.998, 0.965 | .. | 800 | .. |
| Liu et al. (2020) [37] | CT | DL | 2D VGG | AUC | 1.000 (0.999–1.000) |
* 0.997 (0.992–1.000) * 0.999 (0.998–1.000) ** 0.920 (0.891–0.948) |
412 | * 139 * 189 ** 363 |
| Ma et al. (2020) [39] | CT | DL | 2D 4-layer CNN | AUC | 0.9652 | .. | 412 | .. |
| Tonozuka et al. (2020) [42] | EUS | DL | 2D 7-layer CNN | AUC | 0.924 | * 0.940 | 93 | * 47 |
| Qiu et al. (2021) [43] | CT | Radiomics | SVM | AUC | 0.88 (0.84–0.92) |
* 0.79 (0.71–0.87) |
312 | * 93 |
| Chen et al. (2021) [36] | CT | Radiomics | XGBoost | AUC | .. | * 0.98 (0.96–0.99) ** 0.76 (0.71–0.82) |
944 | * 383 ** 212 |
| Chu et al. (2020) [40] | CT | Radiomics | RF | SEN, SPE, ACC | 0.950, 0.923, 0.936 | .. | 380 | .. |
| Chu et al. (2019) [44] | CT | Radiomics | RF | AUC | .. | * 0.999 | 255 | * 125 |
| Li et al. (2018) [41] | 18FDG PET-CT | Radiomics | SVM-RF | SEN, SPE, ACC | 0.952 ± 0.009, 0.975 ± 0.004, 0.965 ± 0.007 | .. | 80 | .. |
| Ozkan et al. (2015) [45] | EUS | Radiomics | ANN | SEN, SPE, ACC | .. | * 0.833 ± 0.112, 0.933 ± 0.075, 0.875 ± 0.047 | 172 | * 72 images |
** external test set, * internal test set. Abbreviations are: DL—deep learning, XGBoost—extreme gradient boost, SVM—support vector machine, VGG—visual geometry group, RF—random forest, ANN—artificial neural network, CNN—convolutional neural network, AUC—area under the receiver operating characteristic curve, SEN—sensitivity, SPE—specificity, ACC—accuracy, Dev. Cohort—development cohort (training + validation).