TABLE 2.
Key publications of AI applications in kidney disease
| Author, year | AI techniques | No of patients | Outcome predicted | Performance | Clinical application use |
|---|---|---|---|---|---|
| Akbilgic et al 16 2019 | Random Forest | 27 615 | Risk of death | AUROC: 0.70‐0.76 | NA |
| Goldstein et al 32 2014 | Random Forest | 826 | Sudden cardiac death | AUROC: 0.78‐0.79 | NA |
| Mezzatesta et al 33 2019 | Support Vector Machine | 1216 | Cardiovascular disease |
Accuracy: 92.15%‐92.25% AUROC: 0.50‐0.74 Precision: 72%‐89% Recall: 73%‐94% |
NA |
| Chauhan et al 39 2020 | Random Forest | 1369 | CKD progression |
AUROC: 0.77‐0.80 PPV: 62% in high‐risk group NPV: 92%‐96% in low‐risk group |
NA |
| Norouzi et al 42 2016 | Artificial Neural Networks | 465 | CKD progression |
MSE: 58.63‐64.00 MAE: 4.77‐5.93 NMSE: 4.77%‐4.88% |
NA |
| Barbieri et al 46 2016 | Artificial Neural Networks | 752 | Anemia management | MAE: 0.59 g/dL | Yes |
| Zhang et al 74 2017 | Random Forest | 83 | Immune fingerprints |
AUROC: 0.993 Sensitivity: 98.5% Specificity: 92.6% |
NA |
Abbreviations: AI, artificial intelligence; AUROC, area under the receiver operating curve; MAE, mean absolute error; MSE, mean square error; NMSE, normalized MSE; NPV, negative predicted value; PPV, positive predicted value.