Table 3.
Performance of automated classical machine learning (ML) and deep leaning (DL) methods for image quality assessment.
| Reference | Image type | Multi-center data? | Learning algorithm (number of features) | Sensitivity | Specificity | Accuracy | PPV | NPV | AUC |
|---|---|---|---|---|---|---|---|---|---|
| (4) | T1w, Brain | Yes | ML (2) | 0.87 | 0.85 | - | - | - | 0.93 |
| (7) | T1w, Brain | No | ML (6) | 0.70 | 0.88 | 0.80 | - | - | - |
| (27) | T1w, Brain | Yes | ML (64) | 0.28 | 0.95 | 0.76 | - | - | 0.70 |
| (5) | T1w, Brain | Yes | ML (190) | 0.91 | 0.84 | 0.84 | 0.09 | 0.99 | - |
| (14) | T1w, head | No | DL | - | - | 0.92±0.08 | - | - | - |
| (14) | T1w, upper abdomen | No | DL | - | - | 0.72±0.05 | - | - | - |
| (15) | T2w, Liver | No | DL | 0.47 – 0.67 | 0.80 – 0.81 | 0.73 – 0.79 | 0.36 | 0.86 – 0.94 | - |
| (16) | T1w, Brain | No | DL | - | - | 0.88 | - | - | - |
| This work | T1w, Brain | Yes | DL | 0.77 | 0.85 | 0.84 | 0.42 | 0.96 | 0.90 |
Empty cells indicate values were not reported.