Table 4 Performance for demographic features prediction stratified by whether each test image is correctly re-identified or not
From: Re-identification of patients from imaging features extracted by foundation models
| Â | CORIS-CFP | CORIS-OCT | MIDRC | |||
|---|---|---|---|---|---|---|
Re-identified? | Yes | No | Yes | No | Yes | No |
N. images | 1747 | 2724 | 30,311 | 32,893 | 7791 | 8226 |
N. patients | 156 | 194 | 166 | 199 | 1528 | 2787 |
Gender is Male | ||||||
 Accuracy | 76.3 | 69.6 | 64.4 | 63.7 | 87.6 | 87.8 |
 AUC-ROC | 82.1 | 76.8 | 70.3 | 68.4 | 94.4 | 94.9 |
Age | ||||||
 R2 | 0.60 | 0.56 | 0.71 | 0.69 | 0.67 | 0.69 |
 MAE | 7.36 | 8.16 | 7.68 | 8.2 | 8.12 | 7.57 |
Race is Caucasian | ||||||
 Accuracy | 88.0 | 88.1 | 86.9 | 83.4 | 78.3 | 78.3 |
 AUC-ROC | 82.2 | 79.3 | 94.1 | 91.2 | 86.0 | 86.5 |
Ethnicity is Hispanic | ||||||
 Accuracy | 90.3 | 85.6 | 81.7 | 80.2 | 71.5 | 70.1 |
 AUC-ROC | 93.3 | 76.0 | 84.1 | 81.6 | 66.4 | 69.8 |
