Table 4.
AUC result overview for all our experiments.
| Pathology | Without non-image features | With non-image features | ||||||
|---|---|---|---|---|---|---|---|---|
| OTS | FT | 1channel | large | OTS | FT | 1channel | large | |
| Cardiomegaly | 72.7 ± 1.8 | 88.5 ± 0.7 | 88.9 ± 0.5 | 89.7 ± 0.3 | 75.9 ± 1.4 | 88.4 ± 0.8 | 90.2 ± 0.4 | 89.8 ± 0.8 |
| Emphysema | 77.8 ± 2.1 | 89.2 ± 1.0 | 87.0 ± 0.8 | 88.3 ± 1.3 | 79.8 ± 1.9 | 89.4 ± 1.2 | 87.4 ± 1.3 | 89.1 ± 1.2 |
| Edema | 84.4 ± 0.6 | 89.1 ± 0.4 | 89.1 ± 0.6 | 88.8 ± 0.5 | 85.7 ± 0.5 | 89.1 ± 0.7 | 89.0 ± 0.6 | 88.9 ± 0.3 |
| Hernia | 78.8 ± 1.4 | 85.5 ± 3.8 | 88.1 ± 4.2 | 87.5 ± 4.5 | 81.9 ± 2.5 | 88.2 ± 3.2 | 89.3 ± 4.4 | 89.6 ± 4.4 |
| Pneumothorax | 77.3 ± 1.3 | 87.0 ± 0.8 | 85.7 ± 0.9 | 85.9 ± 0.9 | 79.1 ± 1.2 | 86.5 ± 0.6 | 85.4 ± 0.7 | 85.9 ± 1.1 |
| Effusion | 79.4 ± 0.4 | 87.1 ± 0.2 | 87.6 ± 0.2 | 87.6 ± 0.2 | 80.6 ± 0.4 | 87.2 ± 0.3 | 87.6 ± 0.2 | 87.3 ± 0.3 |
| Mass | 66.8 ± 0.6 | 82.2 ± 1.0 | 83.3 ± 0.6 | 83.9 ± 0.9 | 68.6 ± 0.6 | 82.2 ± 1.0 | 83.3 ± 0.7 | 83.2 ± 0.3 |
| Fibrosis | 72.0 ± 0.9 | 80.0 ± 0.9 | 79.9 ± 0.8 | 79.2 ± 1.6 | 73.9 ± 0.8 | 80.0 ± 0.9 | 79.6 ± 0.5 | 78.9 ± 0.5 |
| Atelectasis | 71.8 ± 0.6 | 80.3 ± 0.7 | 79.9 ± 0.4 | 79.2 ± 0.7 | 73.2 ± 0.7 | 80.1 ± 0.6 | 79.3 ± 0.6 | 79.1 ± 0.4 |
| Consolidation | 74.3 ± 0.3 | 79.5 ± 0.5 | 80.6 ± 0.4 | 80.0 ± 0.3 | 75.3 ± 0.3 | 79.6 ± 0.5 | 80.4 ± 0.5 | 80.0 ± 0.7 |
| Pleural Thicken. | 68.8 ± 1.0 | 79.0 ± 0.7 | 78.4 ± 0.9 | 78.0 ± 1.1 | 70.8 ± 1.1 | 78.6 ± 1.1 | 78.2 ± 1.3 | 77.1 ± 1.3 |
| Nodule | 65.0 ± 0.8 | 72.6 ± 0.9 | 73.3 ± 0.8 | 75.1 ± 1.3 | 66.5 ± 0.7 | 74.7 ± 0.6 | 74.0 ± 0.7 | 75.8 ± 1.4 |
| Pneumonia | 66.4 ± 2.7 | 74.4 ± 1.6 | 74.3 ± 1.5 | 75.3 ± 2.2 | 68.3 ± 2.3 | 73.3 ± 1.3 | 74.8 ± 1.5 | 76.7 ± 1.5 |
| Infiltration | 65.9 ± 0.2 | 69.9 ± 0.6 | 70.2 ± 0.3 | 70.2 ± 0.5 | 67.0 ± 0.4 | 70.2 ± 0.2 | 70.1 ± 0.5 | 70.0 ± 0.7 |
| Average | 73.0 ± 1.1 | 81.7 ± 1.0 | 81.9 ± 0.9 | 82.1 ± 1.2 | 74.8 ± 1.1 | 82.0 ± 0.9 | 82.0 ± 1.0 | 82.2 ± 1.1 |
| No Findings | 71.6 ± 0.3 | 76.9 ± 0.5 | 77.3 ± 0.3 | 77.1 ± 0.4 | 72.5 ± 0.3 | 76.8 ± 0.4 | 77.1 ± 0.4 | 77.1 ± 0.3 |
In this table, we present averaged results over all five splits and the calculated standard deviation (std) for each pathology. We divide our experiments into three categories. First, without and with non-image features. Second, transfer-learning with off-the-shelf (OTS) and fine-tuned (FT) models. Third, from scratch where “1channel” refers to same input size as in transfer-learning but changed number of channels. “large” means we changed the input dimensions to 448 × 448 × 1. For better comparison, we present the average AUC and the standard deviation over all pathologies in the last row. Bold text emphasizes the overall highest AUC value. Values are scaled by 100 for convenience.