Table 4 The effect of disease probability soft guidance versus disease hard guidance on the performance of our proposed DPE-all.
From: Disease probability-enhanced follow-up chest X-ray radiology report summary generation
Hard threshold | Natural language processing metric | Clinical metric | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
B1 | B2 | B3 | B4 | M | R | C | Acc5 | Acc14 | F1-5 | F1-14 | Rad-F1 | |
0.5 | 63.9 | 57.0 | 51.3 | 46.1 | 38.0 | 59.9 | 1.525 | 67.0 | 80.7 | 47.6 | 43.4 | 25.7 |
0.6 | 63.7 | 56.8 | 51.1 | 45.9 | 38.4 | 60.3 | 1.511 | 66.5 | 80.7 | 47.9 | 43.6 | 26.0 |
0.7 | 63.6 | 56.8 | 51.0 | 45.8 | 38.6 | 60.6 | 1.506 | 66.3 | 80.8 | 48.1 | 43.7 | 26.3 |
0.8 | 63.3 | 56.5 | 50.6 | 45.5 | 39.0 | 60.8 | 1.479 | 65.8 | 80.6 | 48.3 | 43.7 | 26.8 |
0.9 | 62.9 | 56.1 | 50.3 | 45.2 | 39.3 | 60.9 | 1.471 | 65.6 | 80.4 | 47.9 | 43.8 | 27.3 |
Soft | 65.6 | 58.8 | 53.1 | 48.0 | 39.4 | 62.9 | 1.680 | 69.7 | 81.9 | 49.9 | 45.9 | 28.1 |
