Table 10.
The COVID-19 diagnosis and prognosis accuracy (AUC) of different sizes of our Med-MLLM, which are trained on full training data.
| Methods | Year | Retrospective studies | Prospective studies | ||
|---|---|---|---|---|---|
| Diagnosis | Prognosis | Diagnosis | Prognosis | ||
| ClinicalBERT94 | 2019 | 93.2 | 93.4 | 74.8 | 87.5 |
| BioBERT93 | 2020 | 85.9 | 92.2 | 71.2 | 84.6 |
| PubMedBERT92 | 2022 | 95.7 | 93.9 | 76.7 | 88.4 |
| Med-MLLM-Base | Ours | 96.3 | 95.3 | 84.1 | 92.8 |
| Med-MLLM-Large | Ours | 98.0(+1.7) | 97.2(+1.9) | 87.7(+3.6) | 94.9(+2.1) |
All values are reported in percentage (%). The best results are in bold.