Table 3.
Comparison of parameters between the first and second (i.e., AI-supported) diagnostic examinations.
| DNN model | SS-Avg (%) | Accuracy (%) | Sensitivity (%) | Specificity (%) | ||||
|---|---|---|---|---|---|---|---|---|
| 90.8 | 90.3 | 89.8 | 91.7 | |||||
| S-Avg | P-Avg | S-Avg | P-Avg | S-Avg | P-Avg | S-Avg | P-Avg | |
| 1st. Exam. (no AI support) | 82.4 | 69.6 | 88.3 | 80.1 | 71.0 | 47.6 | 93.8 | 91.5 |
| 2nd. Exam. (with AI-supported) | 87.3 | 90.8 | 89.6 | 92.3 | 83.1 | 87.8 | 91.7 | 93.7 |
S-Avg: Average results for radiological specialists.
P-Avg: Average results for radiological practitioners.
DNN model results (sarcoma likelihood and sequences for each patient) were provided to the radiologists for the second examination.
Remarkably, all parameters were superior among radiological practitioners than among radiological specialists, although the difference was not significant (p > 0.05).