Table 3.
ChatGPT’s diagnostic accuracy in nontumor etiologies
| Correct answer (accuracy rate [%]) | ||||
|---|---|---|---|---|
| GPT-4-based ChatGPT | GPT-4V-based ChatGPT | |||
| Final diagnosis | Differential diagnosis | Final diagnosis | Differential diagnosis | |
| Muscle/soft tissue/nerve disorder (n = 12) | 7/12 (58%) | 11/12 (92%) | 2/12 (17%) | 3/12 (25%) |
| Arthritis/arthropathy (n = 10) | 4/10 (40%) | 4/10 (40%) | 1/10 (10%) | 1/10 (10%) |
| Infection (n = 8) | 3/8 (38%) | 5/8 (63%) | 0/8 (0%) | 1/8 (13%) |
| Congenital/developmental abnormality and dysplasia (n = 6) | 4/6 (67%) | 4/6 (67%) | 0/6 (0%) | 0/6 (0%) |
| Trauma (n = 6) | 5/6 (83%) | 5/6 (83%) | 1/6 (17%) | 2/6 (33%) |
| Metabolic disease (n = 5) | 2/5 (40%) | 3/5 (60%) | 0/5 (0%) | 0/5 (0%) |
| Anatomical variant (n = 4) | 3/4 (75%) | 3/4 (75%) | 0/4 (0%) | 0/4 (0%) |
| Others (n = 10) | 4/10 (40%) | 5/10 (50%) | 1/10 (10%) | 3/10 (30%) |
ChatGPT Chat Generative Pre-trained Transformer, GPT-4 Generative Pre-trained Transformer-4, GPT-4V Generative Pre-trained Transformer-4 with vision