Table 1 Baseline characteristics
From: Multimodal GPT model for assisting thyroid nodule diagnosis and management
Categories | Training set (n = 56,285) | Test set 1 (n = 2964) | Test set 2 (n = 1263) |
|---|---|---|---|
Gender | |||
Female | 42,091 (74.8%) | 2173 (73.3%) | 927 (73.4%) |
Male | 14,194(25.2%) | 791 (26.7%) | 336 (26.6%) |
Age, mean (SD) | 51.6 ± 11.3 | 49.3 ± 12.0 | 47.9 ± 12.1 |
Number of nodules | 67,981 | 3376 | 1375 |
ACR TR1 | 816 (1.2%) | 54 (1.6%) | 39 (2.8%) |
ACR TR2 | 1829 (2.7%) | 67 (2.0%) | 62 (4.5%) |
ACR TR3 | 10,945 (16.1%) | 523 (15.5%) | 229 (16.7%) |
ACR TR4 | 28,076 (41.3%) | 1679 (49.7%) | 543 (39.5%) |
ACR TR5 | 26,315 (38.7%) | 1053 (31.2%) | 502(36.5%) |
Number of images | 487,246 | 23,010 | 8617 |
Size, mm (SD) | 20.5 ± 10.1 | 21.6 ± 9.4 | 22.1 ± 9.7 |
Pathological results | |||
Benign | 31,857 | 1775 | – |
Malignant | 36,124 | 1601 | – |
PTC | 32,186 (89.1%) | 1418 (88.6%) | – |
FTC | 2374(6.6%) | 94 (5.9%) | – |
MTC | 943 (2.6%) | 56 (3.5%) | – |
UTC | 621 (1.7%) | 33 (2.0%) | – |
Ultrasound Reports | 48,470 | - | 1263 |
