Table 8 Comparison of unnecessary FNAB rates of the dual-modal US radiomics nomogram and other models
Diagnostic models | Datasets | No. of recommended FNABs | No. of malignant nodules | No. of benign nodules | Unnecessary FNAB rates (%) | P value* |
|---|---|---|---|---|---|---|
ACR TI-RADS | Training dataset (n = 219) | 116 | 75 | 41 | 35.3 (41/116) | < 0.001 |
Validation dataset (n = 93) | 53 | 31 | 22 | 41.5 (22/53) | 0.005 | |
All dataset (312) | 169 | 103 | 63 | 37.3 (63/169) | < 0.0001 | |
Dual-modal US radiomics nomogram | Training dataset (n = 219) | 145 | 124 | 21 | 14.5 (21/145) | NA |
Validation dataset (n = 93) | 62 | 51 | 11 | 17.8 (11/62) | NA | |
All dataset (312) | 207 | 175 | 32 | 15.5 (32/207) | NA | |
Clinical combined with US model | Training dataset (n = 219) | 160 | 137 | 23 | 14.4 (23/160) | 0.980 |
Validation dataset (n = 93) | 65 | 54 | 11 | 16.9(11/65) | 0.894 | |
All dataset (312) | 225 | 191 | 34 | 15.1 (34/225) | 0.908 | |
Dual-modal US Rad-score | Training dataset (n = 219) | 158 | 138 | 20 | 12.7 (20/158) | 0.648 |
Validation dataset (n = 93) | 66 | 57 | 9 | 13.6 (9/66) | 0.515 | |
All dataset (312) | 224 | 195 | 29 | 12.9 (29/224) | 0.440 |