Table 5 Convolutional neural network baseline table
| Â | Training set | Test set | X2 | P -value | ||
|---|---|---|---|---|---|---|
N | % | N | % | |||
Sex | ||||||
 Male | 92 | 24.1% | 32 | 33.7% | 3.163 | 0.075 |
 Female | 290 | 75.9% | 63 | 66.3% | ||
Age (years) | ||||||
  < 45 | 184 | 48.2% | 50 | 52.6% | 0.441 | 0.507 |
  ≥ 45 | 198 | 51.8% | 45 | 47.4% | ||
Combined benign thyroid disease | ||||||
 No | 209 | 54.7% | 48 | 50.5% | 2.126 | 0.547 |
 Nodular goiter | 101 | 26.4% | 32 | 33.7% | ||
 Lymphocytic thyroiditis | 55 | 14.4% | 11 | 11.6% | ||
 Both | 17 | 4.5% | 4 | 4.2% | ||
BRAF V600E gene mutation | ||||||
 No | 79 | 20.7% | 18 | 18.9% | 0.054 | 0.816 |
 Yes | 303 | 79.3% | 77 | 81.1% | ||
Lymph node metastasis | ||||||
 No | 201 | 52.6% | 50 | 52.6% |  < 0.001 | 1.000 |
 Yes | 181 | 47.4% | 45 | 47.4% | ||
Maximum diameter of thyroid nodules | ||||||
  < 10 mm | 167 | 43.7% | 40 | 42.1% | 0.542 | 0.763 |
 10–20 mm | 162 | 42.4 | 39 | 41.1% | ||
  > 20 mm | 53 | 13.9% | 16 | 16.8% | ||
Nodule location | ||||||
 Upper 1/3 | 79 | 20.7% | 19 | 20.0% | 1.573 | 0.814 |
 Middle 1/3 | 118 | 30.9% | 27 | 28.4% | ||
 Lower 1/3 | 63 | 16.5% | 13 | 13.7% | ||
 Isthmus | 49 | 12.8% | 13 | 13.7% | ||
 Whole | 73 | 19.1% | 23 | 24.2% | ||
Aspect Ratio | ||||||
  < 1 | 230 | 60.2% | 44 | 46.3% | 5.453 | 0.020 |
  ≥ 1 | 152 | 39.8% | 51 | 53.7% | ||
Microcalcification | ||||||
 No | 101 | 26.4% | 22 | 23.2% | 0.274 | 0.601 |
 Yes | 281 | 73.6% | 73 | 76.8% | ||
Nodule boundary | ||||||
 Clear | 158 | 41.4% | 34 | 35.8% | 0.764 | 0.382 |
 Blurred | 224 | 58.6% | 61 | 64.2% | ||
Capsular invasion | ||||||
 No | 218 | 57.1% | 50 | 52.6% | 5.020 | 0.081 |
 Close | 55 | 14.4% | 8 | 8.4% | ||
 Invasion | 109 | 28.5% | 37 | 38.9% | ||
Number of lesions | ||||||
 Unifocal | 276 | 72.3% | 63 | 66.3% | 1.031 | 0.310 |
 Multifocal | 106 | 27.7% | 32 | 33.7% | ||