Research Article
Pulmonary Nodule Detection Model Based on SVM and CT Image Feature-Level Fusion with Rough Sets
Table 6
Stability statistics of rough set reduction subsets.
| | Training set/testing set | Accuracy (%) | Sensitivity (%) | Specificity (%) | Running time (s) |
| Before fusion | 50/20 | 97.35 | 94.71 | 100 | 0.4873 | 40/30 | 96.53 | 93.08 | 98.32 | 0.3846 | 35/35 | 95.83 | 92.39 | 97.79 | 0.4254 | 30/40 | 96.16 | 95.58 | 96.74 | 0.3560 | 20/50 | 94.88 | 94.63 | 95.86 | 0.4236 |
| | Mean | 96.15 | 94.08 | 97.742 | 0.4154 |
| After fusion (Rs1) | 50/20 | 99.71 | 99.41 | 100 | 0.2684 | 40/30 | 98.96 | 99.58 | 98.46 | 0.2568 | 35/35 | 98.65 | 99.23 | 98.08 | 0.2382 | 30/40 | 98.37 | 98.60 | 98.14 | 0.2646 | 20/50 | 98.25 | 97.67 | 98.84 | 0.2636 |
| | Mean | 98.79 | 98.84 | 98.70 | 0.2583 |
|
|