Table 3.
The top 5 most frequently selected radiomics features of the three classification tasks based on the optimum RFF models.
| Classification tasks | Top 5 radiomics features | P value | M | (<Mean |>Mean) |
|---|---|---|---|---|
| HR+ vs. HR- | shape_SphericalDisproportion (1st) | 0.0002 | 0.569 | HR+ (70.59% | 29.41%) |
| HR- (57.58% | 42.42%) | ||||
| shape_MinorAxisLength (2nd) | 0.0013 | 0.203 | HR+ (64.29% | 35.71%) | |
| HR- (49.49% | 50.51%) | ||||
| shape_Sphericity (3rd) | 0.0002 | -0.691 | HR+ (72.27% | 27.73%) | |
| HR- (56.57% | 43.43%) | ||||
| glcm_Correlation (4th) | 0.0002 | 0.729 | HR+ (57.56% | 42.44%) | |
| HR- (39.39% | 60.61%) | ||||
| glcm_Imc2 (5th) | 0.0010 | 1.069 | HR+ (57.14% | 42.86%) | |
| HR- (41.41% | 58.59%) | ||||
| TNBC vs. HEBC | firstorder_Entropy (1st) | <10-4 | -1.229 | TNBC (67.44% | 32.56%) |
| HEBC (30.36% | 69.64%) | ||||
| firstorder_MeanAbsoluteDeviation (2nd) | 0.0004 | -0.778 | TNBC (55.81% | 44.19%) | |
| HEBC (30.36% | 69.64%) | ||||
| firstorder_Uniformity (3rd) | <10-4 | 0.838 | TNBC (69.77% | 30.23%) | |
| HEBC (35.71% | 64.29%) | ||||
| glcm_ClusterTendency (4th) | 0.0017 | -0.390 | TNBC (58.14% | 41.86%) | |
| HEBC (28.57% | 71.43%) | ||||
| firstorder_RobustMeanAbsoluteDeviation (5th) | 0.0007 | -0.689 | TNBC (53.49% | 46.51%) | |
| HEBC (32.14% | 67.86%) | ||||
| TNBC vs. non-TNBC | firstorder_90Percentile (1st) | <10-7 | -0.169 | TNBC (62.79% | 37.21%) |
| non-TNBC (25.17% | 74.83%) | ||||
| firstorder_MeanAbsoluteDeviation (2nd) | <10-6 | -0.232 | TNBC (62.79% | 37.21%) | |
| non-TNBC (24.83% | 75.17%) | ||||
| firstorder_RobustMeanAbsoluteDeviation (3rd) | <10-6 | -0.198 | TNBC (62.79% | 37.21%) | |
| non-TNBC (22.45% | 77.55%) | ||||
| firstorder_Entropy (4th) | <10-6 | -0.365 | TNBC (67.44% | 32.56%) | |
| non-TNBC (30.95% | 69.05%) | ||||
| firstorder_RootMeanSquared (5th) | <10-5 | -0.103 | TNBC (60.47% |39.53%) | |
| non-TNBC (27.55% | 72.45%) |
The ‘Mean’ shows the mean of the mean radiomics feature values of the two groups in each classification. The letter of ‘(<Mean | >Mean)’ represents the percentage of patients in the two groups with feature value smaller than or larger than the ‘Mean’ value. Values in bold indicate these features with better discriminative performance.