Table 2.
Summary of the top DRF found after the feature selection phase organized by clinical endpoint and reconstruction and pre-processing method. Asterisks were placed on the iCBCT and sCBCT reconstruction and pre-processing methods with the best performing AUC in their respective categories.
| Clinical endpoint | Reconstruction and pre-processing method | N | Selected features |
|---|---|---|---|
| Acute | iCBCT SHARP VeryLow Llo 1 | 6 | GLCM Correlation, GLCM Homogeneity, GLSZM GLN-VN, Global Skewness, NGTDM Coarseness-VN, Prostate Volume |
| Acute | iCBCT SHARP VeryLow Uni 0 | 3 | GLCM Correlation, NGTDM Coarseness-VN, Prostate Volume |
| Acute | iCBCT STD Med Llo 1 | 4 | GLCM Contrast, GLCM Dissimilarity, GLSZM LZHGE, Prostate Volume |
| Acute | iCBCT STD Med Uni 0* | 4 | GLSZM LZHGE, NGTDM Coarseness-VN, NGTDM Strength, Prostate Volume |
| Acute | sCBCT SHARP Llo 1 | 6 | GLCM Energy, GLRLM LRHGE, GLSZM GLN-VN, Global Kurtosis, Global Skewness, NGTDM Busyness |
| Acute | sCBCT SHARP Uni 0 | 5 | GLRLM GLN-VN, GLSZM GLN, GLSZM GLN-VN, NGTDM Coarseness, Prostate Volume |
| Acute | sCBCT STD Llo 1 | 4 | GLSZM GLN, GLSZM GLN-VN, Global Kurtosis, Prostate Volume |
| Acute | sCBCT STD Uni 0* | 6 | GLRLM GLN, GLRLM GLN-VN, GLSZM ZSN, GLSZM ZSV, Global Kurtosis, Prostate Volume |
| Sub-Acute | iCBCT SHARP VeryLow Llo 1 | 3 | GLCM Contrast, GLCM Correlation, NGTDM Contrast |
| Sub-Acute | iCBCT SHARP VeryLow Uni 0 | 3 | GLCM Correlation, GLRLM RLN, GLSZM ZSN |
| Sub-Acute | iCBCT STD Med Llo 1* | 6 | GLCM Contrast, GLCM Correlation, GLCM Dissimilarity, GLSZM LZLGE, NGTDM Coarseness, NGTDM Strength |
| Sub-Acute | iCBCT STD Med Uni 0 | 5 | GLSZM GLN, GLSZM LZHGE, GLSZM LZLGE, GLSZM ZSV, NGTDM Coarseness |
| Sub-Acute | sCBCT SHARP Llo 1 | 4 | GLCM SumAverage, GLRLM LRHGE, GLSZM GLN, GLSZM LZHGE |
| Sub-Acute | sCBCT SHARP Uni 0 | 5 | GLCM Correlation, GLRLM GLV, GLSZM GLV, GLSZM ZSV, Global Skewness |
| Sub-Acute | sCBCT STD Llo 1 | 7 | GLCM Variance, GLRLM RLN, GLSZM GLN, GLSZM LZLGE, Global Kurtosis, NGTDM Contrast, NGTDM Strength |
| Sub-Acute | sCBCT STD Uni 0* | 4 | GLRLM RLN, GLSZM ZSN, Global Skewness, NGTDM Coarseness-VN |
| ∆IPSS | iCBCT SHARP VeryLow Llo 1 | 6 | GLRLM GLN, GLRLM LRLGE, GLRLM RLN, GLSZM ZSN, IPSS Baseline, NGTDM Contrast |
| ∆IPSS | iCBCT SHARP VeryLow Uni 0 | 4 | GLRLM RLN, GLSZM SZE, IPSS Baseline, NGTDM Coarseness |
| ∆IPSS | iCBCT STD Med Llo 1 | 4 | GLRLM RLN, GLRLM RLV, GLRLM SRE, IPSS Baseline |
| ∆IPSS | iCBCT STD Med Uni 0* | 4 | GLRLM RLN, GLSZM SZLGE, GLSZM ZP, IPSS Baseline |
| ∆IPSS | sCBCT SHARP Llo 1* | 4 | GLRLM GLN, GLRLM RLN, Global Variance, IPSS Baseline |
| ∆IPSS | sCBCT SHARP Uni 0 | 3 | GLRLM RLN, IPSS Baseline, NGTDM Contrast |
| ∆IPSS | sCBCT STD Llo 1 | 6 | GLCM Variance, GLRLM GLN, GLRLM RLN, GLSZM LZE, Global Variance, IPSS Baseline |
| ∆IPSS | sCBCT STD Uni 0 | 6 | GLCM Correlation, GLRLM RLN, GLSZM GLN-VN, IPSS Baseline, NGTDM Coarseness, NGTDM Contrast |