Noninvasive KRAS mutation estimation in colorectal cancer using a deep learning method based on CT imaging

BMC Medical Imaging

Table 2 Performance of Models of the KRAS mutation prediction in testing cohorts

Model		AUC	Sensitivity	Specificity
Radiomics model		0.82	0.7	0.85
ResNet model
Axial directions	ROI only	0.9	0.65	0.83
	ROI + 10 pixels	0.9	0.67	0.83
	ROI + 20 pixels	0.93	0.59	1
	ROI + 30 pixels	0.72	0.67	0.63
Coronal direction	ROI only	0.75	0.79	0.56
	ROI + 10 pixels	0.71	0.83	0.46
	ROI + 20 pixels	0.58	0.45	0.7
	ROI + 30 pixels	0.51	0.7	0.28
Sagittal directions	ROI only	0.72	0.56	0.8
	ROI + 10 pixels	0.69	0.61	0.65
	ROI + 20 pixels	0.61	0.58	0.57
	ROI + 30 pixels	0.54	0.89	0.13

ISSN: 1471-2342