Table 3.

Studies related to the survival prediction performance of radiogenomic models.

Disease	Image modality	Number of patients	Outcome	Results	Study
GBM	MRI	73	Texture features computed from the JIMs of GBM subregions are combined with GLCM and gene expression features are used to build a radiogenomics signature that classifies patients into short or long survival groups	Classification accuracy $\mathrm{AUC}=0.78$	Ref. 44
Breast cancer	DCE-MRI	56	Multiparametric imaging phenotype vector extracted from tumor regions was used to classify tumors at low versus medium versus high risk of recurrence	Classification accuracy $\mathrm{AUC}=0.82$	Ref. 45
Breast cancer	MRI	84	MR imaging phenotype used to evaluate risk of recurrence relative to multigene assay classifications	Prediction accuracy AUC: MammaPrint-0.88	Ref. 46
				Oncotype DX: 0.76
				PAM50: 0.68
Breast cancer	Digital mammograms	71	Radiogenomics signature used to predict Oncotype DX and PAM50 recurrence scores	Prediction accuracy AUC:	Ref. 47
				Oncotype DX: 0.83
				PAM50: 0.78
Breast cancer	FDG-PET/CT	73	Metabolic radiomic signature is associated with Ki67 expression achievement of pathologic complete response NAC and risk of recurrence	Metabolic radiomics patterns of LABC are associated with Ki67 expression (statistically significant $p\text{\hspace{0.17em}value}<0.01$)	Ref. 48
NSCLC	CT	44	Association between 8-week tumor volume decrease and survival	Association with overall survival (Cox model $p$ value-0.01)	Ref. 49
NSCLC	CT	172	Radiogenomic biomarker used to discriminate $\mathrm{ALK}+$ from non-ALK tumors and identify patients with a shorter PFS	Discriminatory power $\mathrm{AUC}=0.894$	Ref. 50