Table 1.

Overview of radiomics based outcome prediction models.

Study	Features	Extraction software (ES) and statistical method (SM)	Outcome prediction (OP) and performance (P)	Number of patients	Tumor characteristics
Yu et al. (64)	MeanBreadth Spherical disproportion	ES: IBEX SM: GLM	HPV status AUC: 0.86667 and 0.91549	315 pts	Oropharyngeal cancers
Ou et al. (71)	24 features signature	ES: Oncoradiomics SM: Logistical regression	5 y survival P: AUC = 0.67 CI (0.58–0.76)	120 pts	stage III – IVb Head and Neck cancer from (71)
Aerts et al. (4)	Statistics Energy Shape compactness 2 Grey level non uniformity Run length non-uniformity	ES: IBEX SM: Logistic regression	Overall Survival C-index: 0.69	545 pts	Lung and head and neck cancers
Anderson (79)	Intensity direct local range max Neighbour intensity difference 2.5 complexity	ES: IBEX SM: Decision Tree model	5 y Tumour control classifier (3 groups)	465 pts	Oropharyngeal cancers
Kann et al. (67)	No prior extraction for the selected model	ES: PyRadiomics SM: Random Forest and CNN	Extra nodal extension AUC = 0.91 (95%CI:0.85–0.97). NPV: 0.95	270 pts	Nodal invasion in resected head and neck cancer
Zhang et al. (81)	8 features signature	ES: Matlab: MRI based features. SM: LASSO and Rad-Score	3 y PFS C-index: 0.737 (95% CI: 0.549–0.924)	118 pts	Nasopharyngeal carcinomas
Li et al. (82)	8 features signature	ES: PyRadiomics on SPAIR T2W MRI SM: ANN	In field recurrence Accuracy: 0.812	306 pts	Nasopharyngeal carcinomas
Zhang et al. (83)	7 features signature	ES: PyRadiomics MRI based SM: Logistic regression	Distant Metastatic MRI based Model AUC : 0.827	176 pts	Nasopharyngeal carcinomas

GLM, General Linear model; CNN, Convolutional Neural Network; LASSO, Least Absolute Shrinkage and Selection operator; SPAIR T2W, spectral attenuated inversion-recovery T2-weighted; ANN, Artificial Neural Netwo.