Table 2.

Radiomics and artificial intelligence (AI) studies in prognostication

References	Imaging modality	Major end point	N (training + test cohorts)	Number of features in the model	Evaluation metric	Analytical method used	Study description	Validation type
Aerts et al. 2014	CT	OS	422 + 225	4	Concordance index = 0.65	Cox proportional hazards regression	Prognostic power of radiomic features and the underlying gene-expression patterns	External validation
Grove et al. 2015	CT	OS	61 + 47	1	HR = 0.31 and 2.36, P-value: 0.008	Cox proportional hazards regression	Prognostic power of newly developed radiomic features	External validation
Tunali et al. 2017	CT	OS	61 + 47	1	HR = 0.40, P-value: 0.014	Cox proportional hazards regression	Prognostic power of newly developed radiomic features	External validation
Coroller et al. 2015	CT	OS and distant metastasis	98 + 84	3	Concordance index = 0.61	Cox proportional hazards regression	Predicting distant metastasis	External validation
Wu et al. 2016a	PET	Distant metastasis	70 + 31	2	Concordance index = 0.71	Cox proportional hazards regression	Predicting distant metastasis	Internal validation with a separate data set
Huang et al. 2016	CT	DFS	141 + 141	5	HR: 1.77 concordance index = 0.691	LASSO	Predicting DFS in early-stage patients	Internal validation with a separate data set
Huynh et al. 2017	CT	OS	131	13	AUC = 0.667	Correlation analysis (Spearman's correlation coefficient)	Early-stage disease recurrence prediction of patients treated with SBRT	No validation
Li et al. 2017	CT	OS	92	2 radiomic + 1 clinical + 1 semantic	Log-rank P-value = 0.0002	Cox proportional hazards regression	Early-stage disease survival prediction of patients treated with SBRT	No validation
Oikonomou et al. 2018	CT + PET	OS	150	7	Log-rank P-value = 0.002	Cox proportional hazards regression	Predict clinical outcome in lung cancer patients treated with SBRT	No validation
Win et al. 2013	FDG-PET/CT	OS	56 + 66	2 radiomic + 1 clinical	Cox regression P-value < 0.001	Cox proportional hazards regression	Predicting OS	External validation
Chae et al. 2014	CT	Prognostic	86	2	AUC = 0.981	ANN	Differentiate preinvasive lesions from invasive pulmonary adenocarcinomas	No validation
She et al. 2018	CT	Prognostic	207 + 195	5	AUC = 0.95	Logistic regression	Differentiate indolent from invasive pulmonary adenocarcinomas	Internal validation with a separate data set

(CT) Computed tomography, (DFS) disease-free survival, (SBRT) stereotactic body radiation therapy, (LASSO) least absolute shrinkage and selection operator, (ANN) artificial neural network, (OS) overall survival, (HR) hazard ratio, (FPR) false positive rate, (FNR) false negative rate.