Table 2.

Overview of the models. The models (M1-M7, column 1) are sorted with increasing level of complexity. Input data (D) to each model is specified in the last two columns.

ID	Model architecture	Model type	Input data type	Input data
M1	Logistic regression	Conventional machine learning	Tabular data	Clinical data (D1), Radiomics data (D2) or all tabular data (D1 + D2)
M2	Random forest	Conventional machine learning	Tabular data	D1, D2 or D1 + D2
M3	Neural network without interaction	Deep learning (FCNN)	Tabular data	D1, D2 or D1 + D2
M4	Neural network with interaction	Deep learning (FCNN)	Tabular data	D1, D2 or D1 + D2
M5	EfficientNet3D CNN	Deep learning (CNN)	Image data	PET/CT (D3)
M6	EfficientNet3D CNN	Deep learning (CNN)	Image data	PET/CT & GTVp (D3)
M7	EfficientNet3D CNN	Deep learning (CNN)	Image data	PET/CT, GTVp & GTVn (D3)

FCNN, fully connected neural network; CNN, convolutional neural network.