Radiomics features (RFs) extract quantitative information from medical images, towards the derivation of biomarkers for clinical tasks such as diagnosis, prognosis, or treatment response assessment. Different image discretization parameters (e.g. bin number or size), convolutional filters, or multi-modality fusion levels can be used to generate radiomics signatures. Commonly, only one set of parameters is used; resulting in only one value or ‘flavour’ for a given RF. We propose ‘tensor radiomics’ (TR) where tensors of features calculated with multiple combinations of parameters (i.e. flavours) are utilized to optimize the construction of radiomics signatures.
An example of a tensor in tensor radiomics. Features describes extracted radiomics features, while the flavour types (A, B, C, D ...) encompass modifications to the original region of interest. Examples of these may be discretization bin size or convolutional filters. Only two variants are shown here, but in practice any number may be used to create higher dimension tensors.