Regularization Techniques in Radiomics: A Case Study on the Prediction of pCR in Breast Tumours and the Axilla

Clinicians have shown an increasing interest in quantitative imaging for precision medicine. Imaging features can extract distinct phenotypic differences of tumours, potentially they can be used as a non-invasive prognostic tool and contribute for a better prediction of pathological Complete Response (pCR). However, the high-dimensional nature of the data brings many constraints, for which several approaches have been considered, with regularization techniques in the cutting-edge research front. In this work, classic lasso, ridge and the recently proposed priority-lasso are applied to high-dimensional imaging data, regarding a binary outcome. A breast cancer dataset, with radiomics, clinical and pathological information as features, was used. The application of sparsity techniques to the dataset enabled the selection of relevant features extracted in MRI of breast cancer patients, in order to identify the accuracy of those features and predict the pCR in the breast and the axilla.