The work conducted in this project has produced an in-depth study of pathology histology images acquired from non-small cell lung cancer patients, with the aim of predicting if the patients will reduce or not the tumour mass during the therapy, offering the chance to deploy personalized approaches for resulting in treatment adaptation. To this end we investigate pathomics as the microscale counterpart of radiomics, studying different deep neural networks as well as transfer learning, data augmentation and data normalization approaches, even in a multimodal fashion. The results obtained are promising, despite the presence of unbalanced data.