This master thesis has two aims. The first main objective is to study and obtain robust voxel-wise (3D) radiomic features for lung cancer and with them compute the habitats (unsupervised clusters) to unravel tumor heterogeneity and facilitate a personalized treatment for each patient and to see the progression of the disease over time.The goal is to obtain robust radiomic features that can be generalized to every database with lung cancer data. The second main objective is to develop a CT-radiomic signature (i.e., imaging phenotype) that will predict the tumor immune profile obtained from the RNA analysis of biopsied sample and allow to identify patients with lung cancer who will respond to immunotherapy.