Although adjuvant radiotherapy in combination with Temozolomide administration has clearly demonstrated the benefit in improving the prognosis of patients with multiforme glioblastoma, radiotherapy as only treatment or in combination with systemic treatment is one of the best supportive in unresectable cases. For recurrent cases, the salvage radiotherapy option (re-irradiation) can be chosen in carefully selected cases so that the benefit is greater than the toxicities. Radiomics, a new subdomain of artificial intelligence (AI), relies on advanced analysis in high-resolution medical imaging to establish diagnostic, prognostic and predictive models for clinical medicine. The variation of the delta-radiomics parameters analyzed within a tumor volume may be via tumor heterogeneity indirectly correlated with the response to treatment. The aim of the study is to propose a delta-radiomic based on entropy algorithm to allow the non-invasive pre-therapeutic identification of patients with unresectable or recurrent multiform glioblastoma who will benefit from irradiation and/or salvage re-irradiation.