Background: Detailed neuro-imaging analyses of PCNSL are limited. Our objective was to evaluate MRI findings in PCNSL patients treated with hdMTX without radiotherapy, with a particular emphasis on T2-FLAIR MRI abnormalities. Methods: We reviewed MRI (T1, T1 post-gadolinium and T2/FLAIR) findings of 85 patients with PCNSL enrolled in a randomized multicenter phase II trial (NCT00503594) evaluating two hdMTX regimens (MPV-A and MTX-temozolomide), conducted in patients aged 60 and older. Response rate (IPCG criteria), number and volume of enhancing lesions, anatomical site, site of relapse and patterns of relapse were analyzed and correlated with outcome. Landmark analyses were performed at 2 and 4 months. Results: Objective responses (OR) were: complete response (CR): 56%, uncertain CR: 4%, partial response: 18%, stable disease: 7%, progressive disease: 15%. On multivariate analysis, location in the posterior fossa (p = 0.008) and tumor volume (p = 0.006), but not lesion number, were significantly associated with survival. OR at the first MRI (2months) and at the end of treatment (4 months) were significantly associated with overall survival in multivariate analysis (p < 0.001 and p = 0.004 respectively). Early versus delayed CR did not impact survival. Relapse in the brain involved the initial enhancing site, a site at distance, or both in 46%, 40% and 14% respectively. At baseline, non-enhancing T2-FLAIR hypersignal lesions at distance from the enhancing tumor site were detected in 18 patients (23%), among whom 16 (88%) displayed a marked decrease ( > 50%) after chemotherapy, confirming their neoplastic nature. Among these 18 patients, ten relapsed, and the initially non-enhancing T2-FLAIR lesions corresponded to the site of an enhancing relapse in half of them (N = 5). Conclusions: Relapse after hdMTX at distance from the initial enhancing tumor is frequent and may occur in the area of non-enhancing T2-FLAIR hyper-intensity identified at distance from the enhancing tumor at baseline. Future response criteria should take into consideration non-enhancing T2-Flair lesions.