Background: The majority of real-world data (RWD) on disease-modifying treatments (DMTs) for multiple sclerosis (MS) is based on clinical outcomes. Performing studies using real-world systematic follow up MRI data poses challenges involved in data gathering, curation, standardization and analysis, thus its value remains unexplored for many DMTs. To address this, research on the use of MRI-RWD to assess DMTs is required to understand how the results compare with clinical trials, and to advance the development of methodology adequate for analysing RWD. In these analyses, we show the first results of a real-world quantified MRI study of dimethyl fumarate (DMF). Objectives: To evaluate the feasibility of using RWD for the assessment of brain atrophy on DMF-treated relapsing MS patients. Methods: MRI data of DMF-treated patients was collected from real-world clinical practice in 2 US MS Centers. A total of 381 time points were collected from 118 RRMS patients, with follow-up duration up to 4 years. Data curation involved the selection of patients with at least 2 time points fulfilling a set of minimal imaging requirements, i.e., presence of T1-weighted and FLAIR series with slice thickness below 5 and 6 mm (respectively), adequate field of view (FOV) and no gross imaging artifacts. Visual quality control was performed to reduce the impact of scanner and protocol switches on atrophy measurements. Annualized Percent-Brain-Volume-Change (aPBVC) was determined with respect to the first available time point using icobrains longitudinal pipeline. Results: A total of 252 (66% time points) were accepted during Quality Control, for a total of 98 patients. Main reasons for rejection included not having T1 and FLAIR series with adequate resolution (55% rejections), followed by protocol switches with accentuated differences in image contrast and resolution (23%). Additional reasons included the existence of a single adequate time point (13%), signal artifacts (5%) and limited FOV (3%). Brain atrophy was analyzed in the patients meeting inclusion criteria. Mean and SD annualized PBVC of all included DMF-treated subjects was -0.29±0.83% per year, with a median value of -0.29%. Conclusions: We identify the challenges inherent in utilizing real world imaging data to address the efficacy of DMF beyond randomized control trials. In particular, quantitative image analysis may be hampered by suboptimal protocols in terms of slice thickness and inconsistency in terms of imaging protocols over time. Despite these challenges, the study shows that atrophy values as detected from DMF-treated subjects in the real world follow those of randomized controlled trials. These findings may facilitate optimizing MS-specific MRI protocols to capture these metrics prospectively.