Clinical Impact Through AI-assisted MS Care - A Retrospective Multi-center Observational Study.

There is a clear need for a data-driven and personalized treatment optimisation tool for people with Multiple Sclerosis (MS), in order to enable/support physicians to deploy appropriate therapeutic measures that will help to better slow down disease progression and eventually, progressive disability worsening. While early diagnosis and prognostic modelling is important to make data-driven recommendations for treatment optimisation, being able to disentangle and monitor the disability accumulation due to 'relapse associated worsening' or due to 'progression independent of relapse activity' will be key to optimizing treatment for the best possible long-term outcomes. The latter strongly depends on the availability of biomarkers that can detect and differentiate between these different forms of disease worsening.

With the RECLAIM study, we focus on gathering a centralized and harmonized dataset, enabling the secondary use of data to support prognosis for people with MS, as well as treatment optimisation in a real-world setting. As such, RECLAIM aims to develop MRI-based tools to better monitor disease progression in people with MS, as well as AI-based models that will support prognosis of individual disease course and treatment response, comprising: (i) a biomarker-based MS progression model, (ii) an MRI-focused generative model to predict brain characteristic evolution, and (iii) an interventional model for treatment optimisation. Additionally, the data will be used to generate further insights on Multiple Sclerosis progression as well as to develop the tools to monitor this progression.

• The representativeness of the harmonised dataset for the MS patient population as evaluated by age range, gender balance, the distribution of country of residence, the distribution of race/ethnicity and the distribution of educational level [Time frame: 4 years]
• The validity of the data through an assessment of the amount of erroneous or impossible data entries for each variable. [Time frame: 4 years]
• The temporal uniformity of each institution's data over time as assessed by the number of changes to variables over time (addition of new variables or variables no longer being captured, alterations to how variables are captured). [Time frame: 4 years]
• The temporal uniformity of the harmonised dataset over time as assessed by the average time between subsequent assessments of each variable. [Time frame: 4 years]
• The presence of contextual information on standard data gathering and analysis processes of each institution [Time frame: 4 years]
• The presence of a unique and pseudonymised patient ID for all data of each patient, allowing to link such data of each patient. [Time frame: 4 years]
• The temporal uniformity of MRI data over time as assessed by the comparability of MRI scans and the average time between subsequent MRI assessments for each patient. [Time frame: 4 years]
• The percentage of MRI data sets which are compliant with the MAGNIMS-CMSC-NAIMS acquisition guidelines. [Time frame: 4 years]
• The percentage of MRI data sets for which the automated quality control process of icobrain ms did not indicate any quality issues upon analysis. [Time frame: 4 years]
• The percentage of patients with a complete disease modifying treatment history available, from the date of diagnosis to the current day. [Time frame: 4 years]
• The percentage of patients with a complete disease history available, from the date of diagnosis to the current day. [Time frame: 4 years]
• The validity and temporal uniformity for disability assessment as clinically determined by EDSS, Functional systems score, T25FWT, 9HPT and SDMT. [Time frame: 4 years]