Multiple Sclerosis (MS) is a chronic, immune-mediated CNS disease, that causes unpredictable inflammation episodes and leads to neurological disability in young adults (Kolčava et al., 2020). This disease is associated with clinically isolated syndrome (CIS), which is characterized by a single episode of demyelination symptoms such as optic neuritis or transverse myelitis (Ohlmeier et al., 2020). MS first presents as CIS in about 85 % of cases and progresses to Clinically Definite Multiple Sclerosis (CDMS) when a second relapse occurs (Kolčava et al., 2020). A previous study found that 45 % of CIS patients converted to MS after a mean follow-up of five years (Brownlee et al., 2015). Long-term studies indicate that 60–70 % of CIS patients will develop CDMS within of 20 years (Fisniku et al., 2008).

Determining the risk of conversion from CIS to CDMS is important, as a quick transition is related to a more severe prognosis and disability. This allows for tailored treatment plans: high-risk patients might benefit from potent treatments, despite their higher risks, while lower-risk patients could receive safer, but less effective treatments (Kappos et al., 2006; Kinkel et al., 2006). Therefore, a model that yields an estimation of the risk for developing CDMS is beneficial in a clinical setting, as it can guide the selection of more individualized treatment options (Rotstein and Montalban, 2019).

Machine learning (ML) techniques have recently gained popularity and are now widely applied in various research areas, including medicine, particularly for classification problems (Boucekine et al., 2013; Zhao et al., 2017). Many studies have focused on predicting the transition from CIS to CDMS using magnetic resonance imaging (MRI) data (Afzal et al., 2022, 2018; Pareto et al., 2022; Yoo et al., 2017; Zhang et al., 2019). However, only a few have made the effort to incorporate MRI, clinical, and demographic data in their models. This underlines the potential for expanding the scope of research in this field.

To the best of our knowledge, this study is the first to develop a ML model that utilizes a combination of MRI characteristics, clinical data, and demographic information to estimate the probability of CIS converting to CDMS. Furthermore, we strive to address the challenge of making the ML model explainable for non-technical individuals in a clinical setting. This approach enhances the predictive power of the model and ensures its practical applicability and understandability in real-world clinical scenarios (Ali et al., 2023).