Multiple Sclerosis is a chronic neurological disease that can cause disability in young adults. Early and accurate diagnosis can optimize therapeutic strategies and slow disease progression. Although deep learning-based methods have shown promising results in the diagnosis of multiple sclerosis through imaging data, challenges such as protecting patient privacy and lack of model explainability limit their application in real-world clinical settings. In this paper we propose a federated learning-based approach integrated with explainability for detection and localisation of multiple sclerosis. The experimental analysis results show that the proposed method is able to obtain interesting classification performances, with an accuracy of 0.985, a precision of 0.979 and a recall measure of 0.978, while maintaining data confidentiality and by providing explainability behind the federated model prediction, as a matter of fact visualization of class activation mapping provides clear and clinically meaningful explanations of predictions.