Abstract
The course of multiple sclerosis begins with a relapsing-remitting phase, which evolves into a secondarily progressive form over an extremely variable period, depending on many factors, each with a subtle influence. To date, no prognostic factors or risk score have been validated to predict disease course in single individuals. This is increasingly frustrating, since several treatments can prevent relapses and slow progression, even for a long time, although the possible adverse effects are relevant, in particular for the more effective drugs. An early prediction of disease course would allow differentiation of the treatment based on the expected aggressiveness of the disease, reserving high-impact therapies for patients at greater risk. To increase prognostic capacity, approaches based on machine learning (ML) algorithms are being attempted, given the failure of other approaches. Here we review recent studies that have used clinical data, alone or with other types of data, to derive prognostic models. Several algorithms that have been used and compared are described. Although no study has proposed a clinically usable model, knowledge is building up and in the future strong tools are likely to emerge.
Highlights
Accepted: 30 January 2021In the last decade, artificial intelligence (AI) approaches—and in particular machine learning (ML)—have been increasingly applied within the medical field, with the hope of increasing diagnostic performance and efficacy of care
It is known that ML performs better when the number of features is adequate to the number of records available for the study [22–24], and that elimination of noninformative features can reduce the effect of noise and the risk of obtaining complex models that overfit data
Several techniques have been developed to determine the importance of features and how they contribute to specific predictions, such as Local Interpretable Model-agnostic Explanations (LIME) [65], DeepLift [66] and, recently, the novel Shapley additive explanations method (SHAP) [67]
Summary
Artificial intelligence (AI) approaches—and in particular machine learning (ML)—have been increasingly applied within the medical field, with the hope of increasing diagnostic performance and efficacy of care. Deep neural network algorithms attain performances similar to those of well-trained radiologists in examining medical images, and AI algorithms are being approved by regulatory agencies (see [3] for a recent meta-analysis) This general frame applies to neurology, with neuroradiology at the forefront of ML application [4]. Patients with mild forms should avoid overtreatment, with substantial benefits in terms of safety, quality of life and overall allocation of resources To this aim, reliable early prognosis would be extremely helpful. The presence of oligoclonal bands in the cerebrospinal fluid at diagnosis may predict worse prognosis with high disability Correlation between these factors and the evolution of MS is established at the population level, none of the prognostic factors or risk scores for early [15–17] or late [18] disease course has been validated. ML approaches have been increasingly tested with regard to their capacity to provide support to patient counseling, prognosis and therapy
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
