A Deep Learning Approach to Diagnosing Multiple Sclerosis from Smartphone Data.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Digital Health

Background:

• Multiple sclerosis (MS) is a central nervous system disorder with diverse symptoms impacting daily life.
• Current MS diagnosis relies on complex clinical assessments and tests to exclude other conditions.
• Objective, long-term symptom monitoring is needed for improved MS assessment.

Purpose of the Study:

• To develop and validate a deep-learning approach for diagnosing MS using smartphone-derived digital biomarkers.
• To investigate the potential of smartphone monitoring for quantifying MS symptom presence and intensity.

Main Methods:

• A deep-learning model combining a multilayer perceptron with neural soft attention was developed.
• The model analyzed long-term smartphone monitoring data from 774 participants.
• Digital biomarkers were extracted from smartphone usage patterns.

Main Results:

• The deep-learning model achieved an area under the receiver operating characteristic curve of 0.88 in distinguishing individuals with and without MS.
• The model demonstrated effectiveness in learning patterns from long-term smartphone monitoring data.
• Results indicate high accuracy in MS detection using digital biomarkers.

Conclusions:

• Smartphone-derived digital biomarkers show promise for objective MS symptom assessment.
• This technology could serve as a supplementary tool for MS diagnosis.
• Future research may integrate digital biomarkers into clinical diagnostic criteria for MS.