The association of environmental exposure with multiple sclerosis severity score: A study based on sequential data...

Area of Science:

• Neuroscience
• Environmental Health
• Data Science

Background:

• Multiple Sclerosis (MS) is a heterogeneous neuroinflammatory disease.
• Predicting MS progression is complex due to individual variability.
• Clinical, demographic, and environmental factors influence MS.

Purpose of the Study:

• To apply Artificial Intelligence (AI) to predict MS Severity Score (MSSS) using clinical and environmental data.
• To investigate the association between environmental exposures and MS progression.
• To evaluate Deep Learning (DL) models for MSSS prediction.

Main Methods:

• Integrated longitudinal clinical records with environmental data (air pollution, weather).
• Employed a hybrid imputation strategy for missing data.
• Utilized Automated Machine Learning (AutoML) for feature selection and evaluated DL models (GRU, LSTM, RNN).

Main Results:

• AutoML identified clinical and environmental variables as key predictors.
• DL models with environmental data performed comparably or better than those with clinical data alone.
• The Gated Recurrent Unit (GRU) model achieved an Area Under the Curve of 0.814, with environmental factors like PM2.5 and NO2 being significant predictors.

Conclusions:

• Incorporating environmental exposures into DL models improves MSSS prediction accuracy.
• Diverse real-world data, including environmental factors, are valuable for MS monitoring.
• Data from approximately one year of monitoring (two prior follow-ups) may suffice for clinically meaningful MS progression predictions.