Towards a predictive model for Guillain-Barré syndrome.

Area of Science:

• Neurology
• Computational Biology
• Medical Informatics

Background:

• Guillain-Barré Syndrome (GBS) presents with varying severity across distinct subtypes.
• Key subtypes include Acute Inflammatory Demyelinating Polyneuropathy (AIDP), Acute Motor Axonal Neuropathy (AMAN), Acute Motor Sensory Axonal Neuropathy (AMSAN), and Miller-Fisher Syndrome (MF).

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in classifying GBS subtypes.
• To compare the predictive performance of C4.5 decision tree, Support Vector Machines (SVM), and k-Nearest Neighbors (kNN) for GBS subtyping.

Main Methods:

• Utilized a dataset of 129 GBS patients, including clinical, serological, and nerve conduction test data.
• Applied C4.5 decision tree, Gaussian kernel SVM, and kNN algorithms for subtype prediction.
• Employed 30 iterations of 10-fold cross-validation (10-FCV) to ensure robust accuracy assessment.

Main Results:

• C4.5 decision tree achieved the highest average accuracy of 0.9211 (±0.0109).
• kNN demonstrated strong performance with an average accuracy of 0.9179 (±0.0041).
• SVM (Gaussian kernel) yielded an average accuracy of 0.9154 (±0.0069).

Conclusions:

• Machine learning models, particularly C4.5, show high accuracy in predicting Guillain-Barré Syndrome subtypes.
• These computational approaches offer a promising tool for objective GBS classification.
• Further research and experimentation are ongoing to refine predictive models.