A Bayesian Network Approach for Friedreich Ataxia Severity Classification using Probability Modelling.

Area of Science:

• Neurology
• Biomedical Engineering
• Data Science

Background:

• Friedreich ataxia (FRDA) lacks objective severity measures, hindering treatment trials.
• Rare disease datasets are small, challenging conventional machine learning.
• Existing quantitative ataxia measures underutilize expert clinical knowledge.

Purpose of the Study:

• To develop an objective severity measure for FRDA using Bayesian Networks.
• To integrate subjective clinical assessments and objective instrumented measurements.
• To address limitations of small datasets and underutilized expert knowledge in FRDA research.

Main Methods:

• Utilized a hybrid learning approach with Bayesian Networks (BNs).
• Incorporated subjective clinical assessments and instrumented upper body movement data.
• Trained the BN model on data from individuals with FRDA.

Main Results:

• The BN model achieved a 0.93 Pearson correlation with low error.
• Predicted clinical scales with high accuracy: 94% for Upright Stability/Lower Limb Coordination.
• Showed 67% accuracy for Functional Staging, Upper Limb Coordination, and Activities of Daily Living.

Conclusions:

• Bayesian Networks offer a viable solution for rare disease severity assessment.
• This hybrid approach effectively combines expert knowledge and objective data.
• The developed model can serve as a clinical decision support system for FRDA severity prediction.