Myo-Guide

Myo-Guide assists clinicians in diagnosing neuromuscular diseases through machine learning analysis of muscle MRI scans, identifying patterns of intramuscular fat replacement to support differential diagnosis.

Key Features:

• Multi-study harmonization pipeline: Standardizes MRI fat content across datasets to minimize source bias.
• Ensemble classification: Uses an ensemble of XGBoost models trained to classify patients based on intramuscular fat replacement, age at MRI and sex.
• Explainable predictions: Adapts the SHapley Additive exPlanations (SHAP) framework to analyse model predictions and identify disease-specific muscle involvement patterns.
• Class-balanced evaluation: Training and evaluation use class-balanced metrics to address class imbalance.

Scientific Applications:

• Neuromuscular disease diagnosis: Classifies patients across 20 paediatric and adult neuromuscular diseases using genetically confirmed cases.
• Differential diagnosis support: Identifies key features relevant for differential diagnosis to aid clinical decision-making.
• Diagnostic ranking comparison: Provides automated diagnostic ranking that can be compared against expert clinician performance.

Methodology:

The tool employs a preprocessing pipeline to standardize MRI fat content across datasets, followed by an ensemble of XGBoost models for classification. The SHAP framework is adapted to analyse predictions and identify disease-specific muscle involvement patterns.