Riemannian Covariance Descriptors for Medical Signal and Image Classification

Key Points

• To evaluate how the Riemannian metric affects classification accuracy in medical diagnostics.
• Compared diagnostic performance of Riemannian metrics versus Euclidean distance.
• Analyzed brain tumour classification from MRI and myocardial infarction detection from ECG.
• Measured accuracy improvements without additional data or training.
• Achieved a 5.67% accuracy increase in brain tumour classification (p=0.016).
• Achieved an 8.15% accuracy increase in myocardial infarction detection (p=0.008).