A convolutional neural network classifier for Parkinson’s disease on neuromelanin-sensitive MRI

Takeaway

  • A novel, convolutional neural network (CNN)-based classifier is more accurate than contrast ratio-based predictions and predictions based on radiomic features computed from the substantia nigra pars compacta (SNc) for accurately diagnosing Parkinson’s disease (PD) on neuromelanin-sensitive MRI (NMS-MRI).

Why this matters ?

  • Current techniques for analyzing NMS-MRI to accurately diagnose PD have limited clinical utility and reproducibility due to the time-consuming nature of computing manually extracted features of the SNc and the requirement for well-defined SNc borders.