Regresar

A Sharper Look: Edge Detection and Light-Weight 3D Vision Transformers for Parkinsonian Disorder Classification on MRI Imaging

Abstract:

Parkinson's disease (PD) is a progressive neurological disorder that affects motor function and is often accompanied by non-motor symptoms, including impulse control disorders (ICDs) and apathy. Early and accurate diagnosis and treatment of ICDs are crucial for improving patient outcomes. This research investigates the advantages of using deep learning and edge detection filter for anatomical pattern recognition associated with Parkinsonian Disorders. We used three transformer-based architectures to analyze the preprocessed MRI T1 weighted images and to compare the strengths and weaknesses of each architecture. The Edge detection filter improves the model accuracy and efficiency for all architectures. The MobileViT-3D achieves a top-1 accuracy of 99.13% on a dataset of 3D MRI images, demonstrating its potential as a promising approach for anatomical pattern recognition associated to PD-related disorders.