eduzhai > Applied Sciences > Engineering >

An Explainable 3D Residual Self-Attention Deep Neural Network FOR Joint Atrophy Localization and Alzheimers Disease Diagnosis using Structural MRI

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Computer-aided early diagnosis of Alzheimer s disease (AD) and its prodromalform mild cognitive impairment (MCI) based on structure Magnetic ResonanceImaging (sMRI) has provided a cost-effective and objective way for earlyprevention and treatment of disease progression, leading to improved patientcare. In this work, we have proposed a novel computer-aided approach for earlydiagnosis of AD by introducing an explainable 3D Residual Attention Deep NeuralNetwork (3D ResAttNet) for end-to-end learning from sMRI scans. Different fromthe existing approaches, the novelty of our approach is three-fold: 1) AResidual Self-Attention Deep Neural Network has been proposed to capture local,global and spatial information of MR images to improve diagnostic performance;2) An explanation method using Gradient-based Localization Class Activationmapping (Grad-CAM) has been introduced to improve the explainable of theproposed method; 3) This work has provided a full end-to-end learning solutionfor automated disease diagnosis. Our proposed 3D ResAttNet method has beenevaluated on a large cohort of subjects from real datasets for two changelingclassification tasks (i.e., Alzheimer s disease (AD) vs. Normal cohort (NC) andprogressive MCI (pMCI) vs. stable MCI (sMCI)). The experimental results showthat the proposed approach has a competitive advantage over thestate-of-the-art models in terms of accuracy performance and generalizability.The explainable mechanism in our approach is able to identify and highlight thecontribution of the important brain parts (e.g., hippocampus, lateral ventricleand most parts of the cortex) for transparent decisions.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×