Litcius/Paper detail

A Novel Residual 3-D Convolutional Network for Alzheimer's disease diagnosis based on raw MRI scans

Varun Ullanat, Vinay Balamurali, A V L Narayana Rao

202113 citationsDOI

Abstract

One of the most widely used deep learning architectures for image classification, Convolutional Neural Networks (CNNs) are used in a diverse range of research areas. Over the past five years, CNNs have been extended for use in disease classification and diagnosis based on body imaging data. In this paper, we propose one such CNN model to diagnose Alzheimer's Disease using raw, volumetric Magnetic Resonance Imaging (MRI) scans. The MRI dataset used contains 857 scans (302 AD and 555 Normal Control) in total and was procured from the ADNI study. The performance of the proposed residual CNN was compared with 3-D ResNet-18 with and without an attention mechanism. Finally, 3-D Gradient-weighted Class Activation Mapping was used to evaluate how effective the models were in recognizing brain regions pertaining to AD. The best performing model obtained an accuracy of 91%.

Topics & Concepts

Convolutional neural networkArtificial intelligenceComputer scienceResidualDeep learningPattern recognition (psychology)Magnetic resonance imagingContextual image classificationResidual neural networkRaw dataNeuroimagingAlzheimer's Disease Neuroimaging InitiativeAlzheimer's diseaseDiseaseImage (mathematics)RadiologyMedicineNeurosciencePathologyPsychologyAlgorithmProgramming languageBrain Tumor Detection and ClassificationAdvanced Neural Network ApplicationsMedical Imaging and Analysis
A Novel Residual 3-D Convolutional Network for Alzheimer's disease diagnosis based on raw MRI scans | Litcius