A DEEP LEARNING-BASED FRAMEWORK FOR ALZHEIMER'S DISEASE DIAGNOSIS AND PROGRESSION PREDICTION FROM MRI IMAGES
Abstract
Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by continuous cognitive and biomarker changes. Identifying predictive biomarkers for the progression from mild cognitive impairment (MCI) to AD is crucial for improving diagnostic accuracy and facilitating targeted drug development. Traditional machine learning methods have been used for AD diagnosis using MRI images, but deep learning (DL) techniques offer superior performance due to their ability to learn complex patterns from large datasets. A significant challenge in AD research is the lack of definitive biomarker signatures to predict which MCI patients will progress to AD. Current methods, such as DenseNet121, ResNet50, VGG 16, EfficientNetB7, and InceptionV3, have limitations in predictive accuracy and the ability to identify distinct neurodegenerative patterns. We developed and validated YoLoV7, a deep learning-based framework that examines neuroanatomical heterogeneity contrasted against normal brain structure to identify disease subtypes through neuroimaging signatures. Using a dataset of 1000 participants and 4000 T1-weighted MRI scans, YoLoV7 identified four patterns of neurodegeneration. We applied this framework to longitudinal data to reveal two distinct progression pathways and assessed the model's performance in predicting the pathway and rate of future neurodegeneration. YoLoV7 identified four neurodegenerative patterns, with Pattern 1 showing a 70% accuracy in predicting slower progression (Pathway A) and Pattern 2 demonstrating an 80% accuracy in predicting rapid progression (Pathway B). Overall, YoLoV7 achieved an accuracy of 85% in predicting clinical progression, outperforming traditional methods such as DenseNet121 (78%), ResNet50 (75%), VGG 16 (72%), EfficientNetB7 (80%), and InceptionV3 (76%). Measures of pattern expression offered complementary performance to amyloid/tau biomarkers in predicting clinical progression.

Authors
M. Rajalakshmi1, C. Karthik2
Thiagarajar College of Engineering, India1, Jyothi Engineering College, India2

Keywords
Alzheimer's Disease, Deep Learning, MRI, Neurodegeneration, Progression Prediction
Yearly Full Views
Published By :
ICTACT
Published In :
ICTACT Journal on Image and Video Processing
( Volume: 14 , Issue: 4 , Pages: 3251 - 3257 )
Date of Publication :
May 2024
Page Views :
Full Text Views :

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.