ORIGINAL RESEARCH article
Front. Neurosci.
Sec. Brain Imaging Methods
Towards Accurate Alzheimer's Detection: Transfer Learning with ResNet50 for MRI-Based Diagnosis
Provisionally accepted
Mohamed Amine Jebli*
Moussa Mourad- The National Engineering School of Sousse (ENISo), Sousse, Tunisia
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
ABSTRACT: In order to improve patient outcomes, Alzheimer's disease (AD), the most common dementia that affects over 50 million people worldwide, requires an accurate and timely diagnosis. Scalability is limited by the labor-intensive manual feature extraction used in traditional machine learning for AD detection via MRI. This article presents an automated method for extracting features from brain MRI scans using the pre-trained ResNet50 convolutional neural network (CNN), which is assessed using Softmax, SVM, and RF classifiers on the ADNI and MIRIAD datasets. The ResNet50-Softmax model outperformed state-of-the-art benchmarks (85.7%-98.59%) with 99% sensitivity and 98% specificity, achieving an impressive 99% accuracy on ADNI (96% on MIRIAD). These findings demonstrate how transfer learning can expedite the diagnosis of AD by providing a high-accuracy, scalable method for clinical neuroimaging and addressing the anticipated threefold increase in AD cases by 2050.
Keywords: Alzheimer's, deep learning, Transfer Learning, ADNI (Alzheimer's DiseaseNeuroimaging Initiative), Classification
Received: 11 Jul 2025; Accepted: 25 Nov 2025.
Copyright: © 2025 Jebli and Mourad. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Mohamed Amine Jebli
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.