Beyond complex architectures: a streamlined CNN pipeline for robust Alzheimer's disease classification from brain MRI
Mohamed Amine Jabli1, Moussa Mourad2
1ENISO, NOCCS, University of Sousse, Sousse, Tunisia. jeblimedamine@gmail.com.
Background And Purpose:
Alzheimer's disease, a common type of dementia, gradually steals memories and impacts daily life as brain cells deteriorate. We explored how Artificial Intelligence (AI) could help spot early signs of Alzheimer's using MRI brain scans.
Materials And Methods:
Our study focused on a deep learning approach, specifically a convolutional neural network (CNN), to distinguish between Alzheimer's, Mild Cognitive Impairment (MCI), and healthy individuals. We used two well-known datasets, OASIS and ADNI, for this work. After carefully preparing the ADNI data (which included 21,324 MRI images: 7,572 from MCI patients, 5,904 from healthy controls, and 7,848 from Alzheimer's patients) and the OASIS data (6,400 MRI images), our model performed exceptionally well.
Results:
The CNN Model achieved an accuracy of 99.67% with the ADNI images and 99.06% with the OASIS images. These encouraging results, which stand up well against other studies, show that our method can effectively analyze large amounts of data and accurately classify Alzheimer's.
Conclusions:
Our main hope is that this kind of technology can give doctors and caregivers better tools to predict and detect the disease, ultimately saving time, reducing costs, and helping those affected by Alzheimer's.
