An Ensembling PLMRI: Alzheimer Detection Via Magnetic Resonance Imaging with Prompt Learning Model

MD Zubeda Banu1
Department of Computer Science and Engineering
Sri Venkateswara College of Engineering,
Karkambadi
Tirupati, India, 517501
md.zubedabanu786@gmail.com

A Ganesh2
Department of Computer Science and Engineering
Sri Venkateswara College of Engineering,
Karkambadi
Tirupati, India, 517501
ganesh.a@svcolleges.edu.in

Abstract— Introduces an end-to-end framework for detecting and classifying stages of Alzheimer’s disease using Magnetic Resonance Imaging (MRI) data. Thesystem leverages Convolutional Neural Networks (CNN) to analyze 6,400 pre-processed sagittal plane MRI slices sourced from Hugging Face. Images arecategorized into four clinically relevant classes: NonDemented, Very Mild Demented, Mild Demented, and Moderate Demented. The methodology follows astructured pipeline beginning with comprehensivepreprocessing, including pixel normalization, label  encoding, and data augmentation techniques to improveconvergence and mitigate overfitting. The CNNarchitecture comprises three convolutional layers with 32, 64, and 128 filters, each activated by Re-LU, followed by max-pooling operations. A dropout regulated dense layer ensures robust feature extraction, capturing structural brain changes such as atrophy. To enhance interpretability and diagnostic transparency,Gradient-weighted Class Activation Mapping (Grad CAM) is integrated, generating heatmaps that highlight criticalbrain regionsinfluencing predictions, particularly the hippocampus. This visualization bridges the gap between automated classification and clinical validation, offering insights into the model’s decision making process. Overall, the proposed system demonstrates the potential of deep learning in medical imaging, providing a scalable and transparent approachfor early-stage Alzheimer’s detection and supporting  clinical decision-making diagnostics. in neurodegenerative
Keywords— MRI, CNN, Re-LU Activation Mapping, Class Activation Mapping.