"DEMENTIA CLASSIFICATION USING DEEP LEARNING APPROACHES"

Abstract

Dementia is a progressive neurological condition that impairs cognitive functions and affects millions worldwide. Early and accurate detection of its various stages is critical for effective management and treatment. This project, titled "Dementia Classification using Deep Learning Approaches," leverages cutting-edge deep learning algorithms, including Convolutional Neural Networks (CNN), MobileNet, and DenseNet, to classify dementia into four distinct stages: No Impairment, Very Mild Impairment, Mild Impairment, and Moderate Impairment. The model is trained on a robust and publicly available dataset of Alzheimer’s disease MRI scans, sourced from Kaggle. Data preprocessing and augmentation techniques ensure the model's resilience and accuracy. Comparative analysis of the three algorithms highlights DenseNet's high performance, achieving superior classification accuracy. The proposed system provides a reliable, efficient, and cost-effective diagnostic tool, empowering healthcare professionals with timely insights for better patient outcomes. By integrating lightweight yet powerful architectures like MobileNet, the project also ensures real-world applicability, especially in resource-constrained environments. This study aims to bridge the gap between early dementia detection and accessible diagnostic technology, paving the way for advancements in medical imaging analysis through deep learning.

Keywords

Dementia, Alzheimer's, MRI, deep learning, CNN, DenseNet, MobileNet, classification, early detection, medical imaging