<p>Dementia is a progressive neurological disorder that disrupts memory, cognitive abilities, and daily activities. Achieving an accurate diagnosis at an early stage is essential for timely treatment and preventive measures. However, conventional diagnostic techniques, such as neuropsychological assessments and manual interpretation of neuroimaging, often involve lengthy procedures, subjective judgments, and susceptibility to error. Emerging deep learning technologies, especially convolutional neural networks, offer a powerful solution by automating MRI scan analysis and enhancing diagnostic precision. An advanced CNN framework was employed to distinguish between four dementia types —non-demented, very mild, mild, and moderate—by analyzing high-resolution structural brain images, enabling robust classification performance. The dataset, which consisted of 6430 MRI scans, underwent structured preprocessing, including resizing, normalization, and noise reduction, to enhance feature clarity. The CNN model was trained using TensorFlow/Keras, employing five convolutional layers with max-pooling, followed by fully connected layers and a softmax classifier for multiclass classification. Adam optimizer and early stopping techniques were used to improve performance and prevent overfitting. The experimental results showed a test accuracy of 99.11%, with high precision and recall across all classes. The confusion matrix confirmed minimal misclassifications, thereby demonstrating the robustness of the model. The research findings from the present investigation show the effectiveness of the CNN model in dementia classification, offering a reliable tool for early diagnosis and clinical decision-making.</p>