A Multi-Stream Convolutional Neural Network Architecture for Alzheimer’s Disease Classification

Alzheimer's disease (AD) represents a progressively advancing neurodegenerative condition that considerably disrupts memory and cognitive abilities, thereby underscoring the necessity for early and precise diagnosis to enable effective intervention. Existing diagnostic techniques encounter difficulties in consistently differentiating AD from the typical aging process owing to similarities in imaging characteristics. A framework based on deep learning, integrated with critical metrics obtained from MRI scans such as cortical thickness, hippocampal volume, and white matter integrity, is proposed in the present research to address this issue and enhance diagnostic precision. The methodology involves a convolutional neural network architecture specifically structured to analyze such metrics, allowing for excellent classification between AD instances and non-AD ones. The model architecture includes specific layers that identify more subtle and unique changes in brain structure related to AD. Detailed overall assessments of the method on a broad level report strong performance, with a test set accuracy of 91.4%, in addition to precision and recall for AD and non-AD classes of 88.2% and 88.9%, respectively. An area under the curve of 0.92 on the test data indicates strong discrimination properties. The present method alleviates current limitations in diagnosis and provides a scalable solution that is reliable for application in clinical settings for the detection of neurodegenerative diseases. The results suggest that this model holds great promise for improving the accuracy of AD diagnosis in healthcare institutions.