Detecting and diagnosing Alzheimer’s disease at an early stage is challenging due to subtle early stage symptoms. In this paper we propose a hybrid deep learning approach where we integrate 3D MRI scans along with their clinical features including Mini Mental State Examination (MMSE) scores, Clinical Dementia Rating (CDR), patient’s educational level, their age and brain volumetric measures to significantly improve diagnostic accuracy. Our architecture consists of custom DenseNet based attention units and parallel clinical data processing pathways. Our model achieved a cross-validation accuracy of 95.18% when tested on OASIS-2 dataset, showing approximately 4% improvement over VGG16(91%) and Random Forest(86%). Hence this approach significantly reduces the false negatives and false positives during early detection of AD thereby demonstrating its potential for clinical deployment.