Alzheimer's Disease (AD) is a neurological disorder defined by a gradual deterioration in cognitive abilities and memory impairment. It significantly impacts individuals’ quality of life and imposes a considerable burden on healthcare systems. Prompt detection, especially during the Mild Cognitive Impairment (MCI) stage, is necessary for earlier interventions and enhancing disease classification. This phase includes Early MCI (EMCI), representing the initial stages of cognitive decline, and Late MCI (LMCI), which indicates progressive deterioration. Deep learning (DL) has the capacity to extract features automatically from image data, presenting significant potential to enhance the diagnosis of AD and MCI. However, challenges such as imbalanced datasets, highdimensional features, and the difficulty of capturing subtle changes in early disease stages have limited the effectiveness of existing models. To address these challenges, this work presents a deep learning-based model using DenseNet-169 architecture integrated with two attention mechanisms—Global Context Network and Triplet Attention. The Global Context Network captures long-range dependencies and contextual information, the Triplet Attention Module extracts cross-dimensional and position-sensitive features. Therefore, the proposed model concentrates on extracting relevant, multi-scale features, which enhances the capacity of the model to classify AD and its early phases more accurately. The proposed model gives an accuracy of 98.8% when evaluated on the ADNI dataset. The conclusions show the efficacy of custom-designed DL models in enhancing the early detection of AD.