Alzheimer’s Disease (AD) poses a significant challenge in healthcare, marked by a gradual decline in cognitive function and memory loss, profoundly affecting daily life, primarily among the elderly. In its early stages, AD can be diagnosed, but predictions are possible only before the illness develops symptoms. Deep learning (DL) may be a viable approach for the early diagnosis of AD. However, its computational complexity may hinder its use in resource-constrained environments. To address this, we explore an optimized lightweight InceptionNet, balancing performance and efficiency. Extensive simulation and testing on benchmark datasets with neuroimaging scans and clinical data demonstrate our method’s efficacy in accurately classifying Alzheimer’s and dementia. These results underscore deep learning’s potential, particularly InceptionNet, in improving diagnostic capabilities for neurodegenerative disorders. Integrating a lightweight InceptionNet variant offers a promising solution for real-world deployment, especially where computational resources are limited.