Advances in AI-based diagnosis of Alzheimer's disease using MRI: a comprehensive survey

Artificial intelligence (AI), especially Deep Learning (DL), has been shown significant in accelerating the detection and diagnosis of neurological disorders via medical imaging. This study is mainly focused on Alzheimer's disease (AD), which reveals distinctive structural modifications observable by Magnetic Resonance Imaging (MRI). Although several studies employing convolutional neural networks (CNNs) and other artificial intelligence models indicate promising diagnostic accuracy, many issues related to methodology exist. This research offers a comprehensive assessment of recent studies (2000-2025) to synthesize the key limitations limiting the clinical application of AI for AD detection using MRI. The study identify the main challenges, namely: (1) restricted access to extensive, curated, and diverse multimodal datasets; (2) elevated model complexity with associated risks of overfitting on small cohorts; (3) insufficient interpretability and clinical validation of AI decisions; (4) computational inefficiency and excessive energy consumption; and (5) challenges in generalizing models across heterogeneous cohorts and imaging guidelines. Our study indicates that modern research frequently emphasizes marginal improvements in accuracy rather than solving these essential translational obstacles. The authors conclude by outlining essential research progressions, highlighting the necessity for federated learning for dealing with data scarcity, the advancement of explainable AI (XAI) frameworks, and the creation of standardized benchmarking protocols to flexible, clinically-adoptable AI methods for early AD detection.