基于增强型灰狼优化器的Facebook人工智能相似性搜索用于阿尔茨海默病诊断

Early diagnosis of Alzheimer's disease (AD) and its prodromal stage, mild cognitive impairment (MCI), requires efficient computational frameworks capable of handling high-dimensional and large-scale 18F-FDG-PET data. To address challenges in feature redundancy and computational efficiency, we propose a unified framework integrating an enhanced Grey Wolf Optimizer (CDL-GWO) with approximate nearest neighbor search based on Facebook AI Similarity Search (FAISS). The proposed CDL-GWO incorporates Logistic chaotic initialization to improve search space coverage, dual random projection to enhance population diversity via inter-solution differences, and Lévy flight to escape local optima, achieving a better balance between exploration and exploitation. Experimental results on the CEC2017 benchmark demonstrate that CDL-GWO outperforms several classical and state-of-the-art metaheuristic algorithms. Combined with FAISS, the framework enables efficient similarity search for large datasets. Validation on 890 subjects from the ADNI 18F-FDG-PET dataset across six AD-related classification tasks shows that the proposed method achieves competitive diagnostic performance (up to 0.943 accuracy) while reducing computational cost by approximately 70% compared to conventional KNN. These results indicate that the proposed framework provides an effective and scalable solution for early AD diagnosis and MCI classification.