Early Diagnosis and Progression of Alzheimer's Disease Based on Long Short-Term Memory Model

Yilin Jiang, Zengyi Yu, Xuanyu Yin, Haidong Guo

Abstract: With the growing aging population, Alzheimer's disease (AD) is becoming more prevalent. However, lack of awareness hinders early detection, leading to 67% of patients being diagnosed in the intermediate stage, missing the optimal intervention period. Therefore, research on early diagnosis of Alzheimer's disease holds practical value. Building on prior work, this study further subdivided the symptoms of Alzheimer's disease, segmenting mild cognitive impairment (MCI) into three categories, such as subjective memory complaints (SMC). This study constructed a clustering model based on neural networks and KNN, and used MAT LAB for model data training, to intelligently diagnose MCI, and ultimately achieving a 90% test accuracy rate. Additionally, considering the dynamic nature of AD progression, this study proposed an improved LSTM neural network method utilizing attention mechanisms to predict the development of patients' conditions across multiple clinical examinations, achieving a test accuracy rate of 88%.