Early Screening and Prediction of Alzheimer's Disease Based on Long-Term and Short-Term Memory Neural Networks

Abstract

With the increase of the proportion of aging Chinese population, the incidence rate has increased in recent years. Because the disease has a latent onset, the course of the disease is slow and irreversible, and early screening and diagnosis of Alzheimer's disease is particularly important. With the development of computer computing power, the exploration of the field of deep learning has gradually unfolded. Because the long short-term memory neural network has a memory unit, it can capture the long-term dependence of time series data and record historical information, which has obvious advantages in disease prediction. LSTM neural networks have obvious advantages in memory, processing lagging data, preventing gradient disappearance, and learning ability, which makes them very suitable for predicting time series data and complex problems such as Alzheimer's disease. This can provide theoretical and technical support for related research, and help improve the accuracy and wide acceptance of predictions. In this paper, the NMR image information data is used to predict the population of patients without Alzheimer's disease (NC) or mild serious disorder (MCI) through the long-short-term memory neural network model, and the probability of disease in the next 3–5 years is obtained, which is also an effective attempt of long-short-term memory neural network in medical prediction.