APPLICATION OF DEEP LEARNING MODELS FOR EARLY DETECTION OF AUTISM SIGNS BASED ON EYE MOVEMENT ANALYSIS

Abstract

The presented scientific study reports the results of the creation and testing of two deep learning models, long short-term memory with a convolutional neural network LSTM+CNN and long short-term memory with an autoencoder LSTM+AE, for autism spectrum disorder diagnosis. The study is directed towards the use of eye tracking technology for recording participants eye movement data in interaction with animated objects. The data were saved in the.npy format of NumPy arrays for convenient subsequent analysis. The algorithms were evaluated in terms of their accuracy, generalization capability, and training time, as confirmed by experts. The primary aim of this study is to enhance the accuracy and efficiency of autism diagnosis. The long short-term memory convolutional neural network and autoencoder-long short-term memory architectures have shown great promise as tools for achieving this goal, with the autoencoder model being especially distinguished by its ability to identify inherent relationships between datasets. Additionally, the paper discusses potential clinical uses of the algorithms and future directions for research.