JMIR Publications

ABSTRACT

Autism Spectrum Disorder (ASD) is a prevalent neurodevelopmental condition that can be quite difficult to diagnose, due to a lack of objective diagnostic methods in the currently used behavioral assessments. In this study, we assessed a variety of deep learning approaches for classification of ASD, utilizing data collected via Inertial Measurement Unit (IMU) hand tracking during goal-directed arm movements. IMU data was recorded from forty-one school aged children both with and without an ASD diagnosis, to track their arm movements during a reach-to-clean up task. The IMU data was then preprocessed using a moving average and z-score normalization to prepare it for the models. We evaluated the effectiveness of different deep learning models using the preprocessed data and a K-fold validation approach. The best result was achieved with a Convolutional Autoencoder (CNN-AE) combined with Long Short-Term Memory (LSTM) layers, reaching an accuracy of 90.21% and an F1-score of 90.02%. Once the CNN-AE+LSTM was determined to be the most effective model for this datatype, it was retrained and evaluated with a patient-separated dataset to assess the generalization capability of the model, achieving an accuracy of 91.87% and an F1-score of 93.66%. Our deep learning approach demonstrates that small scale models can still achieve high accuracy classifying medical data, and shows that our model holds potential for facilitating ASD diagnosis in clinical settings.