Prediction of enhanced bipolar disorder in autistic children using denoising convolutional neural network

AbstractBipolar disorder (BD) causes depression, anxiety, irritability, hyperactivity, and other behavioral changes in autistic children. An accurate BD examination helps the doctors to prescribe the correct treatment and dosage level. Patients with BD have previously undergone the Aberrant Behavior Checklist (ABC) in clinics for examination and which takes only a short amount of time. Continuous monitoring of autistic children is a major problem for physicians when assessing autistic children ’s examinations. In this paper, autistic child BD assessment is performed through the thermal radiometric pixel of facial regions such as the face, eye, mouth, cheek, and forehead thermal images. Thermal images are obtained from continuous monitoring of thermal cameras such as the indoor and outdo or environments. The facial thermal regions have a crumbling effect (CE) and hypomelanotic disorder (HD) in the skin leads to noise in thermal pixel values, resulting in inaccurate measurement. Because of these CE and HD effects, thermal images have both scaled and unscaled noises, as well as white noise. Thermal facial region images are applied with Non-Decimated Wavelet Transform (NDWT) and Denoising Convolutional Neural Network (DnCNN) filters. In comparison to ABC, the suggested NDWT and DnCNN gives an accurate measurement for assessment of BD, emotions, and repetitive behaviors in autisti c children with about 95% of accuracy in identification and examination.
Source: Network Modeling Analysis in Health Informatics and Bioinformatics - Category: Bioinformatics Source Type: research