Autism Detection Using Pre-Trained Models Using VGG And CNN

Abstract

GPU technology efficiently trains and tests deep learning architectures through parallelizable mathematical procedures [1]. GPUs consist of multiple cores that accelerate complex computational operations. The number of processing units (cores) that can operate independently is crucial in determining the level of parallelization possible. There is a substantial difference between GPUs with thousands of cores and CPUs with only four or eight cores. More cores available means an increase in the amount of parallelization possible [2]. When GPUs have a lot of cores, CPU cores operate at a higher frequency. For mathematical procedures in Neural Networks, GPUs are vital [4]. Neural Networks have proven to be very successful in solving several real-life problems.

Ensuring high accuracy while working with diseases, such as Autism Spectrum Disorder (ASD), is crucial. ASD brings developmental disability to the brain and is also associated with genetic conditions [6]. Some causes of ASD are known, while others are still unknown. Patients with ASD exhibit different behavior, communication, interaction, and learning styles compared to ordinary people[5]. While some patients can live and work like ordinary people without any support, many patients require assistance from others to live their life. Some patients may have advanced conversation skills, while others may be nonverbal. Usually, ASD starts to develop before the age of three, but some children may show symptoms early in the 12-month period[4]. In the first 12 or 24 months, they gain knowledge and skills, but later stop learning, making it difficult to communicate with peers and adults, make new friends, and understand complex concepts [8]. People with ASD are also more likely to experience serious issues such as anxiety, depression, and attention-deficit/hyperactivity disorder compared to those without ASD. Early detection of ASD is crucial as it significantly decreases symptoms and has a high chance of improving the quality of life. Medical tests such as blood tests and symptom checking’s are the most common ways of detecting ASD. The symptoms are usually checked by parents or teachers, and healthcare providers[10].