Afrah Imam

How can neural networks effectively model the relationship between fibromyalgia and autism based on behavioral questionnaires?

Started Dec. 13, 2024

Abstract or project description

Fibromyalgia is a chronic pain condition which, like other similar chronic pain conditions, has been explored to discover many connections to autism. Some datasets have been published on such a topic but limited modeling exists. This project aims to utilize neural networks using PyTorch to explore such a condition and how it relates to autism. The dataset for this study consists of behavioral questionnaire data assessing a variety of domains, including sensory issues, autistic traits, and chronic pain experiences. In addition to these measures, formal fibromyalgia diagnoses were recorded. A deep learning approach using neural networks was employed to analyze the data, examining how different factors contribute to the understanding of fibromyalgia in the context of autism. A series of neural networks with varying architectures were trained using PyTorch to predict SAT scores related to autistic traits. The models achieved high training accuracy, with the best-performing architecture (2 layers, 250 units) reaching 98.1% accuracy in predicting scores of the Autism Questionnaire, suggesting a strong ability to capture the relationship between fibromyalgia-related symptoms and autism-related characteristics. These results show that machine learning can help us better understand how fibromyalgia and autism are connected. This could be useful for future research and may help improve how we assess and treat people who have both conditions.