AI Assisted Diagnosis of Mental Health Conditions

Abstract

Mental health disorders among adolescents are increasingly prevalent, highlighting the need for accurate diagnoses and effective treatments. THis study utilized a neural network to analyze symptom overlap in patients with single versus comorbid mental health diagnoses, aiming to improve diagnostic accuracy. It was conducted using Kaggle's Mental Illness Dataset, which includes demographic data, symptoms, and diagnoses from 3,753 patients. The neural network's performance was evaluated, revealing a difference in effectiveness between the original dataset and the single diagnosis subset. Results indicated that the model performed better with the original dataset, which presented distinct patterns of symptom presentation, facilitating more accurate predictions.This distinction emphasizes the importance of considering the complexity of mental health conditions, as symptoms can vary significantly between patients with single and comorbid diagnoses. However, limitations of subsetting patients include oversimplification of symptoms and reduced sample sizes. Future research should focus on larger datasets and changing factors like symptom severity to deeper understand and improve treatment approaches in mental health care.