New AI model could improve accuracy of mental health diagnoses, researchers say

Researchers from Keele University and the University of Nottingham have developed a new artificial intelligence model designed to improve the accuracy of classifying mental health conditions by analysing the emotional content of written text.

The model, known as Emo-MHC, combines machine learning and deep learning techniques with advanced emotion detection to analyse text from sources including clinicians' notes, social media posts and online forums. The researchers say it could help clinicians identify mental health conditions more quickly and accurately, supporting earlier intervention and more effective treatment planning.

Many existing AI tools for classifying mental health conditions rely on natural language processing alongside machine learning, often drawing on self-assessment questionnaires and standardised clinical tests. However, the research team argues these approaches can struggle to identify emotional nuances within text, increasing the risk that important contextual information is overlooked or misinterpreted.

The Emo-MHC model aims to address this by incorporating emotion detection and lexicon-based text analysis, enabling it to better recognise emotional patterns that may be relevant to diagnosis.

The project was jointly developed by Dr Shaily Kabir of the University of Nottingham and Dr Sangeeta Sangeeta, Lecturer in Data Science at Keele University, in collaboration with students Joy Paul and Zerin Jahan.

The researchers evaluated the model using publicly available datasets and found it classified mental health conditions with an accuracy rate of 92 per cent, around eight percentage points higher than the benchmark model used for comparison. The findings have been published in an Institute of Electrical and Electronics Engineers journal.

The team now plans to refine the model further to improve its performance and explore how it could be used in clinical settings to support people experiencing mental health difficulties.

Dr Sangeeta Sangeeta said: “Rates of mental health conditions are increasing, highlighting the urgent need for robust methods that enable accurate early detection. In the era of artificial intelligence and large language models, these technologies have significant potential to support individuals experiencing mental health challenges. Improved diagnostic accuracy can not only reduce the burden on the NHS but also play a critical role in saving lives.”

The researchers emphasise that the technology is intended to support, rather than replace, clinical decision-making by providing healthcare professionals with an additional tool to aid assessment. They believe improved recognition of emotional cues in written information could contribute to more accurate diagnoses and better-informed care for people experiencing mental health conditions.