New method makes AI brain-disorder screening more transparent and computationally efficient

Researchers at Sungkyunkwan University have created a new AI approach called 4DfCF to analyze brain scans and improve diagnosis of neurological disorders. This method examines four-dimensional fMRI data, which records brain activity over time, to uncover complex patterns that relate both where and how brain activity changes. Unlike traditional AI models that only provide a simple yes or no answer, 4DfCF generates visual maps showing which brain regions contributed most to its predictions.

This transparency allows clinicians to better understand the model's reasoning and incorporate it into their decision-making process, rather than simply accepting the AI's conclusion as definitive. The 4DfCF model achieved an impressive 96.28% accuracy across three benchmark datasets containing fMRI scans of individuals with ADHD, Alzheimer's disease, and autism spectrum disorder.

Importantly, the researchers demonstrated that the model could be adapted more easily to other disorders by leveraging transfer learning, where a model trained on one dataset can quickly adapt to another with relatively little additional training. Additionally, the 4DfCF architecture is designed to be computationally efficient, requiring fewer resources than many other AI models while still maintaining high performance.

While these results show promising advancements in AI-assisted brain disorder diagnosis, further validation in real-world clinical settings will be necessary before the technology can be widely adopted in hospitals.