Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session C: Friday, August 15, 2:00 – 5:00 pm, de Brug & E‑Hall

Improving Prediction of Cognitive Abilities through Integrated Resting-State and Task fMRI

Yujing Jiang¹, Victor KS Chan, Jing Jun Wong, Nicole HL Wong, Bolton KH Chau; ¹University of Melbourne

Presenter: Yujing Jiang

Using resting-state fMRI (rs-fMRI) and task-fMRI are two common approaches of predicting cognitive abilities, yet conventional methods offer limited accuracy. We propose a transformer-based framework that unifies rs-fMRI and t-fMRI, drawing inspiration from Large Language Models (LLMs) known for integrating diverse sequential inputs. Instead of treating these modalities separately, our approach encodes both as continuous temporal sequences and applies self-attention to learn Dynamic Activity Signatures that capture neural processes common to spontaneous and task-evoked activity. This unified latent space produces consistent predictions across rs-fMRI and t-fMRI, even in the absence of task data, while eliminating the need for multiple models. Evaluated on the Human Connectome Project (HCP) and Alzheimer’s Disease Neuroimaging Initiative (ADNI) datasets, our framework demonstrated high predictive accuracy and robust generalization, proving the effectiveness of a unified model that seamlessly integrates rs-fMRI and t-fMRI.

Topic Area: Brain Networks & Neural Dynamics

Extended Abstract: Full Text PDF