Resting connectivity predicts task activation in pre-surgical populations

Area of Science:

• Neuroscience
• Medical Imaging
• Cognitive Science

Background:

• Neural processing varies significantly in patients with injury or disease compared to healthy individuals.
• Understanding this variability is crucial for accurate pre-surgical planning, especially for identifying eloquent brain areas.

Purpose of the Study:

• To determine if precise language maps can be inferred from resting-state brain activity in patient populations.
• To investigate the predictive power of resting-state connectivity for individual differences in neural responses, including pathological variability.

Main Methods:

• A predictive model was trained using pairs of resting-state and task-evoked functional magnetic resonance imaging (fMRI) data.
• The model was tested on unseen patients and healthy controls, predicting task activation solely from resting-state data.
• A category fluency task was employed to acquire task-evoked fMRI data.

Main Results:

• Models successfully learned individual variations in language processing from resting-state connectivity features in both patient and control groups.
• Despite greater variability in patients' actual language maps, resting connectivity accurately predicted task activations.
• A model trained on healthy controls alone could predict task activations in patients, demonstrating robustness.

Conclusions:

• Resting-state functional connectivity robustly predicts individual differences in neural responses, even amidst pathological variability.
• This approach offers a reliable method for inferring language maps from non-task-based scans, enhancing pre-surgical planning.
• The findings highlight the potential of resting-state fMRI for understanding neural variability in clinical populations.