Prediction of image interpretation cognitive ability under different mental workloads: a task-state fMRI study.

Area of Science:

• Neuroscience
• Cognitive Science
• Medical Imaging

Background:

• Visual imaging expertise is crucial across various fields.
• Functional magnetic resonance imaging (fMRI) combined with machine learning (ML) can predict cognitive abilities.
• This approach offers a potential method for selecting skilled image interpreters.

Purpose of the Study:

• To investigate the neural basis of image interpretation ability under varying cognitive workloads.
• To assess the efficacy of ML algorithms in predicting image interpretation skills based on fMRI data.

Main Methods:

• Recorded behavioral and neural data (fMRI) from 64 participants during image interpretation tasks.
• Categorized participants into high- and low-ability groups based on performance.
• Applied general linear model (GLM) analysis and a radial basis function Support Vector Machine (SVM) algorithm for prediction and feature importance analysis.

Main Results:

• The high-ability group exhibited significantly higher brain activation in regions including the middle frontal gyrus (MFG), fusiform gyrus, and inferior occipital gyrus compared to the low-ability group.
• The SVM algorithm accurately predicted image interpretation ability (Pearson correlation coefficient = 0.54, R² = 0.31, MSE = 0.039, RMSE = 0.002).
• Activation patterns in the fusiform gyrus and MFG were identified as key predictors of image interpretation ability.

Conclusions:

• This study elucidates the neural underpinnings of image interpretation ability influenced by cognitive load.
• Machine learning effectively utilizes fMRI-derived neural activation features for predicting visual interpretation expertise.