Language task-based fMRI analysis using machine learning and deep learning.

Area of Science:

• Neuroimaging
• Machine Learning
• Computational Neuroscience

Background:

• Task-based functional Magnetic Resonance Imaging (fMRI) is crucial for identifying language-dominant brain regions, particularly for neurosurgical planning near eloquent areas.
• Unstructured fMRI paradigms, like naturalistic fMRI, are gaining interest for language mapping but require advanced analytical techniques due to difficulties in defining task regressors.
• Machine learning (ML) and deep learning (DL) offer potential solutions for analyzing complex fMRI data from these paradigms.

Purpose of the Study:

• To investigate the efficacy of various ML and DL algorithms in identifying brain regions associated with language using task-based fMRI data.
• To evaluate the performance of ML and DL models in classifying voxel-wise fMRI time series for language mapping.

Main Methods:

• Collected fMRI data from 26 individuals across seven task-based language fMRI paradigms.
• Trained ML and DL models to classify voxel-wise fMRI time series.
• Evaluated general machine learning and interval-based methods for their performance in language area identification.

Main Results:

• General machine learning and interval-based methods demonstrated significant promise in identifying language areas through fMRI time series classification.
• The general machine learning method achieved a mean whole-brain AUC of [value], mean Dice coefficient of [value], and mean Euclidean distance of [value] mm.
• The interval-based method achieved a mean whole-brain AUC of [value], mean Dice coefficient of [value], and mean Euclidean distance of [value] mm.

Conclusions:

• This study confirms the utility of diverse ML and DL methods for classifying task-based language fMRI time series.
• These advanced analytical approaches hold significant potential for identifying language activation, especially within unstructured fMRI paradigms.