Functional brain-to-brain transformation without shared stimuli.

Area of Science:

• Neuroscience
• Medical Imaging
• Data Science

Background:

• Combining functional Magnetic Resonance Imaging (fMRI) data across subjects and datasets is essential for neuroscience.
• Current functional transformation methods rely on shared stimuli, which are often unavailable.
• Anatomical mapping alone is insufficient for accurate brain-to-brain transformations.

Purpose of the Study:

• To develop a novel approach for computing functional brain-to-brain transformations without shared stimuli.
• To enable the integration of diverse fMRI datasets collected with different stimuli and scanner resolutions.
• To enhance Image-to-fMRI encoding by leveraging cross-dataset functional alignment.

Main Methods:

• Proposed a novel framework combining brain-to-brain transformation with Image-to-fMRI encoders.
• Enabled learning functional transformations on visual stimuli subjects were not exposed to.
• Demonstrated applicability by improving Image-to-fMRI encoding using a high-resolution dataset to enhance low-resolution data.

Main Results:

• Successfully computed functional brain-to-brain transformations without any shared stimuli.
• Facilitated the merging of fMRI datasets with distinct stimuli and varying resolutions (3T and 7T).
• Showcased improved Image-to-fMRI encoding performance on older, low-resolution fMRI datasets.

Conclusions:

• Presented a general framework for functional alignment across individuals and datasets without shared stimuli.
• Opened new possibilities for integrating and leveraging diverse fMRI collections.
• This method overcomes limitations of existing functional transformation techniques.