Identifying Naturalistic Movies from Human Brain Activity with High-Density Diffuse Optical Tomography

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Engineering

Background:

• Advanced neuroimaging like fMRI decodes complex human experiences but is limited by portability and invasiveness.
• Current wearable brain sensors (EEG, fNIRS) offer limited decoding capabilities, typically restricted to training stimuli.
• There is a need for non-invasive, portable neuroimaging with high-resolution decoding potential for naturalistic settings.

Approach:

• Developed and evaluated model-based decoding using high-density diffuse optical tomography (HD-DOT), an advanced fNIRS technique.
• Utilized a motion energy model to represent visual content for decoding.
• Tested decoding accuracy on novel movie clips presented outside the training set, assessing single-trial performance.

Key Points:

• HD-DOT achieved decoding accuracy significantly above chance for identifying novel visual stimuli.
• Decoding performance remained robust across variations in testing time windows, training/testing sessions, and hemodynamic contrasts.
• Results demonstrate the potential of HD-DOT as a portable fMRI surrogate for detailed brain decoding.

Conclusions:

• HD-DOT facilitates high-resolution, model-based decoding of visual experiences.
• This technology overcomes limitations of current wearable neuroimaging, enabling decoding in naturalistic environments.
• HD-DOT represents a significant advancement towards practical, real-world brain-computer interfaces.