Through a Dog's Eyes: fMRI Decoding of Naturalistic Videos from the Dog Cortex

Area of Science:

• Neuroscience
• Cognitive Science
• Animal Behavior

Background:

• Machine learning and fMRI have advanced visual perception decoding in humans and primates.
• Limited research exists on visual representation in non-primate brains, particularly carnivores.

Purpose of the Study:

• To investigate visual stimulus decoding in domestic dogs using machine learning and fMRI.
• To compare visual processing between dogs and humans using naturalistic video stimuli.

Main Methods:

• Awake functional magnetic resonance imaging (fMRI) was performed on two dogs and two humans watching dog-appropriate videos.
• A neural network (Ivis) was trained to classify video content from brain activity.
• Object-based and action-based classifiers were evaluated.

Main Results:

• Both object-based and action-based classifiers performed well in humans.
• Only action-based classifiers successfully decoded video content from dog brains.
• This represents the first machine learning application to decode naturalistic videos from a carnivore's brain.

Conclusions:

• Dog visual perception differs significantly from human perception, with a greater emphasis on actions over objects.
• The study provides novel insights into the neural representation of visual stimuli in canines.