FROM HUMANS TO DOGS: MODELLING THE BOLD SIGNAL IN COMPARATIVE NEUROIMAGING

Functional MRI (fMRI) with trained pet dogs offers the unique opportunity to study the neural correlates of socio-cognitive processes in awake and unrestrained non-human animals. fMRI measures activity-related changes in blood oxygen concentration that give rise to the blood-oxygen-level-dependent (BOLD) signal. Compared to neural activation, the inflow of oxygenated blood, and thus the BOLD signal time course, is slightly delayed. During neuroimaging analysis, this delay is modelled by a haemodynamic response function (HRF). An inaccurate HRF can lead to both false-positive and false-negative results. In our previous work, the dog BOLD signal peaked earlier than expected with a standard human HRF, and using a tailored dog HRF increased detection power. However, these findings have not yet been replicated and are limited to a single brain region, although BOLD time courses vary across regions. Here, we tested the replicability of the dog HRF across sensory systems and brain regions. We extracted BOLD time courses from visual (n = 41), auditory (n = 18), and tactile (n = 28) stimulation data in N = 45 dogs and compared the performance of the dog and human HRF models (Figure 1A). First findings suggest that the BOLD response peaks 2-3 s earlier in dogs across all tested regions and that a tailored dog HRF increases sensitivity (Figure B-C). Results contribute to improving analysis methods for canine neuroimaging and to further establishing domestic dogs as a model for comparative neuroscience.

Figure 1. Study overview. V/S/A1, primary visual/somatosensory/auditory cortex; SEM, standard error of the mean.