Functional MRI in awake unrestrained dogs

Abstract

Because of dogs' prolonged evolution with humans, many of the canine cognitive skills are thought to represent a selection of traits that make dogs particularly sensitive to human cues. But how does the dog mind actually work? To develop a methodology to answer this question, we trained two dogs to remain motionless for the duration required to collect quality fMRI images by using positive reinforcement without sedation or physical restraints. The task was designed to determine which brain circuits differentially respond to human hand signals denoting the presence or absence of a food reward. Head motion within trials was less than 1 mm. Consistent with prior reinforcement learning literature, we observed caudate activation in both dogs in response to the hand signal denoting reward versus no-reward.

Figure 1. Training and task for dogs in the MRI scanner.

(A) Callie in the training apparatus, which consisted of a replica of the head coil inside a tube of the approximate diameter of the MRI bore. Consistent positioning of the head was achieved by training the dog to place her head in a chin rest molded to the lower jaw from mid-snout to behind the mandible. The chin rest was affixed to a wood shelf that spanned the head coil but allowed enough space for the paws underneath. No restraints were used. The training procedure gradually shaped the desired behavior of placing the head in the rest and not moving through positive reinforcement only. Dogs were free to exit the apparatus at any time. (B) McKenzie inside the real head coil in the MRI. Her handler is giving a hand signal that denotes upcoming “reward.” We used a simple instrumental conditioning task in which the required behavior was to place the head on the chin rest and not move. After a variable interval of approximately 5 s, a hand signal was given that indicated whether a reward would be delivered. The dog had to continue holding still during this period to get the reward. The left hand up indicated a hot dog reward, while both hands pointing toward each other horizontally indicated no-reward. The hand signals were maintained for approximately 10 s. Reward-trials ended by the handler reaching in with the food to the dog. Person in the photograph has given written informed consent for publication.

Figure 2. Motion during canine fMRI.

(A) Timeseries of translations required to correct for motion during the scan sessions. Volume 32 was the target for Callie, and volume 1 was the target for McKenzie. The plots therefore represent the total movement from the target volume. The spikes and breaks occurred when the dog moved its head out of the field of view, which typically happened following a reward. The volumes with artifacts were excluded from further analysis, leaving 62% of the volumes for Callie and 58% for McKenzie. Although the dogs did not place their heads back in exactly the same position, once they did, very little motion was observed. McKenzie exhibited a slow anterior-posterior drift during the second run, but this was sufficiently slow as to not cause movement artifacts during trials. (B) Average motion during a trial, separated by reward and no-reward conditions and after exclusion of volumes with artifacts. Scan volumes are 1610 ms apart. Notably, within-trial motion was less than 1 mm in all directions for both dogs, and no difference between the reward and no-reward conditions was observed.

Figure 3. The caudate is significantly more active to the “reward” hand signal compared to the “no-reward” hand signal.

The same region of activation was observed in both dogs and is identified as the right caudate as indicated on the corresponding slice of each dog's structural image (CD). The structural image has been uniformly scaled to match the size of the brain of the functional images. The underlay of the functional map is the mean of the non-excluded functional images. McKenzie was rotated slightly out of plane, but this was a consistent position in both functional and structural scans. The significance of the peak voxel in this cluster was p<0.01 in Callie and p<0.001 in McKenzie (colorbar indicates t-values and maps are thresholded at p<0.05 to show full spatial extent). The time series of activation was extracted for the cluster (9 voxels in Callie, and 18 voxels in McKenzie after restricting spatial extent with p<0.01), and after adjusting for the other effects in the design matrix (including motion), the average trial response is seen to match a typical hemodynamic response function, which is significantly greater for the “reward” signal than the “no-reward” signal (error bars are +/− 1 s.e.) Bottom: statistical map of the combined model with both dogs, co-registered and overlaid on Callie's structural scan. Activation of the caudate cluster (CD) was significant at p<0.05 after correcting for FDR over the search volume of the ventral brain from olfactory bulb to internal capsule (p<0.01 height and cluster extent>6). Averaged over both dogs, the timecourse of activation in the caudate showed a distinct response to the reward hand signal which differentiates from the no-reward signal (lower right). Scan volumes are 1610 ms apart, indicating a peak in the response 3–5 s after the onset of the reward hand signal.