Functional MRI in Awake Unrestrained Dogs

Citation: Berns GS, Brooks AM, Spivak M ( Functional MRI in Awake Unrestrained Dogs Gregory S. Berns 0 Andrew M. Brooks 0 Mark Spivak 0 Stephan C. F. Neuhauss, University Z urich, Switzerland 0 1 Center for Neuropolicy, Emory University , Atlanta , Georgia , United States of America, 2 Comprehensive Pet Therapy , Atlanta, Georgia , United States of America 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. - As the oldest domesticated species, with estimates ranging from 9,00030,000 years BCE, the minds of dogs inevitably have been shaped by millennia of contact with humans [1,2]. As a result of this physical and social evolution, dogs, more than any other species, have acquired the ability to understand and communicate with humans. A resurgence of research in canine cognition has revealed the range (and variability) of skills such as following pointing and gaze cues [3,4,5], fast mapping of novel words [6], and the conjecture that dogs have emotions [7]. Although the growing list of canine cognitive skills is impressive, how does the dog mind actually work? We are left to infer canine brain function from behavior and ultimately guess at the inner workings of the dog brain. However, the widespread use of functional magnetic resonance imaging (fMRI) to study brain function in both humans and non-human primates has paved the way for potentially understanding how the dog brain works. Here, we report the development of behavioral and technical methods to acquire fMRI data in fully awake, unrestrained dogs. The main challenge of fMRI in dogs comes from subject motion. Historically, the usual approach has been to either anesthetize the animal [8,9] or, as in rats and monkeys, immobilize them [10,11,12,13,14,15]. Clearly, if we wish to understand canine cognition, anesthesia is not an option. Immobilization is technically possible, although ethically objectionable for a dog, and, as we show, unnecessary to acquire useful fMRI data. Instead, because dogs so readily follow human commands, they can be trained to go into an MRI scanner and hold their head still enough for fMRI studies without any restraint. Moreover, they will do this happily with nothing more than positive reinforcement. Because of their 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 [16]. For this reason, we selected a simple discrimination task with two human hand signals for initial study with canine fMRI. Although there is growing evidence that dogs do not need to be conditioned to learn human hand signals, for this first experiment we chose to associate the hand signals with primary rewards to provide a linkage with comparable imaging experiments in both humans and monkeys and to maximize the chance of observing a significant brain response. Importantly, the reward-prediction error hypothesis of the dopamine system provides a concrete prediction of activity in the ventral caudate of the dog. The task was designed to determine which brain circuits differentially respond to hand signals denoting the presence or absence of a food reward. Based on the reinforcement learning literature, we hypothesized that the transfer of reward association to a hand signal would manifest in the ventral striatum [17,18,19,20,21]. Subjects were two spayed, female, domesticated dogs. Callie was a 2 year-old feist of indeterminate pedigree, who had been adopted from a local shelter at the age of 9 months and weighed 12 kg. Apart from basic obedience, she had no specialized training. McKenzie was a 3 year-old border collie and was already well-trained in agility competition and weighed 16 kg. Training and handling for the following procedures were performed by each dogs owner under the supervision of a professional trainer. This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All procedures were approved by the Institutional Animal Care and Use Committee of Emory University (Protocol Number: DAR-2001274-120814). Three fMRI scanning sessions were performed over a period of 6 weeks. Callie participated in all sessions, while McKenzie participated in the last two. The goal of the first session was to familiarize the Callie with the scanner environment and determine the feasibility of acquiring both structural and fMRI data. The goal of the second session was to optimize the scan parameters and to obtain enough fMRI data to evaluate its quality for movementrelated artifacts. It was observed that the onset of each imaging sequence tended to startle the dogs, causing them to move or exit the scanner. This was effectively mitigated in the final session by playing recordings of the scanner noise through the intercom while the dog got settled into the chin rest. The preceding protocol encouraged habituation to the scanner noise and eliminated startle reactions. In the third and final session, the onset was not startling and the dogs didnt move severely when the actual sequence started. This approach allowed us to obtain functional runs long enough for fMRI analysis as well as a high quality structural image. For the final scanning session, we used a simple instrumental conditioning task in which the required behavior was to place the head on the chin rest and not move (Fig. 1). After a variable interval of approximately 5 s, a hand signal was given that indicated the presence or absence of a food reward that would be received. The left hand up indicated a hot dog reward, while both hands pointing toward each other horizontally indicated no reward. The hand signals were chosen to be easily distinguishable and were maintained for approximately 10 s. The dog had to continue holding still during this period. Dogs had been amply trained on these hand signals in the simulator prior to the final scan session. Because the dogs had been trained to go into the head coil in a sphinx position (Fig. 1), the handler gave the hand signals from the head end of the scanner, facing the dog. Trial types were approximately random and alternating (but not predictably) such th (...truncated)