Search

We present practical solutions to applying Gaussian-process methods to calculate spatial statistics for grid cells in large environments. Gaussian processes are a data efficient approach to inferring neural tuning as a function of time, space, and other variables. We discuss how to design appropriate kernels for grid cells, and show that a variational Bayesian approach to log-Gaussian Poisson models can be calculated quickly. This class of models has closed-form expressions for the evidence lower-bound, and can be estimated rapidly for certain parameterizations of the posterior covariance. We provide an implementation that operates in a low-rank spatial frequency subspace for further acceleration. We demonstrate these methods on a recording from grid cells in a large environment.

Spatial working memory and image recognition tests are commonly used to facilitate the diagnosis of hippocampal-related neurological disorders such as Alzheimer’s disease due to their relatively high specificity and sensitivity to damage to the medial temporal lobes compared to standard commonly used clinical tests. Pathological changes in Alzheimer’s disease start years before the formal diagnosis is made, partially due to testing too late. To address this challenge, we developed a novel digital platform, hAge (‘healthy Age’), which integrates double spatial alternation, image recognition and visuospatial tasks for frequent remote unsupervised assessment of spatial and non-spatial working memory. 191 healthy adults (67% females, 18-81 years old) participated in the study. In line with findings using standard laboratory tests, we showed that performance on the spatial alternation task negatively correlated with inter-trial periods and performance levels on image recognition and visuospatial tasks can be controlled by varying image similarity. Importantly, we demonstrated that frequent engagement with the double spatial alternation task leads to a strong practice effect, previously identified as a potential measure of cognitive decline in MCI patients. Finally, we discuss how lifestyle and motivation confounds may present a serious challenge for cognitive assessment in real-world uncontrolled environments.

Although rodents are arguably the easiest animals to use for studying brain function, relying on them as model species for translational research comes with its own set of limitations. Here, we propose sheep as a practical large animal species to use for in vivo brain function studies performed in naturalistic settings. We conducted proof-of-principle deep brain electrophysiological recording experiments using unrestrained sheep during behavioral testing. Recordings were made from cortex and hippocampus, both while sheep performed goal-directed behaviors (two-choice discrimination tasks) and across states of vigilance, including sleep. Hippocampal and cortical oscillatory rhythms were consistent with those seen in rodents and non-human primates, and included cortical alpha oscillations and hippocampal sharp wave ripple oscillations (∼150 Hz) during immobility and hippocampal theta oscillations (5-6 Hz) during locomotion. Recordings were conducted over a period of many months during which time the animals participated willingly in the experiments. Over 3,000 putative neurons were identified, including examples whose activity was modulated by task, speed of locomotion, spatial position, reward and vigilance states, and one whose firing rate was potentially modulated by the sight of the investigator. Together, these experiments demonstrate that sheep are excellent experimental animals to use for longitudinal studies requiring a large-brained mammal and/or large-scale recordings across distributed neuronal networks. Sheep could be used safely for studying not only neural encoding of decision-making and spatial-mapping in naturalistic environments outside the confines of the traditional laboratory but also the neural basis of both intra- and inter-species social interactions.

While rodents are arguably the easiest animals to use for studying brain function, relying on them as model species for translational research comes with its own sets of limitations. Here, we propose sheep as a practical large animal species for in vivo brain function studies performed in naturalistic settings. To demonstrate their experimental usefulness, we performed proof-of-principle deep brain electrophysiological recording experiments from unrestrained sheep. Recordings were made from cortex and hippocampus both whilst sheep performed goal-directed behaviours (two-choice discrimination tasks), and across states of vigilance that included natural sleep. Hippocampal and cortical oscillatory rhythms were consistent with those seen in rodents and non-human primates, and included cortical alpha oscillations during immobility, hippocampal theta oscillations (5-6Hz) during locomotion and hippocampal sharp wave ripple oscillations (∼150 Hz) during immobility. Moreover, we found clear examples of neurons whose activity was modulated by task, speed of locomotion, spatial position, reward and vigilance states. Recordings were conducted over a period of many months. Due to the exceptional stability of individual electrodes we were able to record from some neurons repeatedly across the span of a month. Together these experiments demonstrate that sheep are an excellent experimental animal model to use in longitudinal electrophysiological and imaging studies, particularly those requiring a large brained mammal, large scale recordings across distributed neuronal networks, experimentation outside the confounds of the traditional laboratory, or all the above concomitantly.