A somato-cognitive action network alternates with effector regions in motor cortex.

Motor cortex (M1) has been thought to form a continuous somatotopic homunculus extending down the precentral gyrus from foot to face representations1,2, despite evidence for concentric functional zones3 and maps of complex actions4. Here, using precision functional magnetic resonance imaging (fMRI) methods, we find that the classic homunculus is interrupted by regions with distinct connectivity, structure and function, alternating with effector-specific (foot, hand and mouth) areas. These inter-effector regions exhibit decreased cortical thickness and strong functional connectivity to each other, as well as to the cingulo-opercular network (CON), critical for action5 and physiological control6, arousal7, errors8 and pain9. This interdigitation of action control-linked and motor effector regions was verified in the three largest fMRI datasets. Macaque and pediatric (newborn, infant and child) precision fMRI suggested cross-species homologues and developmental precursors of the inter-effector system. A battery of motor and action fMRI tasks documented concentric effector somatotopies, separated by the CON-linked inter-effector regions. The inter-effectors lacked movement specificity and co-activated during action planning (coordination of hands and feet) and axial body movement (such as of the abdomen or eyebrows). These results, together with previous studies demonstrating stimulation-evoked complex actions4 and connectivity to internal organs10 such as the adrenal medulla, suggest that M1 is punctuated by a system for whole-body action planning, the somato-cognitive action network (SCAN). In M1, two parallel systems intertwine, forming an integrate-isolate pattern: effector-specific regions (foot, hand and mouth) for isolating fine motor control and the SCAN for integrating goals, physiology and body movement.

Abstract Value Encoding in Neural Populations But Not Single Neurons.

An important open question in neuroeconomics is how the brain represents the value of offers in a way that is both abstract (allowing for comparison) and concrete (preserving the details of the factors that influence value). Here, we examine neuronal responses to risky and safe options in five brain regions that putatively encode value in male macaques. Surprisingly, we find no detectable overlap in the neural codes used for risky and safe options, even when the options have identical subjective values (as revealed by preference) in any of the regions. Indeed, responses are weakly correlated and occupy distinct (semi-orthogonal) encoding subspaces. Notably, however, these subspaces are linked through a linear transform of their constituent encodings, a property that allows for comparison of dissimilar option types. This encoding scheme allows these regions to multiplex decision related processes: they can encode the detailed factors that influence offer value (here, risky and safety) but also directly compare dissimilar offer types. Together these results suggest a neuronal basis for the qualitatively different psychological properties of risky and safe options and highlight the power of population geometry to resolve outstanding problems in neural coding.SIGNIFICANCE STATEMENT To make economic choices, we must have some mechanism for comparing dissimilar offers. We propose that the brain uses distinct neural codes for risky and safe offers, but that these codes are linearly transformable. This encoding scheme has the dual advantage of allowing for comparison across offer types while preserving information about offer type, which in turn allows for flexibility in changing circumstances. We show that responses to risky and safe offers exhibit these predicted properties in five different reward-sensitive regions. Together, these results highlight the power of population coding principles for solving representation problems in economic choice.

Development of Biomechanical Response Curves for the Calibration of Biofidelic Measuring Devices Used in Robot Collision Testing.

Collaborative robots (cobots) can be employed in close proximity to human workers without safety fences. The operation mode Power and Force Limiting requires that cobots not exceed the biomechanical limits of ISO/TS 15066 to ensure protection against injuries caused by collisions with them. Collision tests must be performed to prove that cobots cannot exceed the biomechanical limits. Such tests are performed with a biofidelic measuring device that measures contact forces and replicates the biomechanics of the human body. Biomechanical response curves serve as a reference for the calibration of such devices. In order to be able to compare measurements and limits correctly and reliably, the limits and response curves for calibration must be obtained from the same data with the same methodology. In this article, we present a new technique for developing biomechanical response curves, which employs a statistical model we used to calculate biomechanical limits for cobots in a previous study. This technique's development process entails normalizing the data over force, resampling them and then fitting the newly obtained samples to a log-normal distribution. The statistical model makes it possible to produce response curves for the same quantile we used for the limits. Our technique adds a confidence region around each response curve to express the sufficiency of the available data. We have produced response curves for 24 different body locations for which we have calculated limits. These curves will enable manufacturers of cobot testing equipment to calibrate their measuring devices precisely.

Anatomical details affect electric field predictions for non-invasive brain stimulation in non-human primates.

Non-human primates (NHPs) have become key for translational research in noninvasive brain stimulation (NIBS). However, in order to create comparable stimulation conditions for humans it is vital to study the accuracy of current modeling practices across species. Numerical models to simulate electric fields are an important tool for experimental planning in NHPs and translation to human studies. It is thus essential whether and to what extent the anatomical details of NHP models agree with current modeling practices when calculating NIBS electric fields. Here, we create highly accurate head models of two non-human primates (NHP) MR data. We evaluate how muscle tissue and head field of view (depending on MRI parameters) affect simulation results in transcranial electric and magnetic stimulation (TES and TMS). Our findings indicate that the inclusion of anisotropic muscle can affect TES electric field strength up to 22% while TMS is largely unaffected. Additionally, comparing a full head model to a cropped head model illustrates the impact of head field of view on electric fields for both TES and TMS. We find opposing effects between TES and TMS with an increase up to 24.8% for TES and a decrease up to 24.6% for TMS for the cropped head model compared to the full head model. Our results provide important insights into the level of anatomical detail needed for NHP head models and can inform future translational efforts for NIBS studies.

OpenMonkeyChallenge: Dataset and Benchmark Challenges for Pose Estimation of Non-human Primates.

The ability to automatically estimate the pose of non-human primates as they move through the world is important for several subfields in biology and biomedicine. Inspired by the recent success of computer vision models enabled by benchmark challenges (e.g., object detection), we propose a new benchmark challenge called OpenMonkeyChallenge that facilitates collective community efforts through an annual competition to build generalizable non-human primate pose estimation models. To host the benchmark challenge, we provide a new public dataset consisting of 111,529 annotated (17 body landmarks) photographs of non-human primates in naturalistic contexts obtained from various sources including the Internet, three National Primate Research Centers, and the Minnesota Zoo. Such annotated datasets will be used for the training and testing datasets to develop generalizable models with standardized evaluation metrics. We demonstrate the effectiveness of our dataset quantitatively by comparing it with existing datasets based on seven state-of-the-art pose estimation models.

Ultra-high field (10.5T) diffusion-weighted MRI of the macaque brain.

Diffu0sion-weighted magnetic resonance imaging (dMRI) is a non-invasive imaging technique that provides information about the barriers to the diffusion of water molecules in tissue. In the brain, this information can be used in several important ways, including to examine tissue abnormalities associated with brain disorders and to infer anatomical connectivity and the organization of white matter bundles through the use of tractography algorithms. However, dMRI also presents certain challenges. For example, historically, the biological validation of tractography models has shown only moderate correlations with anatomical connectivity as determined through invasive tract-tracing studies. Some of the factors contributing to such issues are low spatial resolution, low signal-to-noise ratios, and long scan times required for high-quality data, along with modeling challenges like complex fiber crossing patterns. Leveraging the capabilities provided by an ultra-high field scanner combined with denoising, we have acquired whole-brain, 0.58 mm isotropic resolution dMRI with a 2D-single shot echo planar imaging sequence on a 10.5 Tesla scanner in anesthetized macaques. These data produced high-quality tractograms and maps of scalar diffusion metrics in white matter. This work demonstrates the feasibility and motivation for in-vivo dMRI studies seeking to benefit from ultra-high fields.

Automated pose estimation in primates.

Understanding the behavior of primates is important for primatology, for psychology, and for biology more broadly. It is also important for biomedicine, where primates are an important model organism, and whose behavior is often an important variable of interest. Our ability to rigorously quantify behavior has, however, long been limited. On one hand, we can rigorously quantify low-information measures like preference, looking time, and reaction time; on the other, we can use more gestalt measures like behavioral categories tracked via ethogram, but at high cost and with high variability. Recent technological advances have led to a major revolution in behavioral measurement that offers affordable and scalable rigor. Specifically, digital video cameras and automated pose tracking software can provide measures of full-body position (i.e., pose) of primates over time (i.e., behavior) with high spatial and temporal resolution. Pose-tracking technology in turn can be used to infer behavioral states, such as eating, sleeping, and mating. We call this technological approach behavioral imaging. In this review, we situate the behavioral imaging revolution in the history of the study of behavior, argue for investment in and development of analytical and research techniques that can profit from the advent of the era of big behavior, and propose that primate centers and zoos will take on a more central role in relevant fields of research than they have in the past.

Minimal specifications for non-human primate MRI: Challenges in standardizing and harmonizing data collection.

Recent methodological advances in MRI have enabled substantial growth in neuroimaging studies of non-human primates (NHPs), while open data-sharing through the PRIME-DE initiative has increased the availability of NHP MRI data and the need for robust multi-subject multi-center analyses. Streamlined acquisition and analysis protocols would accelerate and improve these efforts. However, consensus on minimal standards for data acquisition protocols and analysis pipelines for NHP imaging remains to be established, particularly for multi-center studies. Here, we draw parallels between NHP and human neuroimaging and provide minimal guidelines for harmonizing and standardizing data acquisition. We advocate robust translation of widely used open-access toolkits that are well established for analyzing human data. We also encourage the use of validated, automated pre-processing tools for analyzing NHP data sets. These guidelines aim to refine methodological and analytical strategies for small and large-scale NHP neuroimaging data. This will improve reproducibility of results, and accelerate the convergence between NHP and human neuroimaging strategies which will ultimately benefit fundamental and translational brain science.

An 8-dipole transceive and 24-loop receive array for non-human primate head imaging at 10.5 T.

A 32-channel RF coil was developed for brain imaging of anesthetized non-human primates (rhesus macaque) at 10.5 T. The coil is composed of an 8-channel dipole transmit/receive array, close-fitting 16-channel loop receive array headcap, and 8-channel loop receive array lower insert. The transceiver dipole array is composed of eight end-loaded dipole elements self-resonant at the 10.5 T proton Larmor frequency. These dipole elements were arranged on a plastic cylindrical former, which was split into two to allow for convenient animal positioning. Nested into the bottom of the dipole array former is located an 8-channel loop receive array, which contains 5 × 10 cm2 square loops arranged in two rows of four loops. Arranged in a close-fitting plastic headcap is located a high-density 16-channel loop receive array. This array is composed of 14 round loops 37 mm in diameter and 2 partially detachable, irregularly shaped loops that encircle the ears. Imaging experiments were performed on anesthetized non-human primates on a 10.5 T MRI system equipped with body gradients with a 60 cm open bore. The coil enabled submillimeter (0.58 mm isotropic) high-resolution anatomical and functional imaging as well as tractography of fasciculated axonal bundles. The combination of a close-fitting loop receive array and dipole transceiver array allowed for a higher-channel-count receiver and consequent higher signal-to-noise ratio and parallel imaging gains. Parallel imaging performance supports high-resolution functional MRI and diffusion MRI with a factor of three reduction in sampling. The transceive array elements during reception contributed approximately one-quarter of the signal-to-noise ratio in the lower half of the brain, which was farthest from the close-fitting headcap receive array.

Ultra-high field (10.5 T) resting state fMRI in the macaque.

Resting state functional connectivity refers to the temporal correlations between spontaneous hemodynamic signals obtained using functional magnetic resonance imaging. This technique has demonstrated that the structure and dynamics of identifiable networks are altered in psychiatric and neurological disease states. Thus, resting state network organizations can be used as a diagnostic, or prognostic recovery indicator. However, much about the physiological basis of this technique is unknown. Thus, providing a translational bridge to an optimal animal model, the macaque, in which invasive circuit manipulations are possible, is of utmost importance. Current approaches to resting state measurements in macaques face unique challenges associated with signal-to-noise, the need for contrast agents limiting translatability, and within-subject designs. These limitations can, in principle, be overcome through ultra-high magnetic fields. However, imaging at magnetic fields above 7T has yet to be adapted for fMRI in macaques. Here, we demonstrate that the combination of high channel count transmitter and receiver arrays, optimized pulse sequences, and careful anesthesia regimens, allows for detailed single-subject resting state analysis at high resolutions using a 10.5 Tesla scanner. In this study, we uncover thirty spatially detailed resting state components that are highly robust across individual macaques and closely resemble the quality and findings of connectomes from large human datasets. This detailed map of the rsfMRI 'macaque connectome' will be the basis for future neurobiological circuit manipulation work, providing valuable biological insights into human connectomics.

Novel augmented physical simulator for the training of transcatheter cardiovascular interventions.

BACKGROUND: Training in transcatheter cardiovascular skills today represents a significant challenge because of the complexity of the interventions and an extensive use of multiple live imaging technologies. OBJECTIVES: We describe the design, the face validation, and content validation of a newly developed physical transseptal puncture (TSP) simulator using additive manufacturing techniques and novel imaging simulation solutions. METHODS: The TSP simulator contains a femoral vein catheterization pad, silicon phantoms of the venous system, a replaceable interatrial septum, and cameras to mimic live fluoroscopic and echocardiographic imaging. A validation study was conducted at the University Hospital of Zurich. A total of 14 interventional cardiologists and cardiac surgeons assessed the TSP simulator. Participants performed a TSP on the simulator using standard interventional tools. Face and content validity was demonstrated using a 5-point Likert scale. RESULTS: The TSP simulator is a new training tool for transcatheter cardiovascular interventions. All interventional cardiologists and cardiac surgeons completed the training exercise and scoring. Overall impression was rated (out of 5) 4.04 ± 1.03, haptic feedback scored 4.13 ± 0.82, and the realism of fluoroscopy simulation 4.39 ± 0.79. Usability was rated 4.50 ± 0.63 by the participants, indicating that the simulator could be suitable for training. CONCLUSION: We demonstrated face and content validity of a new simulator for transcatheter cardiovascular interventions. The TSP simulator's usability, haptic feedback, imaging solutions, and the overall impression of its usage were reported as very realistic. The TSP simulator represents a promising tool for simulation-based training using real interventional toolkits in a mimicked radiological environment.

Ultrasonic sensor concept to fit a ventricular assist device cannula evaluated using geometrically accurate heart phantoms.

Future left ventricular assist devices (LVADs) are expected to respond to the physiologic need of patients; however, they still lack reliable pressure or volume sensors for feedback control. In the clinic, echocardiography systems are routinely used to measure left ventricular (LV) volume. Until now, echocardiography in this form was never integrated in LVADs due to its computational complexity. The aim of this study was to demonstrate the applicability of a simplified ultrasonic sensor to fit an LVAD cannula and to show the achievable accuracy in vitro. Our approach requires only two ultrasonic transducers because we estimated the LV volume with the LV end-diastolic diameter commonly used in clinical assessments. In order to optimize the accuracy, we assessed the optimal design parameters considering over 50 orientations of the two ultrasonic transducers. A test bench was equipped with five talcum-infused silicone heart phantoms, in which the intra-ventricular surface replicated papillary muscles and trabeculae carnae. The end-diastolic LV filling volumes of the five heart phantoms ranged from 180 to 480 mL. This reference volume was altered by ±40 mL with a syringe pump. Based on the calibrated measurements acquired by the two ultrasonic transducers, the LV volume was estimated well. However, the accuracies obtained are strongly dependent on the choice of the design parameters. Orientations toward the septum perform better, as they interfere less with the papillary muscles. The optimized design is valid for all hearts. Considering this, the Bland-Altman analysis reports the LV volume accuracy as a bias of ±10% and limits of agreement of 0%-40% in all but the smallest heart. The simplicity of traditional echocardiography systems was reduced by two orders of magnitude in technical complexity, while achieving a comparable accuracy to 2D echocardiography requiring a calibration of absolute volume only. Hence, our approach exploits the established benefits of echocardiography and makes them applicable as an LV volume sensor for LVADs.

Visual Behaviour Strategies of Operators during Catheter-Based Cardiovascular Interventions.

The aim was to gain insights into the visual behaviour and the perceptual skills of operators during catheter-based cardiovascular interventions (CBCVIs). A total of 33 CBCVIs were performed at the University Hospital Zurich by five operators, two experts and three novices, while wearing eye tracking glasses. The visual attention distribution on three areas of interest (AOIs) the "Echo screen", "Fluoro screen" and "Patient" was analysed for the transseptal puncture procedure. Clear visual behaviour patterns were observable in all cases. There is a significant differences in visual attention distribution of the experts compared to the novices. Experts spent 79% of dwell time on the Echo screen and 17% on the Fluoro screen, novices spent 52% on the Echo screen and 40% on the Fluoro screen. Additionally, results showed that experts focused their gaze on smaller areas than novices during critical interventional actions. Operators seem to exhibit identifiable visual behaviour patterns for CBCVIs. These identifiable patterns were significantly different between the expert and the novice operators. This indicates that the visual behaviour of operators could be employed to assist transfer of experts' perceptual skills to novices and to develop tools for objective performance assessment.

Decoding the direction of imagined visual motion using 7T ultra-high field fMRI.

There is a long-standing debate about the neurocognitive implementation of mental imagery. One form of mental imagery is the imagery of visual motion, which is of interest due to its naturalistic and dynamic character. However, so far only the mere occurrence rather than the specific content of motion imagery was shown to be detectable. In the current study, the application of multi-voxel pattern analysis to high-resolution functional data of 12 subjects acquired with ultra-high field 7T functional magnetic resonance imaging allowed us to show that imagery of visual motion can indeed activate the earliest levels of the visual hierarchy, but the extent thereof varies highly between subjects. Our approach enabled classification not only of complex imagery, but also of its actual contents, in that the direction of imagined motion out of four options was successfully identified in two thirds of the subjects and with accuracies of up to 91.3% in individual subjects. A searchlight analysis confirmed the local origin of decodable information in striate and extra-striate cortex. These high-accuracy findings not only shed new light on a central question in vision science on the constituents of mental imagery, but also show for the first time that the specific sub-categorical content of visual motion imagery is reliably decodable from brain imaging data on a single-subject level.

Value signals in the prefrontal cortex predict individual preferences across reward categories.

Humans can choose between fundamentally different options, such as watching a movie or going out for dinner. According to the utility concept, put forward by utilitarian philosophers and widely used in economics, this may be accomplished by mapping the value of different options onto a common scale, independent of specific option characteristics (Fehr and Rangel, 2011; Levy and Glimcher, 2012). If this is the case, value-related activity patterns in the brain should allow predictions of individual preferences across fundamentally different reward categories. We analyze fMRI data of the prefrontal cortex while subjects imagine the pleasure they would derive from items belonging to two distinct reward categories: engaging activities (like going out for drinks, daydreaming, or doing sports) and snack foods. Support vector machines trained on brain patterns related to one category reliably predict individual preferences of the other category and vice versa. Further, we predict preferences across participants. These findings demonstrate that prefrontal cortex value signals follow a common scale representation of value that is even comparable across individuals and could, in principle, be used to predict choice.

Ethological computational psychiatry: Challenges and opportunities.

Studying the intricacies of individual subjects' moods and cognitive processing over extended periods of time presents a formidable challenge in medicine. While much of systems neuroscience appropriately focuses on the link between neural circuit functions and well-constrained behaviors over short timescales (e.g., trials, hours), many mental health conditions involve complex interactions of mood and cognition that are non-stationary across behavioral contexts and evolve over extended timescales. Here, we discuss opportunities, challenges, and possible future directions in computational psychiatry to quantify non-stationary continuously monitored behaviors. We suggest that this exploratory effort may contribute to a more precision-based approach to treating mental disorders and facilitate a more robust reverse translation across animal species. We conclude with ethical considerations for any field that aims to bridge artificial intelligence and patient monitoring.

Neural timescales reflect behavioral demands in freely moving rhesus macaques.

Previous work demonstrated a highly reproducible cortical hierarchy of neural timescales at rest, with sensory areas displaying fast, and higher-order association areas displaying slower timescales. The question arises how such stable hierarchies give rise to adaptive behavior that requires flexible adjustment of temporal coding and integration demands. Potentially, this lack of variability in the hierarchical organization of neural timescales could reflect the structure of the laboratory contexts. We posit that unconstrained paradigms are ideal to test whether the dynamics of neural timescales reflect behavioral demands. Here we measured timescales of local field potential activity while male rhesus macaques foraged in an open space. We found a hierarchy of neural timescales that differs from previous work. Importantly, although the magnitude of neural timescales expanded with task engagement, the brain areas' relative position in the hierarchy was stable. Next, we demonstrated that the change in neural timescales is dynamic and contains functionally-relevant information, differentiating between similar events in terms of motor demands and associated reward. Finally, we demonstrated that brain areas are differentially affected by these behavioral demands. These results demonstrate that while the space of neural timescales is anatomically constrained, the observed hierarchical organization and magnitude is dependent on behavioral demands.

Induced neural phase precession through exogenous electric fields.

The gradual shifting of preferred neural spiking relative to local field potentials (LFPs), known as phase precession, plays a prominent role in neural coding. Correlations between the phase precession and behavior have been observed throughout various brain regions. As such, phase precession is suggested to be a global neural mechanism that promotes local neuroplasticity. However, causal evidence and neuroplastic mechanisms of phase precession are lacking so far. Here we show a causal link between LFP dynamics and phase precession. In three experiments, we modulated LFPs in humans, a non-human primate, and computational models using alternating current stimulation. We show that continuous stimulation of motor cortex oscillations in humans lead to a gradual phase shift of maximal corticospinal excitability by ~90°. Further, exogenous alternating current stimulation induced phase precession in a subset of entrained neurons (~30%) in the non-human primate. Multiscale modeling of realistic neural circuits suggests that alternating current stimulation-induced phase precession is driven by NMDA-mediated synaptic plasticity. Altogether, the three experiments provide mechanistic and causal evidence for phase precession as a global neocortical process. Alternating current-induced phase precession and consequently synaptic plasticity is crucial for the development of novel therapeutic neuromodulation methods.

Neural timescales reflect behavioral demands in freely moving rhesus macaques.

Previous work has demonstrated remarkably reproducible and consistent hierarchies of neural timescales across cortical areas at rest. The question arises how such stable hierarchies give rise to adaptive behavior that requires flexible adjustment of temporal coding and integration demands. Potentially, this previously found lack of variability in the hierarchical organization of neural timescales could be a reflection of the structure of the laboratory contexts in which they were measured. Indeed, computational work demonstrates the existence of multiple temporal hierarchies within the same anatomical network when the input structure is altered. We posit that unconstrained behavioral environments where relatively little temporal demands are imposed from the experimenter are an ideal test bed to address the question of whether the hierarchical organization and the magnitude of neural timescales reflect ongoing behavioral demands. To tackle this question, we measured timescales of local field potential activity while rhesus macaques were foraging freely in a large open space. We find a hierarchy of neural timescales that is unique to this foraging environment. Importantly, although the magnitude of neural timescales generally expanded with task engagement, the brain areas' relative position in the hierarchy was stable across the recording sessions. Notably, the magnitude of neural timescales monotonically expanded with task engagement across a relatively long temporal scale spanning the duration of the recording session. Over shorter temporal scales, the magnitude of neural timescales changed dynamically around foraging events. Moreover, the change in the magnitude of neural timescales contained functionally relevant information, differentiating between seemingly similar events in terms of motor demands and associated reward. That is, the patterns of change were associated with the cognitive and behavioral meaning of these events. Finally, we demonstrated that brain areas were differentially affected by these behavioral demands - i.e., the expansion of neural timescales was not the same across all areas. Together, these results demonstrate that the observed hierarchy of neural timescales is context-dependent and that changes in the magnitude of neural timescales are closely related to overall task engagement and behavioral demands.

OpenApePose, a database of annotated ape photographs for pose estimation.

Because of their close relationship with humans, non-human apes (chimpanzees, bonobos, gorillas, orangutans, and gibbons, including siamangs) are of great scientific interest. The goal of understanding their complex behavior would be greatly advanced by the ability to perform video-based pose tracking. Tracking, however, requires high-quality annotated datasets of ape photographs. Here we present OpenApePose, a new public dataset of 71,868 photographs, annotated with 16 body landmarks of six ape species in naturalistic contexts. We show that a standard deep net (HRNet-W48) trained on ape photos can reliably track out-of-sample ape photos better than networks trained on monkeys (specifically, the OpenMonkeyPose dataset) and on humans (COCO) can. This trained network can track apes almost as well as the other networks can track their respective taxa, and models trained without one of the six ape species can track the held-out species better than the monkey and human models can. Ultimately, the results of our analyses highlight the importance of large, specialized databases for animal tracking systems and confirm the utility of our new ape database.

All animals carry out a wide range of behaviors in everyday life, such as feeding and communicating with one another. Understanding the complex behavior of non-human apes such as chimpanzees, bonobos, gorillas, orangutans, and various gibbons is of great interest to scientists due to their close relationship with humans. Each behavior is made up of a string of poses that an animal makes with its body. To analyze them in a reliable and consistent way, scientists have developed automated pose estimation methods that determine the position of body parts from photographs and videos. While these systems require minimal external input to perform, they need to be trained on a large dataset of high-quality annotated images of the target animals to teach the system what to look for. So far, scientists have relied on systems trained on monkey and human images to analyze ape data. However, apes are particularly challenging to track because their body textures are uniform, and they have a large number of poses. Therefore, for the most accurate tracking of ape behaviors, a dedicated training dataset of annotated ape images is required. Desai et al. filled this gap by creating the "OpenApePose" dataset, which contains 71,868 photographs of apes from six species, annotated using 16 body landmarks. To test the dataset, the researchers trained an artificial intelligence network on separate monkey, human and ape datasets. The findings showed that the network is better at tracking apes when trained on ape images rather than those of monkeys or humans. It is also equally good at tracking apes as other monkey and human networks are at tracking their own species. This is contrary to optimistic expectations that monkey and human models could be generalized to apes. Training the network without images of one of the six ape species showed that it can still track the excluded species better than monkey and human models can. These experiments highlight the importance of species and family-specific datasets. OpenApePose is a valuable resource for researchers from various fields. It can aid tracking of animal behavior in the wild using large quantities of footage recorded by camera traps and drones. Artificial intelligence models trained on the OpenApePose dataset could also help scientists ­ such as neuroscientists ­ link movement with other types of data, including brain activity measurements, to gain deeper insights into behavior.