A chinrest-based approach to measure eye movements and experimental task engagement in macaques with minimal restraint

Background: The use of Rhesus macaques in vision research is crucial due to their visual system ’ s similarity to humans. While invasive techniques have been the norm, there has been a shift towards non-invasive methods, such as facemasks and head molds, to enhance animal welfare and address ethical concerns. New Method: We present a non-invasive, 3D-printed chinrest with infrared sensors, adapted from canine research, allowing for accurate eye movement measurements and voluntary animal participation in experiments. Results: The chinrest method showed a 16% and 28% increase in average trial numbers for Monkey 1 and Monkey 2, respectively, compared to the traditional headpost method. The engagement was high, with monkeys performing over 500 trials per session and initiating a new trial after an average intertrial interval of approximately 1 second. The hit rate improved by about 10% for Monkey 1 in the chinrest condition, and the fixation precision, measured by the standard deviation of gaze positions, was significantly better in the chinrest condition, with Monkey 1 showing a reduction in fixation imprecision from 0.26 ◦ to 0.17 ◦ in the X-axis. Comparison with Existing Methods: The chinrest approach showed significant improvements in trial engagement and reduction in aborted trials due to fixation breaks, indicating less stress and potentially improved data quality compared to previous non-invasive methods. Conclusions: The chinrest method offers a significant advancement in primate cognitive testing by allowing for precise data collection while addressing animal welfare concerns, possibly leading to better scientific outcomes and a paradigm shift in primate research methodologies.


Introduction
Rhesus macaques have played a pivotal role in deepening our comprehension of cognitive brain functions, both in healthy and neurological contexts (Lear et al., 2022;Treue and Lemon, 2023).Particularly in vision research, monkeys are the preferred animal species due to the similarities in visual system organization between them and humans.These non-human primates can be adeptly trained to undertake a myriad of cognitive and perceptual tasks, where they communicate their experiences through specific lever presses or designated eye movements.Typically, this training and testing serve as the foundational steps in studies, preceding invasive brain function measurements or manipulations that are impractical in humans.
When initiating an experimental session that involves advancing electrodes in awake non-human primates to a specific brain structure, it's essential to use head fixation implants, usually affixed to the skull which ensure the stability required for neural recordings.However, this invasive strategy is not without its drawbacks, including heightened risks of chronic wounds, skin retractions, and infections.Such traditional techniques might introduce both pathocentric (pain and suffering) and non-pathocentric strains (excessive instrumentation and significant alterations to appearance or abilities) on the animals, a concept that is incorporated into the ethical evaluation of constraint on animals in Switzerland.These strains not only jeopardize animal welfare but also present ethical dilemmas.
In this study, we leverage these advancements, presenting a 3Dprinted chinrest equipped with infrared sensors that allow to detect the monkey's head.We assess the chinrest's capability to facilitate visually guided tasks paired with eye-movement data.Our analysis juxtaposes this chinrest method with the traditional headpost implantbased testing in two monkeys.These subjects underwent passive fixation and active target detection, conveying their perceptions through saccades.

Chinrest design
We designed the chin rest to be tailored to the monkey head shape using a CT scan.Our goal was to ensure maximum voluntary engagement from the monkeys while reducing pressure points and providing a comfortable resting position with minimal distress and limited movement.The 3D stl files of the chinrest are available upon request.The chinrest was 3D printed in three separate parts, including a central and two peripheral parts that fit together through a sliding mechanism, allowing to adjust the width.As a part of the printing process, we added 7 pairs of aligned apertures to serve as insertion locations for infrared sensors.The chinrest also featured a rectangular metal bar with a slot along its length for adjusting the antero-posterior position and a metal spout connected to plastic tubing for reward delivery, positioned just close enough to the animal's mouth without making contact.This precision and attention to detail ensures the monkey's comfort and engagement during the experimental process.
To ensure full engagement and proper resting position on the chin rest, we incorporated an infrared sensing system (Supplementary material).The system consists of two main optical components -one emitting and the other receiving infrared light -and an electronic threshold detection circuit.The optical components are a TX phototransistor and an RX phototransistor, which work together to detect when the beam of light is interrupted by an obstacle.This interruption causes the impedance of the phototransistor to increase, resulting in a change in voltage that triggers the circuit to output a high signal.The presence of the head in the chinrest then triggers the initiation of the trial by displaying a fixation spot on the computer monitor.
The multiple-sensing circuit is made up of 7 equal channels with an operational amplifier configured as a simple threshold comparator.The sensitivity of the sensors can be adjusted using a potentiometer, which controls the voltage threshold for the output signal.This circuit is compatible with diverse types of optical components and is powered by a 6.2 V voltage for a maximum output voltage of 5 V, making it compatible with common TTL/CMOS devices.The diode D2 serves as an output indicator.The wavelength of the infrared light used is 880 nm.Fig. 1

Eye movement measurement and saccade detection
We utilized an infrared eye tracker (Eyelink 1000, SR research) positioned 50 cm from the monkey's head to measure eye movements.Each session required calibrating the tracker by adjusting the raw eye position signal while the animal fixated on 15 dots spanning 15 • horizontally and 8 • vertically.For untrained animals, we initiated calibration using data from humans or trained animals.Saccades were detected using a modified velocity-based algorithm (Engbert and Kliegl, 2003).After normalizing eye positions and converting to velocities, saccades were identified when velocities exceeded 6x the standard deviation, with a minimum duration of 10 ms and an intersaccadic interval of 20 ms.

Visual stimulation
We provided back projection visual stimulation on a translucent screen using a ProPixx projector at a resolution of 1920×1080 pixels with a refresh rate of 120 Hz (SR Research).The distance to the screen was 57 cm and the screen subtended a total of 1170 mm * 650 mm, which was equal to 91.49 degrees of viewing angle.

Animals
Two adult male rhesus macaques were used for this study (7 and 8 kg).They were housed together in a 45 m 3 home space with access to an outside compartment and natural daylight.The light period was centrally controlled and set to a 12/12 h dark/light cycle (light on between 06:00 am and 06:00 pm).Animals were weighed daily to make sure that their weight was stable.The monkeys were given primate pellets (Granovit SA, Swiss), seeds and vegetables daily and had free access to water.High palatable food items, like nuts or dry fruits were reserved for training.Food reward was used for chair training (Mason et al., 2019).To advance subsequent cognitive task training a reward schedule was developed, in which sweetened fluids were typically given on a trial-by-trial basis and palatable foods intermittently on a reward schedule depending on task goals and performance.All animal procedures were performed in accordance with the Swiss veterinary authorities (cantonal and federal) who reviewed and approved the experimental procedures.

Acclimatization and surgery procedures
Upon arrival at the animal facility, the monkeys were acclimatized to their new environment for 3 weeks.After acclimatization, monkeys were trained to enter a cage in their home enclosure, through which they would access a custom-made primate chair.Once monkeys were fully trained to enter their chair and present their head through the top opening (Mason et al., 2019), they were moved to the lab environment.Training and testing in the lab were both performed using the novel chinrest-based approach and traditional implant-based head fixation.For the latter approach, an individualized, 2 part (Psarou et al., 2023) head post was implanted following previously described procedures (Mason et al., 2019;Ortiz-Rios et al., 2023).

Visual fixation training using the chinrest approach
We trained monkeys by associating chin placement in the chinrest with food rewards, progressively introducing visual stimuli for increased task complexity.This method enabled self-paced learning.The appearance of a visual stimulus linked to chin placement was crucial for eyetracker calibration.During calibration, monkeys fixated on target dots across the screen.By overlaying food stimuli on these targets, we directed the monkey's gaze to the fixation window.After several trials, monkeys associated target fixation with rewards.The two subjects underwent calibration to enhance gaze accuracy, reducing both the visual target size (from 3 • to 0.5 • ) and the fixation window (from 5 • to 2 • ).This also tested the system's head motion sensitivity.To maintain motivation, varied food rewards were given, achieving over 300 correct trials in 1-1.5-hour sessions.Since food rewards could introduce large jaw movements that could impinge on the eye-tracking signal, we provided reward after the offset of a successful trial, and before the onset of a new trial, to avoid contamination of the eye signal by chewing movements.Furthermore, to ensure only minimum chewing, we provided only pellet sized portions of food.

Perceptual report training: The contrast detection task
After animals consistently fixated on targets for over 1 s, they were trained on a yes/no contrast detection task to identify a white disc's presence or absence (Ortiz-Rios et al., 2023) on top of gray background (40 cd/cm 2 ).Trials began with the animal focusing on a red point, which turned green within 300-500 ms, indicating successful initiation.Initial training displayed the disc at maximum contrast with a response dot to its left or right, depending on the disc's presence.As the animal's accuracy reached 90%, the task's complexity increased by introducing varied disc contrasts.Post-initiation, two red response dots appeared alongside a white disc in the bottom left visual field, which could have contrasts ranging from 2.5% to 20%.In half the trials, the disc was absent.The animal's correct response to the disc's presence or absence, by making a saccade to the appropriate red dot, earned a reward.

Engagement
Monkey engagement is necessary to achieve enough trials within the imparted experimental time.We quantified monkey engagement by estimating the average time it takes a monkey to initiate a trial (intertrial time), average session length in minutes, the number of trials the monkey can achieve within an experimental session, the total length of the session and the number of aborted trials due to fixation break.

Task performance and consistency
Successful testing of cognitive function is rooted in reliable abovechance task performance.We quantified performance by calculating the hit rate and false alarms during the contrast detection task.
While satisfactory performance is essential for a successful test session, it is generally desirable to achieve a high degree of behavioral consistency across training and test sessions.We therefore quantified consistency as the absolute difference in hit rate and false alarms between successive days and compared this measure between the two head stabilization methods.

Fixation precision
One key benchmark of the success of our approach is to obtain headfixation quality eye-tracking data.The precision of gaze during fixation is a reliable indicator of how much head movement can contaminate the eye signal.We quantified fixation precision as the standard deviation of positions of the gaze in X and Y coordinates in • of visual angle, around the fixation point, within a 2 • window.A larger standard deviation around the fixation point position shows lower precision.We analyzed fixation stability outside of any micro-saccadic interference.We separately compared microsaccade rates between the chinrest and the headpost conditions.Gaze drift speeds were also compared between conditions.

Saccade landing precision and accuracy
Other reliable indicators of steady eye-tracking are saccade landing accuracy and precision.Hereby accuracy refers to how closely the saccade lands on the intended visual target.A saccade is considered accurate if it lands at the desired location.Therefore, to quantify accuracy, we measured the mean difference between the saccade landing position and intended target in the chinrest versus head-post condition.
On the other hand, a saccade is considered precise if it consistently lands in the same location on repeated attempts.Imprecision is quantified as the standard deviation of saccade endpoint positions across trials.A larger standard deviation indicates a less precise saccade landing.

Results
Our primary objective was to contrast the monkeys' behavior under traditional implant-based head fixation versus the chin-rest approach.Notably, chin-rest training preceded implant-based testing.Despite inherent behavioral differences between the monkeys, the chin-rest method showed consistent effects in both.Consistent effect here means that the observed outcomes or behavioral responses using the chinrest method were reliable and uniform across different monkeys, despite potential variations in their individual behaviors.For Monkey 1, we analyzed 17 sessions pre-and post-headpost implantation.For Monkey 2, we compared 13 sessions (2 weeks) before and 13 (2 weeks) after implantation.

Engagement
Probably the most critical aspect of any behavioral testing approach is that the animal freely engages with the testing situation.Irrespective of the restraint condition (chinrest or implant-based), we found that the two tested monkeys were well engaged in the task, performing on average well above 500 trials in a testing session, aborting less than 10% of all trials and initiating a new trial after a pause (intertrial time) of about 1 s.
There were no statistically significant differences between the two methods concerning the number of successfully performed trials (Fig. 2A).These were on average 16% (M1) or 28% (M2) higher with the chinrest approach (Statistical Table 1a).Session length was not significant between the two conditions for M1.However, M2 had on average slightly shorter sessions after receiving the headpost than M1 (Fig. 2B, Statistical Table 1b).Trial rate (Fig. 2C) was not significantly higher in the chin-rest condition compared to the head-post condition for both and M2 (Statistical Table 1c).
Both monkeys had increased percentages of aborted trials due to fixation break in the head post condition (Statistical Table 1d), however, the increase was significant only for M1 (Fig. 2D).Initiating a trial under the chinrest condition was significantly faster compared to the headpost condition for both M1 and M2 (Statistical Table 1e).Fig. 3

Performance and consistency
The chinrest approach outperformed the headpost approach in terms of hit rate (Fig. 3 A, Statistical Table 1f).)by about 10% for M1.For M2 both conditions yielded similar hit rates.The average difference between hit rates on successive days (Fig. 3B, Statistical Table 1g) was significantly higher for M1, but not for M2.
False alarms represent instances where the monkeys incorrectly identified a stimulus when it was not actually present, i.e. during catch trials.Regarding false alarms (Fig. 3 C, Statistical Table 1h), there were no significant differences for M1 between the two conditions, whereas for M2, the head-fixed condition led to significantly lower false alarms.The average difference between false alarm rates of successive days (Fig. 3D, Statistical Table 1i) did not differ between the 2 conditions either for M1, or for M2.

Fixation stability
A central aspect of any visual function testing is establishing conditions that allow for precise gaze tracking.M1 displayed a significantly higher gaze precision around the fixation point in the chinrest compared to headpost condition for both X (Statistical Table 1j; d=-0.87,p>0.005) and Y (Statistical Table 1k; d=-0.66,p>0.05) axis.
M2 showed not significant differences between chinrest and headpost conditions on either axis (Statistical Tables 1j, 1k).
As an additional measure of fixation stability (Chung et al., 2015), we compared microsaccade rates during the fixation period preceding stimulus onset in chinrest and head-fixed conditions.Microsaccades.As shown in Fig. 4C, both M1 and M2 showed significant increases of microsaccade rates under the head-fixed condition (Statistical Table 1l).
We also compared ocular drift speed between the chinrest and headpost conditions (Fig. 4D) and found that that for both M1 and M2 drift speed significantly decreased in the headpost conditions Statistical Table 1m), likely indicating reduced contamination by head drift.

Saccade landing position precision and accuracy
Another key parameter for successful visual field testing is precise and accurate saccade targeting of visual objects.Accuracy refers to how far the saccade lands from the target, while imprecision refers to the variability of the saccade landing position.We found no statistically different accuracies between conditions on the X and Y axis (Fig. 5B, Statistical Tables 1n, 1o) for M1.M2 showed a greater accuracy of X axis saccade landing positions under the head fixed condition (Fig. 5B, Statistical Table 1n; d=-0.705,p>0.05).In terms of saccade landing precision, we found no statistically significant differences between the chinrest and headpost conditions in either axis for M1 (Fig. 5B, Statistical Tables 1p, 1q).However, M2 showed a statistically significant improvement (16% increase) in y axis landing precision in the headpost condition (Fig. 5B, Statistical Table 1q; d=1.55, p>0.005).

Discussion
Building up on previous developments for canine (Berns et al., 2012;Karl et al., 2020) and macaque (De Luna et al., 2014;Drucker et al., S. Rima et al. 2015;Fairhall et al., 2006;Kawaguchi et al., 2019;Machado and Nelson, 2011;Slater et al., 2016) cognitive testing with minimal restraint, we have further advanced a non-invasive system for training macaque monkeys that uses a chinrest with built-in sensors for position detection.We compared various aspects of task engagement and oculomotor precision in two monkeys that underwent first testing using the chin rest method before undergoing further tests with traditional implant-based head stabilization.In the following, we summarize our major findings, relate them to similar approaches and conclude with a discussion on further improvements.
Both monkeys contributing to our study exhibited enhanced engagement and fewer aborted trials in the chinrest condition, indicating a heightened level of motivation with this method.When it came to performance accuracy, the chinrest condition consistently matched or surpassed the headpost in hit rates.
As for ocular behavior, the chinrest condition resulted in reduced fixation imprecision and fewer microsaccades.The chin-rest approach thus appears equal or perhaps even superior to earlier developments: Slater et al. reported gaze fixation performance during tests with a helmet-based system to be inferior to the one under traditional headfixed condition.Kawaguchi et al. (2019), who compared a snout-based system to traditional head fixation reported comparable gaze fixation stability for both methods.The benefits of the chinrest approach in this study compared to earlier approaches might be attributed to the greater movement freedom, perceived self-control or reduced stress enabled by this method.
Our study has clear limitations that need to be considered.A larger sample size is essential for validating and generalizing our findings.Testing more monkeys and considering individual differences, like age or sex, will offer a broader understanding of the chinrest method's applicability.Potential biases may have arisen from testing monkeys before and after head-post implantation.A direct comparison between the chinrest and traditional cranial implants, using varied monkey groups under consistent conditions, would clarify their respective merits.Furthermore, some of our results may be due to differences in experienced stress, which we have not assessed in this study.While we interpret the observed reductions in trial number (Fig. 2A) and performance (Fig. 3 A) during the headpost condition to the higher movement restriction and potential discomfort associated with headposting, it could be argued that these results could alternatively reflect decreased interest in the task over time.Testing this latter possibility could be done by reversing the testing sequence between the two conditions.
Evidently, the chinrest design can be further refined.Collaborations with engineering and animal behavior experts can enhance sensor accuracy and comfort.Advanced technologies, like real-time 3D head mapping, can improve measurement precision.The new method might be particularly useful for training monkeys to learn new tasks and perform them with a high trial count and gaze precision, before testing using more invasive methods for acutely recording and manipulating brain signals commences.Alternatively, the chinrest-based approach could be very elegantly combined with chronic tethered or wireless neural recording or manipulation approaches.

Table 1
Statistical analysis.Significant metrics and their corresponding figures are reported in Bold and color (Fig. 2,Fig. 3,Fig. 4,Fig. 5) In conclusion, advancing this research on non-invasive methods for macaque cognitive and visual function assessment is pivotal.Comprehensive validation, analysis, and refinement will drive the development of ethical and reliable methods, propelling biomedical research with macaques forward.

Fig. 1 .
Fig. 1. A. Picture of the chinrest mediated eye-tracking setup.B. Top: 3D model of the chinrest showing the 3-part design (the blue medial part serves to connect the red and green lateral parts as well as the rail fitting mechanism in grey.Bottom: 3D print of the chinrest fitted with the infrared sensors on the sides of the chinrest.C. Top: Schematic of monkey position in chair and chinrest.Bottom: function of the chinrest: when the monkey's head interrupts the IR beams, visual stimuli are shown, and the monkey can interact with them.When the monkey removes its head from the chinrest, IR beams are established again and the visual stimuli disappear.

Fig. 2 .
Fig. 2. Monkey engagement. A. Total number of trials achieved within an experimental session.B. Average session length in the chinrest and headpost conditions.C. Average trial rate per session.D. Percent of aborted trials.E. Average intertrial interval per session.Chinrest and headpost conditions are indicated by and , respectively.M1 is shown on top and M2 on the bottom of each figure.