Mymou: A low-cost, wireless touchscreen system for automated training of nonhuman primates

Abstract

Training nonhuman primates (NHPs) to perform cognitive tasks is essential for many neuroscientific investigations, yet laboratory training is a time-consuming process with inherent limitations. Habituating NHPs to the laboratory staff and experimental equipment can take months before NHPs are ready to proceed to the primary tasks. Laboratory training also necessarily separates NHPs from their home-room social group and typically involves some form of restraint or limited mobility, and data collection is often limited to a few hours per day so that multiple NHPs can be trained on the same equipment. Consequently, it can often take a year to train NHPs on complex cognitive tasks. To overcome these issues, we developed a low-cost, open-source, wireless touchscreen training system that can be installed in the home-room environment. The automated device can run continuously all day, including over weekends, without experimenter intervention. The system utilizes real-time facial recognition to initiate subject-specific tasks and provide accurate data logging, without the need for implanted microchips or separation of the NHPs. The system allows NHPs to select their preferred reward on each trial and to work when and for as long as they desire, and it can analyze task performance in real time and adapt the task parameters in order to expedite training. We demonstrate that NHPs consistently use this system on a daily basis to quickly learn complex behavioral tasks. The system therefore addresses many of the welfare and experimental limitations of laboratory-based training of NHPs and provides a platform for wireless electrophysiological investigations in more naturalistic, freely moving environments.

Keywords: Animal welfare; Automated testing; Cognitive neuroscience; Facial recognition; Non-human primates.

Figure 1.

Components of the Mymou system A, The system is fully wireless and uses Bluetooth to communicate with reward delivery systems. B, 3D schematic of the holder that was used to secure the tablet to the outside of the cage. Consisting of 1) tablet, 2) metal locking pin, 3) hole for optional charging cable, 4) tablet holder, 5) sloped funnel, 6) front panel, 7) hole to lock the holder to the cage, 8) reward tubes, 9) hooks to mount holder to cage. C, Photo of the holder depicted in B. Numbers correspond to the same items in B. D, Photo of the microcontroller (1) connected to a 4-channel peristaltic pump (2)

Figure 2.

The Mymou system in use A, One of the non-human primates using the device. B, A selfie taken by the device triggered by Subject O initiating a trial. The stainless steel spouts used for delivery of a juice reward can be seen at the centre bottom. C, An example of subject V receiving reward while attend-ing to the task on screen. D, The distribution of when each NHP would complete trials throughout a single session (n = 7). Error bars represent standard error of the mean. E, The number of trials completed between both NHP’s across all sessions. Note that the tasks they were doing varied dramatically over this period

Figure 3.

A task that was used on the Mymou system A, Subjects were taught a network of 16 stimuli arranged in a 4x4 grid, with each stimulus being associated with its neighbours. From this network 68 unique paths of distance two were present. For example starting at the green highlighted stimulus and ending at the blue highlighted stimulus. B, Schematic of the task that was used to teach the associative network shown in A. When a trial is intitiated a goal location and four possible options to choose from are presented (choice 1). Upon making a choice animations then provide intuitive, engaging feedback to the subject. In total two correct choices were needed to receive a reward. Getting either choice wrong results in trial termination and a timeout of 1-3 seconds

Figure 4.

Subjects learnt effectively on the Mymou system A, Total number of interactions with the system over 18 training sessions across a 3 week period. B, Daily performance over the three week period on the task at navigating two transition problems. C, Same as in B but for when given harder problems three transitions in length

Figure 5.

Online facial recognition A, A cropped selfie of subject O that was used for identification of the subject. B, A cropped selfie of subject V that was used for identification of the subject. C, Schematic of the neural network that was used for facial recognition. Each circle represents a neuron and each arrow represents a weight that modulates the input from the upstream neuron. The network was able to correctly identify subjects with 98.89% accuracy. D, Examples of images correctly identifed by the ANN of subject O (top row) and V (bottom row). E, Timeline of the online facial recognition process, data taken from 10 iterations of the process

Figure 6.

The Mymou system can control multiple reward delivery systems A, After each success trial subjects were given a choice of four different juices to pick as a reward. The amount they picked each different juice across all days tested is displayed. B, Juice preferences over a single day, taken from day 11 in A to show the switching behaviors that were sometimes exhibited. C, Matrix of overall juice preferences for each subject, calculated from all the data shown in A