O mouse, where art thou? The Mouse Position Surveillance System (MoPSS)—an RFID-based tracking system

Existing methods for analysis of home cage-based preference tests are either time-consuming, not suitable for group management, expensive, and/or based on proprietary equipment that is not freely available. To correct this, we developed an automated system for group-housed mice based on radio frequency identification: the Mouse Position Surveillance System (MoPSS). The system uses an Arduino microcontroller with compatible components; it is affordable and easy to rebuild for every laboratory because it uses free and open-source software and open-source hardware with the RFID readers as the only proprietary component. The MoPSS was validated using female C57BL/6J mice and manual video comparison. It proved to be accurate even for fast-moving mice (up to 100% accuracy after logical reconstruction), and is already implemented in several studies in our laboratory. Here, we provide the complete construction description as well as the validation data and the results of an example experiment. This tracking system will allow group-based preference testing with individually identified mice to be carried out in a convenient manner. This facilitation of preference tests creates the foundation for better housing conditions from the animals’ perspective. Supplementary Information The online version contains supplementary material available at 10.3758/s13428-021-01593-7.

However, if a mouse while changing between cages is not detected by one or by 28 both RFID antennas, the simple approach of looking at consecutive detections 29 does not work anymore. In the following, we will explain how to handle these 30 situations by deducing the mouse position from the available data.

64
Our main focus was first, to find the cage changes in which one RFID an-65 tenna did not detect the mouse (B and C), and to correct possible false 66 timestamps (C, as the time belongs to a new cage change when from the 67 antenna's perspective the mouse appears for the first time on this side), and 68 second, to identify cage changes which were completely missed by the anten-69 nas (D, wherever possible, as explained above). To achieve this, we developed 70 an R script to help with the logical reconstruction of the data (available here: 71 https://zenodo.org/record/4650404). At the end, in Table 1 it is shown how the output of the evaluation script of Experiment 1 (validation) looks like.
Our dataset contains a timestamp and the antenna number where the detec-74 tion occurred. Apart from cage changes, around 62 % (see Table 1) of our 75 data points consist of detections we considered "pokes". These are detections 76 in which a mouse is recorded (multiple times) at the antenna without passing 77 through it. This is due to dwelling near the beginning of the tube.

78
In short, the procedure of the R script is as follows (also depicted as a schematic both RFID antennas must have missed the mouse (Fig. 1 D). 94 Second, if there was only one detection between the two safe cage changes, 95 one of the two antennas must have missed the mouse (Fig. 1 B or C). As fast, we decided to accept the introduced uncertainty of a few seconds.
Third, if there was more than one detection between the two safe cage Situation D two times in a row: Although extremely rare, it is possible 136 that a mouse is missed by both antennas when passing through the tube 137 (situation D). In principle this could happen two times in a row leading to an 138 undetectable error based on evaluation of the order of cage changes. E.g., a 139 mouse moves from the left to the right cage, then two cage changes are missed, 140 and the next seen cage change happened logically reasonable from the right to 141 the left cage. However, we could show that situation D (both antennas were 142 missed) is very unlikely and therefore, it is even more unlikely that this occurs 143 two times in a row. detect the mouse (A1 → A2). In this case, the recorded data will provide no 150 hint that the mouse has been in the other cage. However, we did not observe 151 this at all during evaluation and thus deem this situation to be very unlikely.

Customization of the Evaluation Script
Depending on the research question, the evaluation script can be freely cus-  Table 1: Output from evaluation script of the validation experiment (Experiment 1). Detection and cage change are defined as described at the beginning in the glossary. Percentages are calculated as part of the total cage changes each mouse made.