Project Summary/Abstract
Zebrafish, involved in research projects totalling half a billion dollars of NIH funding in 2019, are the fastest
growing animal model of human disease in medical research today. Thus, the health and wellbeing of medical
research fish is paramount as the quality and robustness of experimental data depends on it. Innovation in
zebrafish husbandry systems is crucially needed because current state-of-the-art non intrusive systems only
track water quality measures. No commercial husbandry system has adopted automatic monitoring of fish
behavior using video, despite the fact that most factors crucial in assessing fish health and well being, such as
hunger, anxiety, or light changes, produce well known behavioral signatures that a computer vision system
could detect. Because many such signatures require high temporal resolution, millisecond in the case of
escape movements, video monitoring has been hampered by the need to process enormous data volumes.
Here, the proposed effort aims at building a proof-of-concept system to video monitor a network of 10 custom
tanks, codenamed “canaryTanks”, containing sentinel zebrafish populations strategically located in the racks of
an existing fish facility. Each tank will be equipped with a custom illumination system and a camera that will use
an original "Remanent Imaging" paradigm to capture and process orders of magnitude less data than standard
video while preserving transient, millisecond-scale motion signatures. Threats to fish health—none of which
are tracked by current commercial systems—will be detected through the real time analysis of motion
signatures in the image streams captured by the canaryTanks and will be remotely accessible via browser
through our cloud-based software infrastructure. The first aim is to show that such a system can generate
timely and accurate daily reports of fish activity and sleep (actimetry), feeding times, and changes in room
illumination based solely on processing the behavioral data in the images captured continuously by the
canaryTanks. The second aim is to demonstrate the system’s ability to generate real-time alert notifications
using custom algorithms that automatically detect a range of zebrafish behavioral signatures. A battery of tests,
including disruption of the light cycle, exposure to noise/vibration and gradual increase to out-of-range
temperatures, will validate this proof-of-concept. The future vision is that the canaryTank technology sets new
standards in fish husbandry and provides revolutionary, ubiquitous in-tank recording capabilities that can be
leveraged beyond husbandry for collection of experimental data.