US 12,274,243 B2
Systems and methods of homecage monitoring
Edgar Josue Bermudez Contreras, Lethbridge (CA); Robert James Sutherland, Lethbridge (CA); Majid Mohajerani, Lethbridge (CA); and Surjeet Singh, Lethbridge (CA)
Assigned to Neurocage Systems Ltd., Lethbridge (CA)
Filed by Neurocage Systems Ltd., Lethbridge (CA)
Filed on Jun. 5, 2020, as Appl. No. 16/894,218.
Claims priority of provisional application 62/857,879, filed on Jun. 6, 2019.
Prior Publication US 2020/0383299 A1, Dec. 10, 2020
Int. Cl. A01K 29/00 (2006.01); G06F 18/2413 (2023.01); G06V 10/82 (2022.01); G06V 20/52 (2022.01); G06V 40/10 (2022.01); G06V 40/20 (2022.01)
CPC A01K 29/005 (2013.01) [G06V 10/82 (2022.01); G06V 20/52 (2022.01); G06V 40/10 (2022.01); G06V 40/20 (2022.01)] 14 Claims
OG exemplary drawing
 
1. An automated homecage monitoring system for determining at least a predicted pose and a predicted animal state of at least one animal in the homecage, the homecage system comprising:
a homecage for receiving the at least one animal, operably connected to at least one microcontroller, the at least one microcontroller having at least one sensor for monitoring and collecting data about the at least one animal as sensor input, one of the at least one sensors being a camera for sensing at least one video frame of the animal;
a memory having:
a first predictive model, the first predictive model for predicting a pose estimate of the at least one animal, the first predictive model being a deep neural network; and
a second predictive model, the second predictive model for predicting an animal state, the second predictive model being a recurrent neural network having nodes and transitions for processing the predicted pose estimate and sensor input to predict the animal state;
a processor in communication with the at least one sensor, the memory, and the at least one camera, the processor configured to, in real-time;
receive the sensor input from the at least one sensor and
generate, using the first predictive model, the predicted pose estimate from the sensor input, the predicted animal pose estimates representing a visual feature in the at least one video frame corresponding to a body feature of the animal;
generate, using the nodes and the transitions in the second predictive model, the predicted animal state based on the predicted pose estimate and the sensor input, the animal state including at least one of a behavioral state of the animal, a social state of the animal, a position state of the animal, a sleep state of the animal, and a biological state of the animal, the transitions being triggered by changes in the predicted pose estimate and sensor data to change the predicted animal state;
determine at least one husbandry variable, the at least one husbandry variable comprising a food supply level, a water supply level, a temperature, a humidity value, a bedding quality metric, and a nesting quality metric;
determine a report based on the at least one of the animal state, the pose estimate, and the sensor input, and
output the report to a server by network communication.