System and method for characterizing and monitoring health of an animal based on gait and postural movements

1 . A method for monitoring health of an animal, comprising:

accessing an image feed recorded during a health session by an optical sensor defining a field of view intersecting a working field;

detecting the animal in a first location within the working field, at a first time during the health session, in the image feed; and

in response to detecting movement of the animal from the first location toward a second location within the working field in the image feed during a first test period succeeding the first time and within the health session:

extracting a first timeseries of position data, representing position of a set of anatomical features of the animal during the first test period, from the image feed;

deriving a first movement profile for the animal based on the first timeseries of position data, the first movement profile representing movement of the animal during the first test period;

accessing a baseline movement profile defined for the animal;

characterizing a difference between the first movement profile and the baseline movement profile; and

in response to the difference exceeding a threshold difference:

interpreting an abnormality in the first movement profile for the animal;

predicting a causal pathway for the abnormality based on characteristics of the abnormality; and

selecting a mitigation protocol, in a set of mitigation protocols, for the animal based on the causal pathway.

2 . The method of claim 1 :

wherein accessing the image feed recorded by the optical sensor comprises accessing the image feed recorded by the optical sensor integrated into a training apparatus configured to dispense units of a primary reinforcer toward a range of locations within the working field via a dispenser integrated into the training apparatus;

further comprising, at a second time succeeding the first time, in response to detecting the animal at the first location in the image feed, dispensing a first unit of the primary reinforcer toward the second location within the working field, the second location intersecting a target pathway extending from the first location to the second location; and

wherein extracting the first timeseries of position data from the image feed in response to detecting movement of the animal from the first location toward the second location during the first test period succeeding the first time comprises extracting the first timeseries of position data from the image feed in response to detecting movement of the animal from the first location toward the second location during the first test period succeeding the second time.

3 . The method of claim 2 :

further comprising, at a third time preceding the first time, dispensing a second unit of the primary reinforcer toward the first location within the working field;

wherein detecting the animal at the first location at the first time comprises detecting the animal within a threshold distance of the first location at the first time succeeding the third time; and

wherein dispensing the first unit of the primary reinforcer toward the second location in response to detecting the animal at the first location comprises, in response to confirming consumption of the second unit of the primary reinforcer by the animal and in response to detecting the animal within the threshold distance of the first location, dispensing the first unit of the primary reinforcer toward the second location.

4 . The method of claim 1 :

wherein predicting the causal pathway for the abnormality based on characteristics of the abnormality comprises:

deriving a confidence score for the causal pathway based on characteristics of the abnormality; and

based on the confidence score, predicting the causal pathway; and

wherein selecting the first mitigation protocol based on the causal pathway comprises selecting the first mitigation protocol based on the causal pathway and the confidence score.

5 . The method of claim 4 :

wherein predicting the causal pathway based on the confidence score comprises, in response to the confidence score exceeding a first threshold confidence, predicting the causal pathway; and

further comprising, in response to the confidence score exceeding a second threshold confidence and falling below the first threshold confidence:

selecting a health analysis protocol configured for execution by the animal to evaluate the abnormality; and

in response to detecting execution of the health analysis protocol by the animal, within the working field, in the image feed during a second test period succeeding the first test period:

deriving a second timeseries of position data representing position of the set of anatomical features during execution of the health analysis protocol;

updating the first movement profile for the animal based on the second timeseries of position data;

characterizing a second difference between the first movement profile and the baseline movement profile;

characterizing a second confidence score for the causal pathway based on the second difference and characteristics of the abnormality; and

in response to the second confidence score exceeding the first threshold confidence, predicting the causal pathway for the abnormality.

6 . The method of claim 1 :

wherein detecting movement of the animal from the first location toward the second location comprises detecting movement of the animal from the first location toward the second location along a first target pathway defining a first orientation of the animal relative the optical sensor;

wherein deriving the first movement profile representing movement of the animal during the first test period comprises deriving the first movement profile representing movement of the animal along the first target pathway during the first test period; and

further comprising, during the health session:

detecting the animal in a third location within the working field, at a second time succeeding the first test period during the health session, in the image feed; and

in response to detecting movement of the animal from the third location toward a fourth location within the working field, along a second target pathway defining a second orientation of the animal relative the optical sensor, in the image feed during a second test period succeeding the second time during the health session:

extracting a second timeseries of position data, representing position of the set of anatomical features of the animal during the second test period, from the image feed;

deriving a second movement profile for the animal based on the second timeseries of position data, the second movement profile representing movement of the animal along the second target pathway during the second test period; and

deriving a composite movement profile for the animal for the health session based on the first movement profile and the second movement profile.

7 . The method of claim 1 :

wherein predicting the causal pathway for the abnormality based on characteristics of the abnormality comprises:

estimating a first confidence score for the causal pathway based on the difference and characteristics of the abnormality;

estimating a second confidence score for a second causal pathway based on the difference and characteristics of the abnormality; and

in response to the first confidence score exceeding the second confidence score, predicting the causal pathway for the abnormality; and

wherein selecting the mitigation protocol based on the causal pathway comprises selecting the mitigation protocol based on the causal pathway and the first confidence score.

8 . The method of claim 1 :

wherein interpreting the abnormality in the first movement profile comprises interpreting the abnormality for a first subset of anatomical features, in the set of anatomical features, represented in the first movement profile; and

wherein predicting the causal pathway based on characteristics of the abnormality comprises predicting the causal pathway based on the difference and the first subset of anatomical features.

9 . The method of claim 1 :

wherein extracting the first timeseries of position data comprises:

in the image feed, extracting a first set of positions of the set of anatomical features at a third location, arranged between the first location and the second location, at a second time during the first test period;

accessing a first time value corresponding to the second time;

in the image feed, extracting a second set of positions of the set of anatomical features at a fourth location, arranged between the third location and the second location, at a third time during the first test period;

accessing a second time value corresponding to the third time; and

generating the first timeseries of position data based on the first set of positions, the second set of position, the first time value, and the second time value; and

wherein deriving the first movement profile based on the first timeseries of position data comprises:

deriving a set of position changes of the set of anatomical features based on the first set of positions and the second set of positions; and

deriving the first movement profile based on the first time value, the second time value, and the set of position changes.

10 . The method of claim 1 :

wherein extracting the first timeseries of position data for the set of anatomical features of the animal comprises:

extracting a first position of a first anatomical feature, in the set of anatomical features, from a first image, in the image feed, recorded at a second time and depicting the animal at a third location between the first location and the second location within the working field;

extracting a second position of a second anatomical feature, in the set of anatomical features, from the first image in the image feed;

extracting a third position of the first anatomical feature from a second image, in the image feed, recorded at a third time, succeeding the second time, and depicting the animal at a fourth location between the third location and the second location within the working field; and

extracting a fourth position of the second anatomical feature from the second image in the image feed; and

wherein deriving the first movement profile for the animal based on the first timeseries of position data comprises:

deriving a first position change, in a set of position changes, for the first anatomical feature based on the first position and the third position;

deriving a second position change, in the set of position changes, for the second anatomical feature based on the second position and the fourth position;

calculating a time interval between the second time and the third time; and

deriving the first movement profile for the animal based on the set of position changes and the time interval.

11 . The method of claim 1 :

wherein accessing the image feed recorded by the optical sensor during the health session comprises accessing a video feed recorded by the optical sensor during the health session;

wherein detecting the animal in the first location and at the first time in the image feed comprises detecting the animal in the first location and at the first time in a first frame, in the video feed, captured at the first time; and

wherein extracting the first timeseries of position data from the image feed comprises extracting the first timeseries of position data from a sequence of frames, in the video feed, captured during the first test period.

12 . The method of claim 1 :

wherein detecting the animal in the first location and at the first time in the image feed comprises detecting the animal in the first location and at the first time in a first static image, in the image feed, captured at the first time; and

wherein extracting the first timeseries of position data from the image feed comprises extracting the first timeseries of position data from a sequence of static images, in the image feed, captured during the first test period.

13 . The method of claim 1 , wherein interpreting the abnormality in the first movement profile comprises:

deriving a health score for the animal based on the difference between the first movement profile and the baseline movement profile; and

in response to the health score falling below a threshold score, interpreting the abnormality in the first movement profile.