Systems and methods for monitoring and treating diabetic foot ulcers
In some aspects, a computer-implemented method is disclosed for gathering and processing sensor data to identify a risk of impacting or causing a skin injury. The computer-implemented method can include: receiving, via a computer network, sensor feature data representing output of a user sensor configured to be worn on a limb of a user; generating activity classification model output data using the sensor feature data and an activity classification model, the activity classification model output data representing likelihoods that the sensor feature data corresponds to each of a plurality of different activity classifications; determining an activity classification from the activity classification model output data; and transmitting, via the computer network, display data representing the activity classification to a computing device configured to present the display data.
1 . A system for gathering and processing sensor data to identify a risk of impacting or causing a skin injury, the system comprising computer-readable memory and one or more processors, the system being configured to:
obtain sensor feature data comprising a first set of feature data and a second set of feature data, the first set of feature data representing output of a user sensor worn on a limb of a user, the user sensor not coupled to an offloading device worn on the limb, the second set of feature data representing output of an offloading device sensor coupled to the offloading device;
determine an activity classification using the first set of feature data input into an activity classification model,
wherein the activity classification model is configured to determine a first degree of similarities between the first set of feature data and each of a plurality of different activity classifications, and wherein the activity classification is determined based on the first degree of similarities;
determine an offloading usage classification using the first set of feature data and the second set of feature data input into an offloading usage model,
wherein the offloading usage model is configured to determine a second degree of similarities between motion of the user sensor associated with the first set of feature data and motion of the offloading device sensor associated with the second set of feature data, and wherein the offloading usage classification is determined based on the second degree of similarities;
from a plurality of events represented in the sensor feature data, identify a set of events for which at least one of 1) the activity classification satisfies an acceleration threshold or 2) the activity classification satisfies a duration threshold and the offloading usage classification indicates that the offloading device was not used during an event;
generate display data comprising the set of events, the set of events associated with the risk of impacting or causing the skin injury on the limb; and
transmit the display data to a computing device and cause the computing device to present the display data.
2 . The system of claim 1 , wherein the skin injury comprises a diabetic foot ulcer or a venous leg ulcer.
3 . The system of claim 1 , wherein the user sensor comprises a first accelerometer and a first magnetometer, and the offloading device sensor comprises a second accelerometer and a second magnetometer, the first set of sensor feature data representing output from the first accelerometer and the first magnetometer, and the second set of sensor feature data representing output from the second accelerometer and the second magnetometer.
4 . The system of claim 1 , being further configured to:
determine a subsequent activity classification using a third set of feature data input into the activity classification model, the third set of feature data representing output of the user sensor; and
determine from the activity classification and the subsequent activity classification that a change in an activity of the user has occurred.
5 . The system of claim 1 , being further configured to:
determine a subsequent offloading usage classification using a third set of feature data and a fourth set of feature data input into the offloading usage model,
wherein the third set of feature data represents output of the user sensor, and the fourth set of feature data represents output of the offloading device sensor; and
determine from the offloading usage classification and the subsequent offloading usage classification that a change usage of the offloading device by the user has occurred.
6 . The system of claim 1 , being further configured to:
generate activity model input data using the first set of feature data;
generate activity classification model output data using the activity model input data input into the activity classification model; and
identify, from the activity classification model output data, a data element having a value that satisfies a classification criterion to determine the activity classification, the activity classification being associated with the data element.
7 . The system of claim 1 , wherein the plurality of different activity classifications comprises a motion classification, a no motion with static weight classification, a no motion with variations in force classification, and a laying down classification.
8 . The system of claim 1 , wherein the offloading usage model is configured to determine the second degree of similarities from the degree to which motion and orientation of the user sensor corresponds to motion and orientation of the offloading device sensor.
9 . The system of claim 1 , wherein the activity classification model comprises a decision tree, a neural network, or a support vector machine.
10 . The system of claim 1 , wherein the activity classification model comprises a first decision tree configured to assign the first set of feature data to one of the plurality of different activity classifications for the first degree of similarities, and the offloading usage model comprises a second decision tree configured to determine whether motion and orientation of the user sensor corresponds to motion and orientation of the offloading device sensor for the second degree of similarities.
11 . The system of claim 1 , wherein the limb comprises a leg of the user, and the user sensor is worn below a knee of the leg of the user.
12 . A computer-implemented method for gathering and processing sensor data to identify a risk of impacting or causing a skin injury, the computer-implemented method comprising:
receiving, via a computer network, sensor feature data representing output of a user sensor worn on a limb of a user, the user sensor not coupled to an offloading device worn on the limb;
generating, by one or more computer processors, activity classification model output data using the sensor feature data input into an activity classification model, the activity classification model output data representing likelihoods determined by the activity classification model that the sensor feature data corresponds to each of a plurality of different activity classifications,
wherein the plurality of different activity classifications comprises:
a first activity classification representing motion of the limb; and
a second activity classification representing no motion of the limb with weight loading the limb;
determining, by the one or more computer processors, an activity classification from the activity classification model output data associated with the risk of impacting or causing the skin injury on the limb; and
transmitting, via the computer network, display data comprising the activity classification to a computing device and causing the computing device to present the display data.
13 . A non-transitory computer readable medium having an application stored thereon for gathering and processing sensor data to identify a risk of impacting or causing a skin injury, the application, when executed by one or more computer processors, causing the one or more computer processors to:
receive, via a computer network, sensor feature data representing output of a user sensor worn on a limb of a user, the user sensor not coupled to an offloading device worn on the limb;
generate activity classification model output data using the sensor feature data input into an activity classification model, the activity classification model output data representing likelihoods determined by the activity classification model that the sensor feature data corresponds to each of a plurality of different activity classifications,
wherein the plurality of different activity classifications comprises:
a first activity classification representing motion of the limb; and
a second activity classification representing no motion of the limb with weight loading the limb;
determine an activity classification from the activity classification model output data associated with the risk of impacting or causing the skin injury on the limb; and
transmit, via the computer network, display data comprising the activity classification to a computing device and cause the computing device to present the display data.
14 . The non-transitory computer readable medium of claim 13 , wherein the plurality of different activity classifications comprises a third activity classification representing no motion of the limb with variations in force on the limb.
15 . The non-transitory computer readable medium of claim 13 , wherein the plurality of different activity classifications comprises a third activity classification representing the user is lying down.
16 . The non-transitory computer readable medium of claim 13 , wherein the activity classification model comprises a decision tree, a neural network, or a support vector machine.
17 . The non-transitory computer readable medium of claim 13 , wherein the activity classification model comprises a decision tree configured to assign the sensor feature data to one of the plurality of different activity classifications.
18 . The non-transitory computer readable medium of claim 13 , wherein the application, when executed by the one or more computer processors, further causes the one or more computer processors to:
determine a subsequent activity classification using second sensor feature data input into the activity classification model, the second sensor feature data representing output of the user sensor; and
determine from the activity classification and the subsequent activity classification that a change in an activity of the user has occurred.
19 . The non-transitory computer readable medium of claim 13 , wherein the application, when executed by the one or more computer processors, further causes the one or more computer processors to:
receive, via the computer network, second sensor feature data representing output of an offloading device sensor coupled to the offloading device; and
determine an offloading usage classification using the sensor feature data and the second sensor feature data input into an offloading usage model,
wherein the offloading usage model is configured to determine a degree to which motion of the user sensor corresponds to motion of the offloading device sensor.
20 . The non-transitory computer readable medium of claim 19 , wherein the offloading usage model is configured to determine the degree to which motion and orientation of the user sensor corresponds to motion and orientation of the offloading device sensor.
21 . The non-transitory computer readable medium of claim 19 , wherein the application, when executed by the one or more computer processors, further causes the one or more computer processors to:
receive, via the computer network, third sensor feature data representing output of the offloading device sensor coupled to the offloading device;
determine a subsequent offloading usage classification using third sensor feature data input into the offloading usage model, the third sensor feature data representing output of the offloading device sensor; and
determine from the offloading usage classification and the subsequent offloading usage classification that a change in usage of the offloading device by the user has occurred.