System and method for improving cardiovascular health of humans
An estimate of a functional capacity such as VO2Max is made by applying the vital signs of a monitored human to a trained encoding neural network producing a cardio profile vector. The vector is applied to a trained functional capacity (VO2Max) neural network to estimate the functional capacity. Once estimated, an action is taken.
1. A system for treating and improving the cardiopulmonary health of humans, the system comprising:
an encoding network configured to generate a cardio profile vector responsive to input of a duration of vital sign data;
a functional capacity estimation network configured to generate an estimate of a cardiorespiratory functional capacity of a person based upon an input of a cardio profile vector generated for that person; and
a control circuit
wherein the encoding network is trained using an excluded-variable estimation network configured to receive as input a duration of vital sign data and a cardio profile vector, and generate an estimate of at least one vital sign excluded from the input,
wherein the encoding network is trained to produce cardio profile vectors that are effective to enable the excluded-variable estimation network to make accurate estimates of the at least one vital sign excluded from input to the excluded-variable estimation network, by propagating excluded-variable estimation error as a cost function that is applied to and trains the encoding network;
and wherein said control circuit is configured to receive a duration of monitored data from a sensor worn by a monitored human subject, apply the monitored data to the encoding network after being trained to produce a current cardio profile vector, apply the current cardio profile vector to the functional capacity estimation network after being trained to produce a current estimate of functional capacity of the monitored human subject; and
based upon the current estimate of functional capacity, performing one or more actions selected from the group consisting of:
the control circuit displays the current estimate in a time series with prior estimates to a clinician for review of possible health changes in the monitored human subject;
the control circuit compares the current estimate to prior estimates and testing for a change whereupon an alert is triggered to a clinician to investigate the health of the monitored human subject;
the control circuit compares the current estimate to prior estimates and testing for a change whereupon a questionnaire is sent to the monitored human subject;
the control circuit compares the current estimate to prior estimates and testing for a change whereupon an entry is made in the medical record of the monitored human subject indicating a change in health of the monitored human subject occurred;
the control circuit compares the current estimate to estimates from other monitored human subjects, in order to quantify health affects between at least a control group receiving a first intervention and a test group receiving a second intervention in a clinical trial; and
the control circuit transmits an electronic control signal that controls the operation or setting a parameter of a medical device associated with treating or monitoring the monitored human subject.
2. The system of claim 1 , wherein the excluded-variable estimation network comprises a function generator network and an estimation network.
3. The system of claim 2 , wherein the estimation network receives a time series of a subset of selected cardiopulmonary variables, the subset omitting at least one excluded variable from the selected cardiopulmonary variables, and generates an estimate of the at least one excluded variable, wherein the estimation neural network is configured to use parameters received from the function generator network to structure the estimation neural network and generate the estimate.
4. The system of claim 1 , wherein the encoding network is trained by application of training data and in a plurality of iterations, and for each of the iterations and for each of a plurality of training human subjects,
a first duration of training data comprising a substantially contiguous time series of selected cardiopulmonary variables from a training human subject is input to the encoding network;
a second duration of the training data, substantially nonoverlapping in time with the first duration, comprising a substantially contiguous time series of the selected cardiopulmonary variables from the training human subject, provides input to the excluded-variable estimation network and wherein the excluded-variable estimation network provides the estimate of the at least one excluded variable.
5. The system of claim 1 , wherein the functional capacity estimation network comprises a feed forward neural network.
6. The system of claim 1 , wherein the functional capacity estimation network generates the current estimate of functional capacity using one of: a linear regression technique, a decision tree, a random forest estimation, or a gradient boosting model.
7. The system of claim 1 , wherein the functional capacity is VO2Max.
8. The system of claim 1 , wherein the vital signs comprise one or more of: a heart rate, a measure of physical activity or motion, a time domain heart rate variability, a respiration rate, a tilt angle, an A-fib probability, and a sample time difference.
9. The system of claim 1 , wherein the current cardio profile vector comprises a plurality of real numbers, each of the real numbers associated with one or more of the vital signs.
10. The system of claim 1 , wherein the encoding network is a neural network.
11. A method, the method comprising:
training an encoding network to produce cardio profile vectors, the cardio profile vectors being effective, after application to an excluded-variable estimation network, to enable the excluded-variable estimation network to make accurate estimates of vital signs being withheld from an input of the excluded-variable estimation network, the excluded-variable estimation network responsively propagating an error as a cost function that is applied to and trains the encoding network;
applying a duration of monitored data from a monitored human subject to the encoding neural network after being trained to produce a current cardio profile vector and applying the current cardio profile vector to a functional capacity estimation network that, after being trained, produces a current estimate of functional capacity of the monitored human subject;
based upon the current estimate of functional capacity, performing one or more actions selected from the group consisting of:
displaying the current estimate in a time series with prior estimates to a clinician for review of possible health changes in the monitored human subject;
comparing the current estimate to prior estimates and testing for a change whereupon an alert is triggered to a clinician to investigate the health of the monitored human subject;
comparing the current estimate to prior estimates and testing for a change whereupon a questionnaire is sent to the monitored human subject;
comparing the current estimate to prior estimates and testing for a change whereupon an entry is made in the medical record of the monitored human subject indicating a change in health of the monitored human subject occurred;
comparing the current estimate to estimates from other monitored human subjects, in order to quantify health affects between at least a control group receiving a first intervention and a test group receiving a second intervention in a clinical trial; and
controlling the operation or setting a parameter of a medical device associated with treating or monitoring the monitored human subject.
12. The method of claim 11 , wherein the excluded-variable estimation network comprises a function generator network and an estimation network.
13. The method of claim 12 , wherein the estimation network receives a time series of a subset of selected cardiopulmonary variables, the subset omitting at least one excluded variable from the selected cardiopulmonary variables, and generates an estimate of the at least one excluded variable, wherein the estimation neural network is configured to use parameters received from the function generator network to structure the estimation neural network and generate the estimate.
14. The method of claim 11 , wherein the encoding network is trained by application of training data and in a plurality of iterations, and for each of the iterations and for each of a plurality of training human subjects,
a first duration of training data comprising a substantially contiguous time series of selected cardiopulmonary variables from a training human subject is input to the encoding network;
a second duration of the training data, substantially nonoverlapping in time with the first duration, comprising a substantially contiguous time series of the selected cardiopulmonary variables from the training human subject, provides input to the excluded-variable estimation network and wherein the excluded-variable estimation network provides the estimate of the at least one excluded variable.
15. The method of claim 11 , wherein the functional capacity estimation network comprises a feed forward neural network.
16. The method of claim 11 , wherein the functional capacity estimation network generates the current estimate of functional capacity using one of: a linear regression technique, a decision tree, a random forest estimation, or a gradient boosting model.
17. The method of claim 11 , wherein the functional capacity is VO2Max.
18. The method of claim 11 , wherein the vital signs comprise one or more of: a heart rate, a measure of physical activity or motion, a time domain heart rate variability, a respiration rate, a tilt angle, an A-fib probability, and a sample time difference.
19. The method of claim 11 , wherein the current cardio profile vector comprises a plurality of real numbers, each of the real numbers associated with one or more of the vital signs.
20. The method of claim 11 , wherein the encoding network is a neural network.