Patent Application
20090312621
The present invention relates to a method and device for monitoring a physiological parameter.
About two hundred million people suffer from diabetes mellitus. Diabetes is a chronic condition in which the body does not properly regulate or utilize the insulin, which results in abnormal (too high/too low) levels of glucose in the blood. A tight glycemic monitoring is desirable for these patients. Presently, methods for observing the glucose levels require, for example, the finger-stick blood sample, which is further processed by a glucose meter. But observation alone is not sufficient for glycemic control. Any diabetic decision support system should be able to predict future values of the blood glucose with a time horizon of 30 to 60 minutes as accurately as possible. Within the predicted time the patient can then take the appropriate actions, as for example food intake or insulin injection, so that theses actions can become effective in time.
There are several problems which must be resolved by a prediction method. First of all, the dynamics of glucose-insulin interaction is rather complicated. Even continuous measurement of the blood glucose provides only part of the actual information about the state of the system. Secondly, fluctuations in blood glucose levels are intrinsically non-stationary and depend on external factors like food and insulin injections. Thirdly, even for the same patient, given the same food intake and insulin injections, the glucose dynamics can be different due to variation in a large number of parameters, as for example body temperature or physical activity, which are known to affect the glucose metabolic rates.
For example, in U.S. Pat. No. 6,923,763 B1 a prediction method is proposed based on a complex dynamic model that includes measured glucose values, heart rate and heart rate variability to try to estimate the blood glucose level in the future and estimate the risk of hypoglycemia. The approach is based on dynamic modeling of the glucose-regulation. One drawback of the prediction by dynamical models for glucose evolution is that the models are non-linear, complex and contain a large number of physiological parameters which cannot be measured.
It is therefor an object of the present invention to provide a method for monitoring a physiological parameter with a more reliable prediction of the evolution of the physiological parameter.
The above objective is accomplished by a method for monitoring a physiological parameter, in particular for monitoring a blood glucose level, comprising
The logging comprises preferably a non-invasive, continuous method of measuring the physiological parameter, e.g. for glucose measurements. Continuous in this sense means, that, if required, the measurements are performed about every 5 to 10 minutes, preferably automatically.
An interval of autonomous evolution in the sense of this invention is meant to be a time interval with a relatively simple evolution of the physiological parameter, e.g. the blood glucose level. Such intervals are primarily characterized by the absence of drastic external influences like food intake, insulin injections, physical exercises, etc. Within an interval of autonomous evolution one is able to use a parametric model for the physiological parameter, like glucose concentration. The person skilled in the art understands that the parametric model can be derived from statistical analysis of continuous traces of the evolution of the physiological parameter, which can be described well approximated over a limited time interval. An advantage of the inventive method is, that a limited number of parameters and models of reduced complexity can be used, despite the complexity of underlying processes.
Within an autonomous interval, the evolution of the physiological parameter is predictable by the following parametric model:
G(t)=Q(t; p1, . . . , pk),
where G(t) is the level of the physiological parameter at time t, which is predicted by the value Q at time t and with the parameters p1, . . . , pk. The parametric model is used to extrapolate, i.e., predict the future values of the physiological parameter, especially the blood glucose concentration.
For the interval of autonomous evolution, which comprises a plurality of previous measurements (at times tk-i, for i=1 . . . m), an optimal set of parameters or parameter vector {right arrow over (p)}k-1 according to the parametric model is preferably determined by the Least Squares Method:
The prediction according to the inventive method comprises the steps of
A new measurement (tk, G(tk)) and a multitude of previous measurements of the interval of autonomous evolution {(tk-i, G(tk-i))}, with i running from 1 to m, together form a new interval upon which the test is conducted. The person skilled in the art understands, that in this test, the new measurement is compared to the previously predicted evolution of the physiological parameter. The interval of autonomous evolution is then adapted accordingly. For example the interval of autonomous evolution is narrowed such that the parametric model is able to predict the evolution of the physiological parameter, i.e. to predict the new measurement. The person skilled in the art also understands that the modified interval of autonomous evolution comprises at least the new measurement. When the interval of autonomous evolution is adapted, the parametric model is preferably adapted as well. The adapted parametric model is then used for the prediction of the future evolution of the physiological parameter.
The future evolution, using the optimal set of parameters, is predicted by the parametric model:
Ĝ(tk+Δt)=Q(tk+Δt; {right arrow over (p)}k),
where Δt is the desired time horizon of 30 to 60 minutes.
Advantageously, the inventive parametric model, in principle, is not physical, i.e. the parametric model is primarily selected on the basis of adequate description of glucose traces, and not necessarily using complex models describing the full glucose regulatory system. Nevertheless, incorporation of at least parts of physical models is not excluded by the present invention. It is beneficial to use non-physical models, because they are simpler and contain fewer or no unknown parameters. For a complex non-linear dynamical system, which does not admit an analytic solution, proper estimation of the unknown parameters is much more difficult, if not infeasible. This can be done for type 1 diabetes, which is easier to model than type 2 or type 3 diabetes. The inventive method is advantageously applicable to glucose monitoring and prediction in general, and does not assume a particular type of diabetes.
The interval of autonomous evolution is preferably adapted by removing at least one measurement of the plurality of measurements of the interval of autonomous evolution. The person skilled in the art understands, that the oldest measurements of the interval of autonomous evolution are preferably removed first. The advantage of this method is, that an interval of autonomous evolution is not terminated, once the test is negative, but prediction is continued with a modified, smaller interval of autonomous evolution.
In a preferred embodiment of the present invention the test is conducted again, now testing whether the new measurement together with the measurements of the adapted interval form an interval of autonomous evolution. Furthermore preferred, the steps of removing at least one measurement and conducting the test is repeated until the adapted interval is an interval of autonomous evolution.
A preferred way of obtaining a new interval of autonomous evolution is, to repeatedly remove only the oldest measurement and then conduct the test, so that the longest possible, new interval of autonomous evolution can be determined. If, however, the plurality of measurements fails the test until only the new measurement is left, no interval of autonomous evolution can be determined. According to the inventive method, a new interval of autonomous evolution can be determined earliest after the next measurement.
A number of quantitative characteristics can be used to evaluate the quality of the new interval of autonomous evolution. Preferably, in the test an error is calculated by comparing the physiological parameter of the new measurement to a predicted physiological parameter of a preceding iteration:
εk=|G(tk)−Q(tk; {right arrow over (p)}k-1)|
If the error does not exceed a predefined threshold, the interval of autonomous evolution is extended by adding a new observation value.
In another preferred embodiment, a new set of parameters for the parametric model is determined, wherein the parametric model with the new set of parameters optimally describes the evolution of the physiological parameter over the interval of autonomous evolution including the new measurement.
The new set of parameters is preferably determined after the Least Squares Method:
Then, a distance between a new vector, defined by the new set of parameters and a former vector, defined by the set of parameters of the preceding iteration is calculated:
Again, if the distance does not exceed a predefined threshold value, the new measurement is added and the current interval of autonomous evolution is extended.
If the measurement is rejected in the test, the current interval of autonomous evolution is adapted. Most advantageously the new measurement can be added to the plurality of measurements forming the interval of autonomous evolution on the right, and the oldest measurements can be removed from the left until a smaller plurality of measurements meets the criterion of autonomous evolution.
In a preferred embodiment of the inventive method, a step of alarming a patient in case the predicted evolution of the physiological parameter is below or above a dangerous level. Advantageously, a patient suffering from diabetes can be alarmed early enough to prevent hypoglycemia or hyperglycemia.
Alternatively or additionally, an advice for the patient is provided, suggesting an agent or an amount of an agent to be administered, in particular an amount of insulin. The inventive system advantageously provides a decision support to the patient.
In a further preferred embodiment, a likelihood value is associated to the predicted evolution of the physiological parameter, the prediction of the evolution of the physiological parameter being discarded if the likelihood value is below a certain, predefined threshold value. The likelihood value advantageously reflects the trust in the accuracy of the prediction. For example, if the interval of autonomous evolution used for the prediction of glucose concentration is too small (too short), or, the interval was just adapted, this reduces the likelihood value. On the other hand, if the interval of autonomous evolution is relatively long and the set of optimal parameters has not changed drastically in a few last iterations, the likelihood value is higher. An advantage of the embodiment is to avoid prediction at all if the likelihood value is too small. The primary goal of the inventive monitoring and prediction of, for example, glucose concentrations is to prevent hypoglycemic and hyperglycemic events, i.e. events where glucose levels become dangerously low or dangerously high. These events typically occur at the end of long intervals of autonomous evolution. Hence, according to the present invention such predictions will have high likelihood values. Advantageously, extremely low or high glucose predictions cause an alarm for the patient and call for an appropriate action and a large number of false alarms is prevented.
An important aspect of the prediction is the stability of the results. Variations in the predicted evolution, which are faster than typical variations of glucose concentrations, should be avoided. In a preferred embodiment of the invention, the step of predicting the evolution of the physiological parameter is conducted using a multitude of parametric models for the adapted interval of autonomous evolution and obtaining the prediction as a weighted sum of the predictions of the multitude of parametric models:
wherein the weights wj of the parametric models are selected adaptively, preferably depending on a quality of the prediction of each parametric model in the previous iterations and/or depending on the number of iterations each parametric model has been used. For example, the weight wj for the j-th parametric model reflects the quality of the prediction of the j-th model demonstrated so far and/or the time period the model during which the model has been used. If the weight of a particular parametric model drops below a certain predefined threshold value the parametric model is preferably discarded (or its weight is set to zero) and a substitute parametric model is added to the multitude of parametric models. One of the advantages of this embodiment is to increase the stability of the prediction. Machine Learning theory provides methods for collecting predictions from several models (“experts”) in one super-model (“super-expert”), which can outperform each of the individual “experts”.
Preferably a number of threshold values, particularly the threshold value for the weight of the parametric models and/or the threshold for the likelihood value are adapted. Advantageously, the variability between different individuals is met by this embodiment. Again, Machine Learning techniques may advantageously be used, to adapt the threshold values.
In a further preferred embodiment the inventive method comprises the step of adapting the interval of autonomous evolution and/or the parametric model depending on external parameters known to be affecting the physiological parameter. The external parameters are preferably associated to ingestion, administration of insulin, physical activity and/or emotional stress of the patient. These major factors, which influence the intervals of autonomous evolution in glucose dynamics, are preferably detected and the parametric model is adapted appropriately. This advantageously enhances the accuracy of the predicted evolution, as the intervals of autonomous evolution can be updated earlier, for example.
In a still further preferred embodiment, the external parameters are manually inputted by a patient or by an auxiliary. For example, a device can be equipped with “food”, “insulin”, “exercise” buttons, which will advantageously provide information on forthcoming changes in the interval of autonomous evolution. Alternatively calorie input forms, food lists and/or databases can be used.
According to another preferred embodiment at least one of the external parameters is detected by at least one sensor. Sensors could be heart rate sensors, heart rate variability sensors, insulin pump sensors or food intake sensors. The reliability of the prediction is advantageously enhanced, as it does not depend on the accurateness of the input of the patient, who may forget to give or give wrong input on food intake, exercise or amounts of injected insulin.
In a preferred embodiment the physical activity is determined by an accelerometer, more preferably by an accelerometer summarizing the acceleration in three directions in space. In a further preferred embodiment the physical activity is determined by analyzing a heart rate of a patient by a heart rate sensor. With relation to exercise, these embodiments advantageously provide relatively accurate readings for a parameter which is difficult to quantify for a patient.
Another aspect of the present invention is a monitoring device for monitoring a physiological parameter of a patient, in particular for monitoring a blood glucose level, comprising a physiological parameter measuring means and a computing means, the computing means having a data storage for storing measurements and a processor unit for calculating a prediction of the evolution of the physiological parameter, using the method according to the present invention. The monitoring device advantageously provides reliable prediction of the evolution of the physiological parameter.
The monitoring device preferably comprises input means for inputting external parameters which are known to affect the physiological parameter, such as ingestion, administration of insulin, physical activity and/or emotional stress of the patient. The input means preferably comprises at least a button for manual input of each external parameter. In this embodiment the quality of the predicted evolution is enhanced, as information on forthcoming changes in the interval of autonomous evolution is provided by the patient.
Preferably the monitoring device comprises output means for outputting alarm messages and/or advice messages to the patient. The inventive monitoring device thus provides a (diabetic) decision support system with the ability to predict future evolution of the physiological parameter, especially the blood glucose level, with a time horizon of 30 to 60 minutes.
Still another aspect of the present application is a monitoring device for a physiological parameter of a patient, in particular for monitoring a blood glucose level, comprising a physiological parameter measuring means and a computing means, the computing means having a data storage for storing the blood glucose measurements and a processor unit for calculating a prediction of the evolution of the physiological parameter, the monitoring device being connected to a detector for detecting a physical activity of the patient, preferably a three-axial accelerometer.
It is known that food intake, insulin uptake and exercise influence the blood glucose level most. The reliability of the prediction of the monitoring device usually depends on the accurateness of the input of the user. If he or she forgets to give or gives wrong input on food intake, exercise or amounts of injected insulin the model/device will give false predictions. With relation to exercise it is difficult anyhow to quantify the input. The monitoring device according to the present invention overcomes the abovementioned disadvantages by providing a monitoring device that comprises an detector for detecting a physical activity of the patient. This activity sensor senses the motions of a body and therewith provides a continuous input with relation to the physical activity of a person suffering from diabetes. In this way the input with regard to the exercise level is advantageously reliable.
The detector is preferably wearable by the patient, more preferably the detector is arranged in or at a belt, a watch or a mobile phone. This makes it easy to wear the detector all day long, which advantageously provides objective and detailed information on the duration and intensity of the activity of the patient (sleep, rest, sitting, exercise etc).
Preferably, the monitoring device in combination with the detector is initialized once to learn the specific reaction of a patient's activity on his or her blood glucose level, also in relation to prior food- and insulin-intake. Preferably, a first time usage requires an individual initialization of the model. For example, the user wears the detector together with the monitoring device and performs a set of standardized activities of different intensities. The prediction of the evolution of the physiological parameter is adapted during this initialization phase on an individual basis of the response of the different activity levels on the physiological parameter, in particular the blood glucose level. This information is used to update the parameter prediction model. An existing blood glucose level prediction device can be extended with the detector to improve the reliability of the blood glucose level evolution prediction.
These and other characteristics, features and advantages of the present invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, which illustrate, by way of example, the principles of the invention. The description is given for the sake of example only, without limiting the scope of the invention. The reference figures quoted below refer to the attached drawings.
The present invention will be described with respect to particular embodiments and with reference to certain drawings but the invention is not limited thereto but only by the claims. The drawings described are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes.
Where an indefinite or definite article is used when referring to a singular noun, e.g. “a”, “an”, “the”, this includes a plural of that noun unless something else is specifically stated.
Furthermore, the terms first, second, third and the like in the description and in the claims are used for distinguishing between similar elements and not necessarily for describing a sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments of the invention described herein are capable of operation in other sequences than described of illustrated herein.
Moreover, the terms top, bottom, over, under and the like in the description and the claims are used for descriptive purposes and not necessarily for describing relative positions. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments of the invention described herein are capable of operation in other orientations than described or illustrated herein.
It is to be noticed that the term “comprising”, used in the present description and claims, should not be interpreted as being restricted to the means listed thereafter; it does not exclude other elements or steps. Thus, the scope of the expression “a device comprising means A and B” should not be limited to devices consisting only of components A and B. It means that with respect to the present invention, the only relevant components of the device are A and B.
Fundamental work on glucoregulatory systems demonstrate without a doubt a great variation of possible dynamic evolutions of the blood glucose level 1, one of which is exemplarily depicted in
In
In
According to an embodiment of the invention, the parametric model 3 is additionally adapted depending upon external parameters 5, which are derived from an input means 50 and/or from a detector 6. The input means 50 is preferably equipped with buttons 51, 52, 53, which are pressed by a patient upon certain activities in order to provide information on forthcoming changes in the interval of autonomous evolution to the monitoring device. For example button 51 is used for ingestion, button 52 for insulin administration and button 53 for physical exercises. Detector 6 preferably detects physical activity of the patient automatically, for example by way of a three-axial accelerometer worn by the patient.
The present invention is related to a method and device for monitoring a physiological parameter, like the blood glucose level, using prediction of future evolution of the physiological parameter based on continuous traces. The present method and device can be employed as a decision support system for diabetic patients.
| Number | Date | Country | Kind |
|---|---|---|---|
| 06118605.2 | Aug 2006 | EP | regional |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/IB07/52971 | 7/26/2007 | WO | 00 | 2/3/2009 |
