WO2014180941A1

WO2014180941A1 - Method and device for non-invasive real-time control of inner body temperature variables during therapeutic cooling or heating

Info

Publication number: WO2014180941A1
Application number: PCT/EP2014/059421
Authority: WO
Inventors: Aleksandra RASHKOVSKA; Roman TROBEC
Original assignee: Jozef Stefan Institute
Priority date: 2013-05-08
Filing date: 2014-05-08
Publication date: 2014-11-13
Also published as: SI24357A

Abstract

A method and device for non-invasive real-time control of inner body temperature variables during therapeutic cooling or heating using computer simulations, machine learning and control techniques. The invention refers to a method and device that enable control of inner body temperatures according to different therapeutic protocols. The technical problem that the invention solves is the control of inner body temperatures (hidden temperatures) that are difficult or impossible to be measured. In that context, a predictive model is used to estimate (predict) the values of the controlled inner temperature variables based on smaller number of other variables whose measurement is more feasible, i.e. temperatures on the body surface. However, simulations are usually resource and time consuming. The predictive model is constructed using advanced methods for data analytics to capture the correlation between the hidden variable and the measurable ones in data resulting from preliminary computer simulation of the system for different input simulation parameters.

Description

METHOD AND DEVICE FOR NON-INVASIVE REAL-TIME CONTROL OF INNER BODY TEMPERATURE VARIABLES DURING THERAPEUTIC COOLING OR HEATING

The subject of the invention is non-invasive real-time control of inner (hidden) body temperature variables during therapeutic cooling or warming, which preferably uses cross-fertilization of computer simulations, machine learning and control techniques. The invention does not refer to a therapeutic method, but to a method and device which enable control of inner temperatures according to different therapeutic protocols.

Background

Therapeutic cooling or warming of body parts, e.g., head, knee, elbow, etc., is a medical treatment that adjusts patients' body part temperature to help reduce tissue oedema, inflammatory reaction, hematoma formation and pain, reducing the need for pain medications and enabling faster rehabilitation after injury or surgery. It is routinely used in postoperative treatment in orthopaedics, traumatology, surgery, pain prevention in sport, etc., and is generally a topical treatment. Patients' responses to the treatment can differ considerably. Different patients need different therapeutic protocols, depending on their body constitution, regulatory systems and on the conditions in the environment where the therapy is performed. Consequently, the heat therapy should be controlled depending on the individual patient's response and the conditions of the environment where the therapy is performed, which raises the need for "smart" cooling devices for personalized therapy. The patient's response is different primarily because of various changes in the blood flow in the tissue, which significantly influences the temperature condition of the whole body, including the inner parts of the body. The influence of the environment is mainly due to the ambient temperature and the way how the body part that is cooled or heated is exposed to that influence. For non-invasive control of inner body temperature that cannot be measured, we require a method and device which enable control of the internal temperature based on temperatures on the body surface whose measurement is more feasible.

Most often control systems are based on the principle of feedback whereby the physical quantity to be controlled is compared to a desired reference physical quantity and the discrepancy is used to compute corrective control action. The key in feedback control is measuring the variable being controlled to compare it with a target value. However, measurements in biomedicine are often difficult to perform because human subjects are involved. Particularly in clinical procedures, in-vivo measurements are often not as accurate as desired, difficult, dangerous, or even impossible to perform, especially if deep tissues or vital organs are in question. Moreover, non-invasive medical procedures are emerging in the search for more reliable, less expensive, and risk free medical technology for the future. The technical problem that the invention solves is the case when inner body temperatures that have to be controlled for improved therapeutic treatment, are difficult or impossible to be measured.

State of the art

Therapeutic cooling is often performed using either ice chips or refrigerated gel- packs, or a cryotherapeutic cuff cooled by a liquid with a specified flow rate and temperature maintained by an external cooling device or by a container with ice water. Therapeutic heating is performed analogously, but at a different temperature. The shape and type of the cooling/heating cuff, and the temperature and/or flow rate of the cooling/heating liquid, can influence the amount of extracted/added energy from/to the treated part of the body.

Different devices for therapeutic cooling are already known. Most are based on a circulating cooling liquid in a pad/cuff and incorporate temperature, flow, or pressure control of the cooling liquid, but not for means of controlling the treatment depending on the patient's response. They are flexible and adaptive to the cooled portion of the body, as presented in the patent US4335726 - Therapeutic device with temperature and pressure control, 1982, or fitted around a joint for applying therapeutic temperature regulated compression to the joint and providing support for the limb, as in the patent US7744551 - Temperature regulated compression brace, 2010.

Some form of temperature control using feedback from the temperature of the outgoing circulating liquid has been subject of the patents US5097829 - Temperature controlled cooling system, 1992, and US5241951 - Therapeutic non-ambient temperature fluid circulation system, 1993. In the patent US5097829, the subject of the invention is an improved temperature control fluid circulating system for automatically cooling that uses temperature control fluid in a thermal blanket. The thermoelectric cooling device has a cold side and a hot side. The temperature control fluid is cooled by the cold side of the cooling device and its temperature is modified depending on the sensed temperatures of the fluid flowing within the blanket. Similar, in the patent US524195, a device therapeutically treating a desired region of a patient's body uses a non-ambient temperature fluid circulating through a pad having a tortuous fluid pathway positioned on the treatment region. The device has fluid inlet and outlet lines, each having an end connected to the pad and an opposite end positioned in a reservoir containing the non-ambient temperature fluid. Temperature control of the pad is enabled by an in-line valve and temperature monitor positioned in the outlet line.

Other therapeutic devices control the cooling based on the measured temperature of the cooled body surface. In the patent US5634940 - Therapeutic structure and methods, 1997, the therapeutic device includes: a shield adapted to be placed in intimate surface to surface contact with the body portion to be cooled, a regulator for passing a fluid through the shield at a regulated temperature and/or flow rate, a control for the flow and temperature of the fluid, and a temperature sensor for modifying the temperature and flow of the fluid in response to the monitored temperature of the body portion. A similar apparatus for controlling the temperature of an area of the body is the subject of the patent US5190032 - Apparatus for controlling the temperature of an area of the body, 1993.

Finally, invasive control therapeutic methods and devices are also known. In the patent CN2242696 - Automatic cooling device for automatic circulation blood therapeutic machine, 1996, the automatic cooling device is an automatic blood circulation therapeutic machine. Blood circulation is kept under appropriate environmental temperature, and auto chemotherapy is ensured when the blood is transfused back. Another invasive therapeutic treatment has been subject of the patent US8061 149 - Temperature control case, 201 1 , where embodiments of a temperature control case and related methods and systems are disclosed. Some embodiments include both internal and external temperature sensors to allow the device to predict internal temperature changes and adjust heating/cooling elements accordingly before the internal temperature is unduly impacted by the ambient temperature.

However, none of the above mentioned therapeutic devices incorporates a method for non-invasive real-time control of inner temperatures during the therapeutic treatment. To the best of our knowledge, a concrete technical solution for therapeutic cooling or warming that incorporates controlling of inner temperature variables whose measurement is not feasible or would introduce invasiveness into the system, is not known so far. Moreover, the use of simulation or predictive modelling, or the combination of the two, for estimation of controlled inner temperature variables in therapeutic control systems, has not yet been known. The subject of our invention is a new method and device for non-invasive real-time control of inner temperature variables during heat therapy. The method uses predictive models build upon results from pre simulated therapy.

Computer simulations can provide safe and inexpensive insight into physiological processes. In recent decades, computer simulations have proved a great help in understanding and solving a variety of problems in science. Advances in computer technology enable simulations of natural phenomena that cannot be subject to experiments in reality because of ecological, hazardous or financial obstructions. With the use of computer simulation, it is possible to calculate, analyse, and visualize both stationary and time dependent temperature fields in living biological tissues. The temperature of human tissue and organs is an important factor, which influences the therapeutic approaches in physiology, sports, cryotherapy, and many other fields. Preliminary, a large number of simulations scenarios of the therapeutic procedure are performed to obtain a data set that describes the operation of the system under various conditions. The data set includes values of inner and outer system variables. Appling methods for data analytics on the data set, such as machine learning, we can build a predictive model that estimates the value of inner variable from the values of outer variables that can be measured. The predicted value of the inner variable can be now used to control that variable in a classical feedback control system.

Summary of the Invention

Aspects of the present invention provide a device for non-invasive real-time control of an inner body temperature variable during therapeutic cooling or heating and a method for non-invasive real-time control of an inner body temperature variable during therapeutic cooling or heating as set out in the independent claims.

Optional and preferred features of these aspects are set out in the dependent claims.

In the context of the invention, preliminary computer simulation of the system (in our case - the therapeutic treatment) can be used for estimation of the controlled inner temperature variable based on other system variables whose measurement is more feasible. In the case of biomedical systems, if the controlled output cannot be measured because of the non-invasive nature of the system, then we should measure other non-invasive variables that can be used to estimate the output, for example temperatures on the body surface or temperature gradients between the cooling pad and the skin. However, simulations are usually resource and time consuming which does not go in the favour of real-time control systems. Real-time control, especially in biomedical systems, requires a solution usually supported by a mini on-board computer that cannot process large amount of data nor perform computationally demanding operations.

This can avoid the need for expensive real-time monitoring using MRI or CT imaging for temperature guidance, or the need for invasive measurements of actual inner body temperature variables. Therefore, machine learning methods are preferably used to construct a predictive model that provides a prediction for the inner temperature variable (output) in a far shorter time with satisfying accuracy. Namely, we use simulation to generate a substantial amount of data for different input simulation parameters for the task of machine learning: capture the correlation between the inner non-measurable variables and the outer non-invasive measurable ones. There are two main advantages of using predictive models: a) the predictive model can be used to simplify the simulation model and elucidate the most important correlations between the measurable (input) variables and non-measurable (output) variables, and b) the predictive models are typically more efficient in terms of memory and computational complexity.

The task that we are solving belongs to the predictive modelling setting, where the goal is to predict the value of one property of the examples, called a target attribute or output, using the values of the remaining properties, called descriptive or input attributes. Furthermore, machine learning can incorporate also methods for feature ranking and feature selection. The feature ranking methods provide an ordered list of the input variables by their relevance or importance for the output/target variable. The feature selection methods provide a subset of the input variables that are the most relevant for the output variable. These two methods are closely connected: feature selection can be performed by taking the top-k ranked features as provided by the feature ranking method. The invention uses these methods in order to investigate how many input variables are sufficient for making a good prediction and to which extent we need to measure them in time. The benefit is, for example, reduction of the number of sensors for measurements of input variables and construction of even smaller and more efficient predictive models.

The invention is presented in details with the figures, which show:

Figure 1 : Schematic diagram of the construction of the predictive model. Figure 2: Schematic diagram of the classical feedback control system construction and the invented system for non-invasive real-time control of inner body temperature variables.

Figure 3: Schematic diagram of an industrial application of the invented method and device.

Detailed description of the invention

According to Figure 1 , by performing preliminary simulations of the system with different input simulation parameters, we generate a large amount of simulation results E for different values of the input parameters P. The simulated data contain values for a large number of inner and outer system variables. We analyze the simulated results and describe the system with a smaller set of significant variables. Using methods for advanced data analytics, like machine learning, we build a predictive model D that captures the correlation between the inner variables that cannot be measured and the outer variables that can be measured with sensors. We select a subset of the measured variables that performs the best in predicting the value of the controlled inner variable. The predicted value of the controlled variable with satisfying accuracy replaces the value of the input variable in the control loop. In general, there can be more controlled inner variable.

In Figure 2, the classical feedback control loop is shown with dashed lines. The feedback loop of the invention is shown with solid lines. The basic feedback control loops has three components: a plant B to refer to the main process to be controlled, a sensor C to measure the output of the plant B, i.e. the controlled output variable 4, and a controller A that takes the difference 2 between the reference 1 and the measured output 5 to change the inputs 3 to the plant B and consequently control the output behaviour of the process. The classical feedback control loop with sensor C for the inner controlled variable is replaced with the components of the invention: sensors C1 to obtain non-invasively measured values 7 of the outer variables (for example, temperatures), and predictive model D that gives predicted output 8, i.e. estimated value of the controlled inner (hidden) variable. The controller A, based on the difference 2 between the reference value 1 and the predicted value 8 of the inner controlled variable, controls the inputs 3 to the plant B and consequently controls the output behaviour of the process.

An example of industrial application is shown in Figure 3. If the controlled output variable cannot be measured, for example, the temperature inside a knee, the solution that is the subject of the invention is to add a predictive model D to the feedback control loop to obtain estimation of the inner knee temperature 9 instead of the actual inner knee temperature that cannot be measured non-invasively. The predicted inner knee temperature 9 is estimated based on the measured values 6 of other outer system/process variables whose measurement is more feasible. Sensors C1 are added for non-invasive measurement of the outer variables 6 (for example, temperature on the body surface). Preliminary, simulation model of the system/process is used to provide a reliable data for different input simulation parameters to build the predictive model D using machine learning methods. The controller A, based on the difference 2 between the desired value 10 and the predicted value 9 of the inner knee temperature, controls the cooling temperature in the pad, representing the cooling system 3a, wrapped around the knee, representing the plant B.

The methods for computer simulation and construction of the predictive model are already known and are therefore not subject of the invention. However, the type of selected methods may affect the accuracy and stability of the control process. The sensors for measuring outer variables and their location are not explicitly defined in the invention. The claims of the invention are not limited or dependent of the above mentioned options.

Example of industrial application

A practical application of the invention on a problem in the domain of biomedicine - real-time control of the inner knee temperature during cryotherapeutic treatment after anterior cruciate ligament (ACL) reconstructive surgery of a knee is shown in Figure 3. It is evident from clinical praxis that patients' responses on the topical knee cooling differ significantly regarding inner knee temperatures. Assuming that the inner knee temperature should be kept constant and equal in all patients, the cooling methodology has to be able to adapt to different patients who need different cooling protocols, depending on their body constitution, regulatory systems and on the environmental conditions. A "smart" cooling device has to adjust to any individual patient's response and different therapeutic protocols.

The method for control of the inner knee temperature is based on computer controlled cryotherapy with pre-programmed protocols in terms of heat extraction intensity and treatment time. The goal is automatic control of the temperature inside the knee, representing the plant B, by changing the cooling temperature in the pad, representing the cooling system 3a, as input 3. The cooling temperature of the pad is changed according to the output of the controller A. The controlled inner knee temperature is hidden variable since we aim to achieve non-invasiveness of the biomedical system. Therefore, a few non-invasive temperature sensors C1 on the knee surface measure the values 7 of outer variables, like the value 6 of the temperature on the skin or the actual heat flux between the knee and the cooling pad, and provide physiological response of the patient as a feedback for the control process. The sufficient number of sensors is determined using machine learning methods. To control the deep knee temperature successfully in an arbitrary knee without measuring it, we need to predict the value 8 of the inner variable, i.e. the value 9 of the deep temperature, from the non-invasively measured outer variables 7, i.e. the temperatures 6 on the knee skin, using predictive model D obtained by knowledge extraction from the simulated data E. The light-weighted predictive models can be now implemented on a low performance computer that controls the cryotherapy. Computer simulations of heat transfer in biological tissues, using spatial models of human knee, will form the ground base for providing the simulated data E.

The described method and device for control of hidden variables is general and applicable in any system based on control of inner temperatures in any part of the body and therefore does not limit the scope of the presented invention to the presented application example. A different choice of simulation mode or predictive model construction does not affect the basic idea for control of variables that are difficult or impossible to be measured based on outer variables whose measurement is more feasible. Also, the cooling/heating method or the location and selection of sensors for measuring the outer variables do not limit the scope of the presented invention.

An exemplary embodiment of the invention provides a device for non-invasive realtime control of inner body temperature variables during therapeutic cooling or heating, consisted of:

- a pad with controlled temperature, representing the cooling/heating system (3a), wrapped around the treated body part, representing the plant (B);

- at least one sensor (C1 ), placed on the surface of the treated body part, which measures outer variables non-invasively (7).

- predictive model (D) for estimating the value of the controlled inner variable (8);

- controller (A) that takes the difference (2) between the reference value (1 ) and the predicted value (8) of the inner variable to controls the temperature in the cooling/heating system (3a) and change the inputs (3) to the plant B and consequently control the output behaviour of the process.

The above device may also have the feature that it uses machine learning methods to construct a predictive model, based on the results (E) from computer simulation of the therapeutic process for different simulation parameters (P), which estimates the value of the controlled inner temperature variable (8) from other variables whose measurement is more feasible (7).

A further exemplary embodiment provides a method used by the device for noninvasive real-time control of inner body temperature variables during therapeutic cooling or heating, with the feature, that it is based on feedback control loop that uses predictive model (D) to estimate the value of the controlled inner variable (8) from other variables (7) measured with sensors (C1 ) placed on the body surface, using the correlation between the measurable and non-measurable variables in the results (E) from computer simulation of the therapeutic process for different input simulation parameters (P).

The above method may also have the feature that the machine learning method uses data (E) generated from computer simulation of the cooling or heating therapy for different input simulation parameters (P), e.g. input signals, initial and boundary conditions, or any combination of them.

Furthermore, the above method may have the feature that the predictive model (D) is built upon data for other variables of the process (B), and their correlation obtained from the simulation results (E) of the system, or in a broader sense, the predictive model is built with advanced data analysis of the simulation results (E).

Claims

1 . A device for non-invasive real-time control of an inner body temperature variable during therapeutic cooling or heating, comprising:

- a temperature controlled pad (3a), representing the cooling/heating system (3a), arranged to be wrapped around a treated body part;

- at least one sensor (C1 ), arranged to be placed on the surface of the treated body part, and to measure at least one outer variable non-invasively (7).

- a predictive model (D) for estimating the value of the controlled inner body temperature variable (8) from the measurement of said outer variable(s);

- a controller (A) that takes the difference (2) between the reference value (1 ) and the predicted value (8) of the inner variable and controls the temperature in the cooling/heating system (3a) by changing the inputs (3) to the pad and consequently controls the output behaviour of the process.

2. The device according to claim 1 , wherein the predictive model is constructed prior to use of the device using machine learning methods, based on the results (E) from computer simulation of the therapeutic process for different simulation parameters (P), which estimates the value of the controlled inner temperature variable (8) from other variables whose measurement is more feasible (7).

3. The device according to claim 1 or claim 2 further comprising a memory and a processor, wherein the memory stores the predictive model (D) and the processor estimates the value of the controlled inner body temperature variable (8) using the predictive model.

4. A method for non-invasive real-time control of inner body temperature variables during therapeutic cooling or heating, including the steps of using a predictive model (D) to estimate the value of a controlled inner body temperature variable (8) from at least one outer variable (7) measured with a sensor (C1 ) placed on the body surface, and using the estimated value in a feedback control loop to change the temperature inputs to a temperature controlled pad which is wrapped around a treated body part, wherein the predictive model is generated offline prior to the performance of the method using the correlation between measurable and non-measurable variables in the results (E) from a computer simulation of the therapeutic process for different input simulation parameters (P).

5. The method according to claim 4, wherein the predictive model (D) is constructed by using a machine learning method which uses data (E) generated from computer simulation of the cooling or heating therapy for different input simulation parameters (P), e.g. input signals, initial and boundary conditions, or any combination of them.

6. The method according to any one of claim 4 or claim 5, wherein the predictive model (D) is built upon data for other variables of the heating or cooling process (B), and their correlation obtained from the simulation results (E) of the system, or the predictive model is built with advanced data analysis of the simulation results (E).