WO2007141345A1

WO2007141345A1 - Device for supervision and stimulation intended to fight sleep apnea

Info

Publication number: WO2007141345A1
Application number: PCT/EP2007/055723
Authority: WO
Inventors: Alfredo Hernandez; Julio Cruz; Guy Garrault
Original assignee: Universite De Rennes 1
Priority date: 2006-06-09
Filing date: 2007-06-11
Publication date: 2007-12-13
Also published as: FR2908624A1

Abstract

The invention concerns a device for supervision and stimulation of at least a sleep apnea prone newborn infant, comprising at least a sensor for acquisition of at least a representative signal of the heart rate of said newborn infant, means for treatment of said at least a signal and means for kinesthetic stimulation of said newborn infant. Said means of treatment deliver to said means of stimulation a variable control signal, enabling control of the amplitude and/or frequency of the stimulation applied to the newborn infant from the moment that said treatment means detect a variation or an abnormal modification of said heart rate.

Description

SUPERVISION AND STIMULATION DEVICE FOR FIGHT AGAINST THE APNEA OF THE

SLEEP

1. Field of the invention

The field of the invention is that of monitoring (or "monitoring", according to a terminology more English-speaking but commonly accepted in the technical field of the invention) physiological signals of individuals, and more specifically newborns.

The invention thus relates to the control of rapné / bradychardie in newborns, and in particular premature newborns.

2. State of the art

The prior art discloses systems for remote monitoring of cardiac signals, for example, enabling a patient, from home, to transmit the ECG signals (electrocardiogram) acquired by himself on his person by means of a housing. equipped with sensors, via a communication network, for example the Internet, to his or her doctors.

Thus, the ECG data is communicated (most often in a completely secure manner) by the device, also called housing, equipping the patient, to one or more client applications or remote servers, executed, for example, on the computer terminal. medical staff in charge of the follow-up.

If this type of technical solution remains suitable for certain patients with

20 of cardiac pathologies allowing remote monitoring, a first disadvantage of this type of system is related to the fact that it is the patient himself who must, according to a given prescription, set up the sensors, then initiate the transmission of data, to the medical staff, the latter taking cognizance, most often, physiological information transmitted by the patient, in time

25 deferred.

As a result, such a system is totally unsuitable for a tele-monitoring context applied to a newborn, who is physically unable to participate actively. It is also not adapted to handle an emergency problem, for example during a sleep apnea that may

30 occur untimely in the newborn. To overcome such a disadvantage, surveillance solutions, or

"Monitoring", (it should be noted that the term "monitoring" and its equivalent "monitoring", commonly accepted in the field of the invention, are used interchangeably in the following description) adapted to a hospital context have been proposed. They are based on the implementation of a housing equipped with sensors able to communicate by means of a wired connection, with a cardiac monitoring station allowing the doctor, for example through an application developed in JAVA programming language, handling, processing and analysis of the patient's ECG physiological signals.

However, such an approach has the main disadvantage of requiring the implementation of a heavy and expensive hardware architecture, in that it requires the implementation of as many sensor housings as patients to monitor, and connect all these boxes of sensors. acquisition of ECG signals by "network" cables to one or more monitoring stations.

In addition, such an approach according to the prior art also remains most often unsuited to medical emergency contexts.

The aforementioned solution has been extended in the form of a 12-channel remote monitoring system, relying on the Internet network, including for the exchange of physiological data of the intra-hospital ECG type.

Such a system has also been made compatible with the various connectors of equipment for the acquisition and communication of physiological data, so as to make it more generic and more rational in terms of manufacturing costs. Such a system according to the prior art, however, still has the disadvantage of being limited in terms of real-time processing of data acquired on a patient and therefore remain unsuitable for use in an emergency context.

In an attempt to overcome these various disadvantages of prior art technical solutions, yet another solution has been proposed. She rests, on the implementation, on the patient, of a set of communicating sensors without wires, directly connected via a wireless communication box by means of a real-time wireless connection, to a station or a terminal running a software application advanced treatment of physiological signals (ECG for example) received from the patient.

A major drawback of such a technical solution lies in the fact that all the data processing takes place only at the level of the software application and, without prior knowledge of any physiological specificities of the patient. Consequently, such a solution remains difficult to implement in contexts of an emergency nature since it does not allow proactive behavior of the system in real time. It is only after delocalized processing by the monitoring application that alerts or actions can be triggered, in a period which may sometimes prove to be too long, or incompatible with the measures that it would be necessary to take. at the level of the monitored patient.

Moreover, all these known systems pose a crucial problem in the application of the fight against apnea of newborns. Indeed, when an apnea / bradychardia is detected, it is necessary to act immediately, which supposes the presence of a large number of specialized personnel. It has been thought to automate these systems, producing an action (sound or movement for example), leading to wake up the newborn to resume his breathing and heart rate. However, these systems have another disadvantage, since they lead to wake up often the newborn, which is harmful to his health, especially for premature babies. It is known that a prolonged sleep is particularly favorable to motor development of newborns. 3. Objective of the invention

The invention described here proposes in particular to overcome the disadvantages of the technical solutions of the prior art. An object of the invention is to provide a device adapted to newborns, for the fight against apnea / bradychardia.

Another object of the invention is to provide such a device, which optimizes the comfort, health and development of newborns. Another object of the invention is to provide such a device, which are more effective than known techniques, particularly in emergency situations.

A further object of the invention is to provide such a device that is simple to implement, both from the point of view of software and hardware, while being low cost or at least reasonable.

A particular object of the invention, in a particular embodiment, is to provide such a device that can simultaneously monitor a large number of newborns.

4. Presentation of the invention These objectives, as well as others which will appear later, are achieved by means of a device for supervising and stimulating at least one newborn suitable for apnea of the sleep, comprising at least one acquisition sensor of at least one signal representative of the heart rate of said neonate, means for processing said at least one signal and kinesthetic stimulation means of said newborn.

According to the invention, said processing means deliver to said stimulation means a variable control signal, making it possible to control the amplitude and / or the frequency of the stimulation applied to the newborn, since said processing means detect a variation or an abnormal change in said heart rate.

Thus, the invention is based on a new approach to stimulating newborns, based on stimulation adapted to a given situation. It is no longer an "all-or-nothing" approach according to the prior art, but a consideration of the situation, taking into account, for example, previously stored information and databases. knowledge, for optimize in real time the stimulation provided, depending on the detected bradychardia.

Advantageously, said control signal can take at least two values, a value controlling a sufficient stimulation to wake up the newborn and a value controlling stimulation not waking the newborn.

In other words, the invention proposes that certain stimulations be subliminal only, so as not to cause awakening. This approach makes it possible to combat the morbidity due to lack of sleep. According to a preferred embodiment of the invention, said processing means take account, for the generation of said control signal, of an evolution of said signal or signals representative of the heart rate over a predetermined period of time.

Advantageously, said processing means take into account, for the generation of said control signal, reference data relating to said newborn and / or to a set of newborns stored in a database.

Preferably, said processing means comprise learning means implementing: remote means for recording said at least one signal representative of the heart rate and the events detected; means for calculating at least one criterion for variability in the evolution of said at least one signal representative of the heart rate. In this case, said processing means may in particular comprise means for comparing said at least one variability criterion with at least one predetermined threshold of detection and / or presumption of the occurrence of an abnormal event in said newborn, of so that when said at least one threshold of presumption and / or detection is crossed, said control signal is appropriately generated. According to a particular embodiment, said processing means are housed partly in at least one acquisition box installed near a newborn and for another part in at least one remote computer terminal, putting in place in particular a visualization application. In this case, said processing means and said terminal communicate advantageously via a wireless link.

In a particular implementation, said processing means implement at least one multi-agent execution environment, comprising at least the four types of software agents belonging to the group comprising: a principal agent, providing global management operation of said device;

a communication agent, able to establish a transmission protocol between said processing means and at least one computer terminal;

a data acquisition agent, able to read at least some data acquired by at least one acquisition card;

an application agent, able to carry out the processing of at least some of said data acquired by said acquisition agent.

This multi-agent approach proves to be particularly simple and effective, and adapted to variable implementations depending on the situations encountered. In a particular aspect, said sensors advantageously comprise three ECG sensors for each newborn.

Advantageously, the device of the invention comprises a plurality of housings each intended for a newborn, each comprising at least one sensor and stimulation means, said housings cooperating with a computer terminal comprising means for storing data.

This allows for efficient implementation in hospitals, allowing simultaneous monitoring of several newborns. 5. Figures

Other characteristics and advantages of the invention will emerge more clearly on reading the following description of an embodiment. preferred, given as a simple illustrative and nonlimiting example, and the appended drawings, among which:

- Figure 1 schematically illustrates the principle of the device of the invention; FIG. 2 illustrates a multi-agent structure for implementing an embodiment of the invention;

FIG. 3 illustrates an example of means for implementing a multi-agent environment, comprising a microcontroller and integrable in the autonomous and self-adaptive remote monitoring system of FIG. 2;

FIG. 4 is a general representation of a remote monitoring system according to the invention;

- Figure 5 shows different modes of operation of a remote monitoring system according to the invention; FIG. 6 presents a general hardware architecture of an intelligent remote monitoring system according to the invention;

FIG. 7 is an illustration of the multi-agent execution environment for the microcontroller of the system of FIG. 5;

FIG. 8 shows a set of agents involved in the software architecture of a remote monitoring system according to the invention;

FIG. 9 gives an illustration of the main interface of a visualization and multi-patient remote monitoring console;

FIG. 10 shows a multi-agent architecture of a monitoring station according to the invention; - Figures 11 and 12 show different hardware components implemented in a remote monitoring system according to the invention;

FIG. 13 gives an overall view of a human-machine monitoring interface fitted to a remote monitoring station according to the invention.

6. description of an embodiment 6.1. general principle The invention therefore proposes a new approach to the fight against apnea in newborns, based in particular on an adaptive approach, according to which the stimulation is adapted, notably in amplitude and / or frequency, as a function of physiological measurements and reference data. Figure 1 illustrates the structure of a device according to the invention.

It comprises one or more sensors 11, intended to be placed in the body of the newborn. These sensors 11 deliver measurement signals 12 representative of a measurement of a physiological information, for example a

ECG. In a particular embodiment, three ECG measurements are thus obtained.

These signals 12 are transmitted to processing means 13, for example a microcontroller, which determine a control (advantageously comprising two components 14A and 14F) of stimulation means 15.

(tactile vibrator), ensuring a suitable stimulation, that is to say having a 14A amplitude and a frequency 14F adapted to each particular situation.

The processing means 13 take into account the previous measurements, stored in a memory or on a server 16, and general data 17, such as rule bases for example.

The commands 14A and 14F are therefore variable, for example beat to beat, and may in particular increase if the bradycardia increases, and decrease if the bradychardia decreases. They are chosen in particular so as to be weak enough not to wake up the newborn (subliminal stimulation), when the situation does not justify it.

6.2 Implementation of the invention using a multi-agent architecture It is in particular to allow the application of adaptive therapies on the newborn, or patient, to better manage the tasks performed by the device , to distribute the tasks according to their complexity / priority, to simplify the hardware and software development ("firmware" (software embedded in the hardware) and "software" (software running in the computer)). In relation with FIG. 2, a multi-agent architecture 61 is selected for implementing the solution according to the invention.

By intelligent agent, we mean a computer object (program) that performs tasks in the manner of an automaton, according to the state of its environment and the "behaviors" associated with it.

The agents 61 are capable of a certain autonomy and, in particular, can "dialogue" with other agents (having complementary behaviors) to perform complex tasks.

Multi-agent architectures are particularly useful for facilitating the integration of knowledge into IT applications.

In relation with FIG. 3, the technical approach adopted authorizes the creation of a new complete multi-agent environment, applicable to an electronic device equipped with a microcontroller 71 and makes it possible to propose a new electronic architecture facilitating the development of the hardware. and software embedded in the hardware, or "firmware", as well as the integration of knowledge.

This results in the advent of a new intelligent remote monitoring system able to communicate wirelessly, both autonomous and self-adaptive for the monitoring of physiological signals, for example ECG acquired on a monitored patient.

It is also translated by the implementation of a new process of automatic adaptation of the activity of the system according to the context and of a new complete system (hardware & software embedded in the hardware

(firmware) & software) enabling adaptive stimulation monitoring, detection and interruption of bradycardia of newborns.

In relation to Figure 4, the proposed system is composed of the following elements:

one or more intelligent devices 80 for monitoring and stimulating a set of patients 83; an application 81 for visualizing, configuring, processing and controlling all the devices 80 that can operate on one or more computers 82 at the same time, locally or remotely.

In connection with Figure 5, typically, one or more devices (operating as master 91, or slave 92) may be connected to patient 90 and operate in one of three ways:

An autonomous modality 93, in which each device 91 acquires the data 94, performs the signal processing locally, or adaptively applies therapy 94 and records certain events in its local memory.

A master-slave modality 95, in which each device 92 (slave) carries out data acquisition and wireless transmission 97 to one or more computers 98 (masters). The application 99 on the master computer allows, among other things, the viewing, recording, signal processing 99b and the automatic or manual control of the therapy applied by the device.

- A hybrid modality (not shown in the figure), where the device is able to respond to a number of pathological and / or physiological events (and adjusts the therapy accordingly), but communicates 96 with an application master for decisions that are more difficult to make, requiring, for example, more important means of calculation or access to databases or knowledge. In relation with FIG. 6, each intelligent device 100 consists of four interconnected electronic cards 101, 102, 103, 104: a commercially available "processor card" 101, which contains a microcontroller 105 ATMegal28, RAM and Flash memories , RTC clock, etc. ;

an "acquisition card" 102, specific to the problem being addressed, which enables the acquisition of the physiological data and their analog-digital conversion; a "communications card" 103 that allows communication between different devices and the master application;

a "motherboard" 104, developed by the inventors, which performs the following functions: o Interface 106 between the different cards by means of serial or parallel protocols, o Management 108 of the additional analog and digital inputs and outputs, o Management of the supply of the device 109 (batteries, DC power supply).

In relation with FIG. 7, a multi-agent execution environment 111 (also called a "framework") and a function library in C or C ++ language, adapted for use in microcontrollers, were first created, using a simplified message communications protocol, which still allows communication with standards such as FIPA-ACL. The developed environment manages the exchange of messages between the different agents and establishes the order of execution of the different behaviors of the agents, using a FIFO approach. The execution of the behaviors is based on the "life cycle" of the agents, defined by the FIPA, and includes the states of initialization, activation and waiting. The behaviors of an agent can be executed only during the activation state.

In connection with FIG. 8, once the multi-agent execution environment 110 has been developed, four types of agents 121, 122, 123, 124 adapted for the main tasks of the device have been defined and implemented: principal, responsible for the overall management of the device and, in particular, the functionalities of the motherboard; it manages the power supply of the device, the control of the real-time clock (RTC), the non-volatile memory and the inputs / outputs of the motherboard (buttons, diodes (still called "leds"), actuators (still called "Actuators")). In addition, this agent contains a list of possible clinical applications with the hardware (also called "hardware") available in the device.

A communication agent 122, which establishes a transmission protocol between the device 100 and the computer. This protocol allows error checking and a variable data size, regardless of the means of transmission used (USB, bluetooth, etc.); a data acquisition agent 123, which has cyclic type behaviors capable of reading the data acquired from the different acquisition cards integrated in the device; a clinical application agent 124, which performs the treatment of the acquired data, the detection of events or the adaptation of the therapy, depending on the clinical application for which it was designed.

An essential and additional advantage of such a multi-agent execution environment 111 (framework) concerns its relatively easy implementation within one or more microcontrollers embedded in the device according to the invention.

It is important to emphasize that a given device may have several agents 124 of different clinical applications.

At the start of the runtime environment, the primary agent and the communication agent are created. If there are monitoring applications available in the communications rank of the device, the master agent transmits to all these applications (using the communication agent) the characteristics of the devices, which include a unique identifier and a list possible clinical applications with the hardware available in the device. For example, a device with 3 ECG pathways may be useful for monitoring the bradycardia of newborns, but also to detect certain cardiac arrhythmias in patients who have suffered a myocardial infarction. The methods of signal processing and the possible therapeutic actions applied to the patient will be different for these two applications. After initialization, the principal agent waits for user input to define the operation mode (autonomous, master-slave or hybrid) and the desired clinical application (ie bradycardias, arrhythmias, etc.). This interaction with the user can be performed either directly on the device by means of the push buttons or through the graphical interface of the application. The primary agent then creates a data acquisition agent corresponding to the available acquisition hardware and a clinical application agent, depending on the user's choice to start the monitoring process. These agents will send messages continuously during data monitoring. The clinical application agent contains, among other things, a behavior associated with the analysis of the data received from the acquisition agent and a behavior that controls, through the main agent, the operation of the digital and analog outputs of the device . This last behavior is particularly useful for the application of therapeutic stimulation or the generation of alarms. Due to the fact that these two behaviors work concurrently, the application of the stimulation can be continuously modified and optimized according to predefined parameters or physiological knowledge included in the clinical application agent.

In relation with FIG. 9, an example of a main interface 131 and a monitoring interface 132 of a multi-patient viewing and supervision console 130 is presented.

Such a console is developed using the standard multi-agent environment JADE (registered trademark), which allows the automatic adaptation of the graphical interfaces 131 and 132 (developed in object-oriented language JAVA - registered trademark) according to the application clinic. Such a console 130 advantageously makes it possible to view, record and process all or part of the physiological data acquired from one or more patients simultaneously.

It also makes it possible to configure and control the various intelligent patient monitoring devices, while managing the databases 187,

182 of patient data and knowledge (respectively).

Such a console 130 also advantageously offers the possibility of estimating in real time, the optimal stimulation to bring to the patient, once one or more alarms have been triggered. This console 130 according to the invention offers the additional major benefit of allowing the simultaneous supervision and monitoring of a large number of hospitalized patients in real time, which also contributes to reducing the material and human costs in the patients. hospitals and / or maternities. In relation with FIG. 10, the multi-agent architecture of the station (software) is presented. The main agents defined are:

The main agent 141, which manages the entire application and the multi-patient display console 146. It also allows the creation of graphical interfaces according to the context, for example, for the creation of a list of possible clinical applications, according to the detected intelligent monitoring devices.

The central agent 142, which makes it possible to export the most relevant information of each patient 143 to the central multi-patient visualization console, to follow the main monitored parameters, the alarms and the detections of the risk events of the patient. all patients followed.

The monitor agent 144, responsible for creating the user interface 145 specific to a patient, a clinical problem and a given monitoring device. He is also responsible for recording data and distributing analysis and communication tasks to other agents. The graphical interface 145 created by the monitor 144 operates in real time and presents the acquired signals 133, the detected parameters 134, their temporal evolution, etc.

The acquisition agent 147 is an "intermediate agent", which carries out the transfer of information with the "communication" agent of the intelligent device 123 according to the invention, by translating the FIPA-ACL messages.

(used in the application) in the format of the protocol developed for the microcontroller, in the context of the present invention. This agent 147 has a graphical interface which facilitates access to the parameters of the associated device 135. This interface is displayed via the monitor agent 144.

The processing agent 1400 is responsible for analyzing the data from the acquisition agent 147 (in the master-slave operating mode). In the behaviors of this agent 1400, it is possible to easily implement the various methods of information processing, by means of the use of a C ++ library coupled to the application by a 1401 JNI interface (" Java Native Interface "in English or" Java Native Interface "in French - with" Java ", registered trademark).

It should be noted that, in a particular embodiment, the method and / or system according to the invention can be applied to the detection and treatment of bradycardias in preterm infants. It is recalled that, unfortunately, apnea and bradycardia are dangerous for the premature newborn, which sometimes requires invasive resuscitation maneuvers, which induce an extension of the length of hospitalization.

Such apneas are also involved in long-term neurological morbidity (frequent awakenings).

Now, it is clear today that the detection of such apneas by scope and manual stimulation requires too long delays in intervention. In addition, detection and computer-controlled stimulation, for example of PC type (fixed amplitude) proposed by some systems of the prior art require always to wake up the newborn.

Thus, the use of the remote monitoring system according to the invention advantageously allows the early detection and characterization (autonomous or hybrid) of the bradycardias, then the application of an adaptive stimulation (beat to beat) which allows:

- optimization of stimulation (if possible subliminal)

- Increasing the stimulation intensity for important bradycardias

- The decrease of stimulation intensity for lower bradycardias.

In connection with FIG. 11, for the example of the detection of bradycardias in the newborn, a card of acquisition of 3 channels of electrocardiography (ECG), with the labels EN 60601 (type BF) and

AAMI / ANSI (AAMI EC 13-92) and available on the market (MCC ECGBoard), was chosen.

Communications with other devices or with the master application is based on "BlueTooth" technology, using an OEM 164 card. These two cards 160 and 164 are connected by serial link

161, 162, 163 to the motherboard 165 and, thereafter, to the processor board 166, with, for example, an ATmega 128 microcontroller. A kinesthetic stimulator 167

(Touch vibrator), for example the brand "Audiological Engineering" (registered trademark) is applied to the skin of the newborn and connected to the analog output of the motherboard. The motherboard 165 also allows the application of auditive stimulation 1600, with a built-in speaker. It may also include means 168 for visual stimulation, of the diode ("LED") type.

It is important to emphasize that in this document, a particular application of this system for the early detection of large bradycardias of the premature newborn is presented. Indeed, the entire system has already been developed at the prototype stage for this application.

Moreover, in terms of targeted clientele, there is currently no functional automatic system in neonatal intensive care unit. The expected economic and industrial benefits are therefore potentially significant in the context of hospitals or private clinics.

In terms of public health and clinical outcomes, rapid inhibition of bradycardia by stimulation is expected to reduce the need for resuscitation and ventilator-assisted intubation, reduce the length of hospital stays, and reduce home monitoring needs and therefore improve the quality of life of babies. 6.3 Further Implementation Details of the Multi-Agent Architecture

The proposed system falls within the general framework of distributed intelligent monitoring and responds to the growing need to perform continuous (tele) monitoring of physiological data of healthy or pathological subjects and to provoke an optimal adaptive action (possibly therapeutic), according to the state of health of the patient, locally or remotely. The "intelligent" aspect of the system is associated with the fact that it is able to adapt to the surveillance context and to the clinical problem studied, while allowing to dynamically optimize stimulation on the patient.

One of the main difficulties currently encountered for the development of this type of device is associated with the integration of problem-specific knowledge, to detect risk events or to define and / or apply the optimal therapy. This problem is particularly important when implementing these methods in a microcontroller embedded in the device.

One of the main originalities of this system is associated with the proposal of a multi-agent architecture, embedded in the microcontroller, to facilitate the introduction of knowledge, the communication with others devices or applications of information processing and application development on this architecture.

The application example presented in this document is based on the monitoring of the bradycardias of the premature newborn. These bradycardias are defined by a drop in heart rate of at least 33% from baseline. If bradycardias are frequent events, which may occur alone or more often (80% of cases) associated with apnea and / or desaturation, they place the newborn in immediate danger in the absence of appropriate intervention. . Experience shows that intervention times remain long for babies in distress, even if they are performed very efficiently.

The application of the proposed system in this context allows: i) real-time remote and wireless ECG recording of prematurity, ii) extraction of its cardiac variability curve, iii) early detection of bradycardia and iv) adaptive activation of a kinesthetic stimulator. The expected results are particularly important clinically as it is considered that automatic detection and stimulation of newborns should minimize the risk of prolonged and / or profound apnea-bradycardia, decrease the use of resuscitation or intubation maneuvers and shorten the duration of hospitalization.

The proposed system consists of the following elements:

- one or more intelligent monitoring and stimulation devices;

an application for viewing, configuring, processing and controlling the set of devices, which can operate on one or more computers at the same time, locally or remotely.

Typically, one or more devices can be connected to the patient and operate in one of three ways:

- An autonomous mode, in which each device acquires the data, carries out the processing of the signals locally, applies or not a Adaptive therapy and records certain events in its local memory.

- A master-slave mode, in which each device (slave) realizes data acquisition and transmission to one or more computers (masters). The application on the master computer allows, among other things, viewing, recording, signal processing and automatic or manual control of the therapy applied by the device.

- A hybrid modality, where the device is able to respond to a certain number of pathological events (and adapts the therapy accordingly), but communicates with a master application for more difficult decisions to take, requiring, for example, means more important calculations or access to databases or knowledge. Figure 5 shows two possibilities for monitoring newborns with the proposed system: the autonomous mode and the master-slave mode.

Two modes of operation of the proposed system: on the left an autonomous application in which the intelligent devices realize the acquisition, the treatment and the stimulation without interaction with the outside; on the right a master-slave application in which the processing and the decision-making are carried out in the computer (master application).

The following is a more detailed description of the various components of the system according to the invention.

Each intelligent device consists of four interconnected electronic cards (Figure 6):

- a "processor card", available on the market, which contains an ATMegal28 microcontroller, RAM and Flash memories, RTC clock, etc.

- an "acquisition card", specific to the problem addressed, which allows the acquisition of physiological data and their analog-digital conversion; - a "communications card" that allows communication between different devices and the master application;

- a "motherboard", developed by us, which performs the following functions: - Interface between the different cards by means of serial or parallel protocols,

Management of additional analog and digital inputs and outputs,

Power management of the device (batteries, DC power supply).

This modular hardware architecture facilitates the generalization of the device to several different applications, by a simple change of cards.

The motherboard allows the connection of a set of cards specific to the problem dealt with.

For the example of the detection of bradycardias in the newborn, we have chosen a 3-channel acquisition card for electrocardiography (ECG), with the labels EN 60601 (type BF) and AAMI / ANSI (AAMI EC 13 -92) and available on the market (MCC ECGBoard). Communications with other devices or with the master application is based on BlueTooth technology, using an OEM card. These two cards are connected by serial link to the motherboard and then to the processor card. A kinesthetic stimulator (tactile vibrator) of the brand Audiological Engineering is applied on the skin of the newborn and connected to the analog output of the motherboard. The motherboard also allows the application of auditory stimulation, with a built-in speaker. A prototype is presented figure 12, for the detection of the bradycardias of the newborn. In the center, there are three interconnected main cards and superimposed (motherboard 171, processor card 172 and ECG card 173). The various functions of the device (acquisition, communication, stimulation, etc.) are implemented in the microcontroller (processor card) by means of an embedded program (firmware). The development of this type of embedded software is often tedious and difficult to generalize. One of the originalities of the approach presented here lies in the proposal of a multi-agent architecture, implemented on the microcontroller.

The application of multi-agent architectures for real-time intelligent monitoring of continuous processes has been an active research topic during the last decade. An intelligent agent is a computer object (program) that performs tasks in the manner of a PLC, depending on the state of its environment and the "behaviors" associated with it. They are capable of a certain autonomy and, in particular, can "dialogue" with other agents (with complementary behaviors) to perform complex tasks. Multi-agent architectures are particularly useful for facilitating the integration of knowledge into IT applications.

Several message exchange languages have been proposed and new standards have recently emerged, such as the Knowledge Query and Manipulation Language (KQML), and more recently, the FIPA-ACL (ACL for Agent Communication Language) standard. However, none of these architectures is adapted to the constraints associated with microcontroller programming. We have therefore developed a framework of multi-agent execution (framework) and a library of functions C adapted for use in microcontrollers, by using a simplified protocol of communications of messages, which still allows the communication with standards like the FIPA-ACL.

The developed environment manages the exchange of messages between the different agents and establishes the order of execution of the different behaviors of the agents, using a FIFO approach. The execution of the behaviors is based on the "life cycle" of the agents, defined by the FIPA, and includes the states of initialization, activation and waiting. The behaviors of an agent can be executed only during the activation state. Once the execution environment was developed, we defined and implemented four types of agents adapted for the main tasks of the device (figure 8):

the main agent, responsible for the overall management of the device and, in particular, the functionalities of the motherboard; It manages the power supply of the device, the control of the real-time clock (RTC), the nonvolatile memory and the inputs / outputs of the motherboard (buttons, LEDs, actuators). In addition, this agent contains a list of possible clinical applications with the hardware available in the device; - The communication agent, which establishes a transmission protocol between the device and the computer. This protocol allows an error check and a variable data size, regardless of the means of transmission used (USB, bluetooth, etc.);

the data acquisition agent, which has cyclic type behaviors capable of reading the data acquired from the different acquisition cards integrated in the device;

- The clinical application agent, which performs the treatment of acquired data, the detection of events or the adaptation of the therapy, depending on the clinical application for which it was designed. It is important to emphasize that a given device may have several agents of different clinical applications.

At the start of the runtime environment, the primary agent and the communication agent are created. If there are monitoring applications available in the communication rank of the device, the master agent transmits to all these applications (using the communication agent) the characteristics of the devices, which include a unique identifier and a list possible clinical applications with the hardware available in the device.

After initialization, the principal agent waits for a user input to define the operation mode (autonomous, master-slave or hybrid) and the desired clinical application (ie bradycardias, arrhythmias, etc.). This interaction with the user can be performed either directly on the device by means of the push buttons or through the graphical interface of the application. The primary agent then creates a data acquisition agent corresponding to the available acquisition hardware and a clinical application agent, depending on the user's choice to start the monitoring process. These agents will send messages continuously during data monitoring.

The clinical application agent contains, among other things, a behavior associated with the analysis of the data received from the acquisition agent and a behavior that controls, through the main agent, the operation of the digital and analog outputs of the device . This last behavior is particularly useful for the application of therapeutic stimulation or the generation of alarms. Due to the fact that these two behaviors work concurrently, the application of the therapy can be continuously modified and optimized according to predefined parameters or physiological knowledge included in the clinical application agent ("therapy" behavior).

For the above-described example of bradycardia, the data acquisition agent receives the ECG signals from the card and transmits this information to a specific clinical application agent (bradycardia agent).

The latter contains behaviors allowing: i) to detect bradycardias (based on algorithms presented in the literature or proposed in the laboratory) and U) to stimulate the newborn in an adaptive way, according to the observed response. More particularly, kinesthetic stimulation is performed at different amplitudes and frequencies, depending on the evolution of the state of the newborn and its response to the stimulation applied. The adaptive definition of this therapy can be achieved by a transfer function, a rule base, an interval cutting optimization algorithm (such as the golden section), a "black box" model or a physiological model, linking the measured heart rate to the optimal stimulation parameters.

As an illustrative example of implementation of the invention, an intelligent monitoring multi-platform intelligent monitoring application has been developed, also using a multi-agent approach.

This application is coded in Java and uses the libraries of the Java Agent Development (JADE) project, which is an open-source multi-agent system development environment (also called "open-source" in a more English-language terminology). In particular, JADE offers advanced support for the FIPA-ACL standard, as well as syntax validation tools for messages between agents. The information processing methods included in the platform are developed in C ++ and associated with the station using JNI interfaces. The application can operate on one or more computers at the same time, locally or remotely and allows: automatic adaptation of the graphical interface according to the clinical application context and available and detected devices; viewing, recording and processing of data received from the devices; - initial configuration and control of the set of devices;

- low management of patient data and low knowledge;

- the detection of events using more complex methods than those implemented in the device; the estimation of an optimal and adaptive therapy and the remote control of the devices.

The multi-agent architecture of the station is shown in Figure 10. The main agents defined are: the main agent, which manages the entire application and the multi-patient visualization console. It also allows the creation of graphic interfaces depending on the context, for example, for the creation of a list of possible clinical applications, depending on the detected intelligent monitoring devices.

The central agent, which allows you to export the most relevant information from each patient to the central multi-patient visualization console, to monitor the main monitored parameters, alarms and detections of risk events for all patients followed. The monitor agent, responsible for creating the specific patient-specific user interface, clinical problem, and monitoring device. He is also responsible for recording data and distributing analysis and communication tasks to other agents. The graphical interface created by the monitor works in real time and presents the acquired signals, the detected parameters, their temporal evolution, etc. The acquisition agent is an "intermediate agent", which carries out the transfer of information with the "communication" agent of the intelligent device, by translating the FIPA-ACL messages (used in the application) in the format of the protocol that we have developed for the microcontroller. This agent has a graphical interface that facilitates access to the parameters of the associated device. This interface and is displayed through the monitor agent.

The processing agent is responsible for analyzing the data from the acquisition agent (in the master-slave operating mode). In the behaviors of this agent, it is possible to easily implement the various methods of information processing proposed in the laboratory, by means of the use of a C ++ library coupled to the application by a JNI interface.

A succinct presentation of the use of the application 180 is proposed below:

(AT). When the application 180 starts, the main agent is created. It detects the different possible clinical applications by interrogation of 181 available intelligent devices and a 182 knowledge base, or base of Géranipales, which represents the different clinical problems treated by the application.

(B). This agent presents a graphical interface with a context sensitive palette 185, which lists the possible clinical applications according to the set of devices.

(VS). The user selected in this palette 185 the desired application and a dialogue box 186 allows him to integrate patient data, or select a patient already available in the database 187 patient data. The databases 187 and data 182 are implemented on a MySQL (registered trademark) manager.

(D). The different agents are created for each clinical application and for each patient: a central agent (which communicates with the multi-patient central viewing console 180) and the acquisition, treatment and monitor agents.

Once the monitoring application 188 has been initiated, the user can view the acquired signals 189 and the detected parameters of the patient 1800, activate and modify the parameters of the various signal processing methods 1801 and manually configure or control the device 1802. the inventors' knowledge, there is currently no other technology available to manufacturers that makes it possible to include in the microcontroller of a medical device the knowledge specific to the problem addressed and having the characteristics of genericity and ease of use described in this document. Compared to the specific application in bradycardia, several embedded neonatal stimulation applications have been proposed. None of these applications proposes an adaptive stimulation capable of, for example, increasing the stimulation amplitude for larger bradycardias and to decrease this stimulation for weaker bradycardias. Thus, from a technological point of view, the main advantage of this invention is associated with the use of a multi-agent environment, implemented in the embedded software ("firmware") of a microcontroller, allowing easy communication with a standard multi-agent system, defined at software level (computer). This approach facilitates the integration of problem-specific knowledge into the device and the communication between the different devices and a master application. In addition, it significantly reduces the development time of firmware and software. Another advantage is associated with the generality of the hardware design, which facilitates the extension to other applications by simply changing the card. This aspect also reduces the time and costs of hardware development.

For the bradycardia application of neonates, the proposed adaptive stimulation has several advantages over a "classic" fixed stimulation. Such adaptive stimulation achieves an interesting compromise between the minimum amplitude necessary to obtain a response (thus favoring the rest periods, strongly disturbed by the current methods) and a greater amplitude (allowing a faster response and a decrease duration of bradycardia). In addition, there is currently no automatic functional system in neonatal intensive care unit.

Claims

1. A device for supervising and stimulating at least one newborn suitable for sleep apnea, comprising at least one acquisition sensor for at least one signal representative of the heart rate of said newborn, means for of treating said at least one signal and kinesthetic stimulation means of said newborn, characterized in that said processing means deliver to said stimulation means a variable control signal, making it possible to control the amplitude and / or the frequency of the stimulation applied to the newborn, since said processing means detect abnormal variation or modification of said heart rate.

2. Monitoring and stimulation device according to claim 1, characterized in that said control signal can take at least two values, a value controlling a sufficient stimulation to wake up the newborn and a value controlling a stimulation not waking the patient. new born.

3. Monitoring and stimulation device according to any one of claims 1 and 2, characterized in that said processing means take into account, for the generation of said control signal, an evolution of said signal or signals representative of the frequency heart rate over a predetermined period of time.

4. Monitoring and stimulation device according to any one of claims 1 to 3, characterized in that said processing means take into account, for the generation of said control signal, reference data relating to said newborn and / or to a set of newborns stored in a database.

5. Monitoring and stimulation device according to any one of claims 1 to 4, characterized in that said processing means comprise learning means implementing: means for remotely recording said at least one signal representative of the heart rate and the events detected;

means for calculating at least one criterion for variability in the evolution of said at least one signal representative of the heart rate.

6. Monitoring and stimulation device according to claim 5, characterized in that said processing means comprise means for comparing said at least one variability criterion with at least one predetermined threshold of detection and / or presumption of the occurrence of an abnormal event in said newborn, so that when said at least one threshold of presumption and / or detection is crossed, said control signal is appropriately generated.

7. Supervision and stimulation device according to any one of claims 1 to 6, characterized in that said processing means are housed in part in at least one acquisition box installed near a newborn and for another part in at least one remote computer terminal, implementing in particular a visualization application.

8. Supervision and stimulation device according to claim 7, characterized in that said processing means and said terminal communicate via a wireless link.

9. Monitoring and stimulation device according to any one of claims 1 to 8, characterized in that said processing means implement at least one environment (110) multi-agent execution, comprising at least the four types software agents belonging to the group comprising:

a principal agent (121) providing global management of the operation of said device;

a communication agent (122) capable of establishing a transmission protocol between said processing means and at least one computer terminal;

an agent (123) for acquiring data, able to read at least some data acquired by at least one acquisition card (102); an application agent (124) capable of processing at least some of said data acquired by said acquisition agent (123).

10. Monitoring and stimulation device according to any one of claims 1 to 9, characterized in that said sensors comprise three ECG sensors for each newborn.

11. Monitoring and stimulation device according to any one of claims 1 to 10, characterized in that it comprises a plurality of housings each for a newborn, each comprising at least one sensor and stimulation means, said boxes cooperating with a computer terminal comprising data storage means.