FR3113824A1 - Method and device for monitoring body analyte concentration - Google Patents

Abstract

The present invention relates to a method for monitoring the body analyte concentration in an individual, implemented by a processing unit of a body monitoring device associated with a sleep sensor, comprising the steps of:obtaining (300) a sleep state of the individual, based on a measurement acquired by the sleep sensor, determining (400) the relevance of a health alert, the health alert being determined as relevant under the two cumulative conditions following:- a predicted body analyte concentration (C) associated with the individual at a future time is outside of a normal range;- the obtained sleep state of the individual is a state conducive to receiving the health alert, and, if the health alert is determined to be relevant, issuing (500) an alert signal by the body monitor. Figure for abstract: Figure 3

Description

Method and device for monitoring body analyte concentration

TECHNICAL FIELD OF THE INVENTION

The present invention falls within the field of body monitoring devices, for example for recording and monitoring biochemical parameters in humans.

The invention finds an advantageous application in particular for ready-to-wear or "wearable" devices of the electronic watch or electronic bracelet type, used daily for real-time monitoring of the blood sugar of a diabetic patient or another body analyte.

In particular, the invention relates to an improved method for monitoring bodily analyte concentration in an individual, an improved body monitoring device configured to perform such a method, a computer program product for implementing such a method, and computer readable means for carrying out such a method.

STATE OF THE ART

The monitoring of many chronic diseases in humans requires a daily reading of biochemical parameters and quantities.

For example, health monitoring of a diabetic patient requires daily blood glucose readings of the diabetic patient. The usual protocol consists of regularly performing a puncture, for example at the end of the finger, to cause a drop of blood to bead up, then to perform the daily blood glucose measurement using the drop of blood thus obtained.

However, such a puncture is unpleasant and invasive. There is thus a risk that the diabetic patient does not perform the punctures regularly enough. The quality of the patient's blood glucose monitoring may be degraded. In addition, blood glucose monitoring depends on the rigor of the patient, who must generally carry out the punctures themselves on a daily basis.

To avoid these drawbacks, automated systems known as CGM for “Continuous Glucose Monitoring” have been proposed for automatically carrying out, at regular intervals, a blood sugar measurement at the level of the interstitial liquid between the cells of the skin. The glycaemia of the interstitial fluid is in fact very close to the glycaemia of the blood plasma. CGM systems can be ready-to-wear devices, also called "wearables" according to English terminology, placed in continuous contact with the patient's skin. These include, for example, electronic watches or electronic bracelets. Interstitial fluid level measurements have the advantage of simple and minimally invasive monitoring of patient blood glucose.

It is also possible to determine with the help of some of these devices a predictive blood sugar level of the wearer in the next few hours. Such CGM systems make it possible to predict in advance possible excursions beyond a hyperglycemia threshold or below a hypoglycemia threshold. The predictive glycemia is for example obtained according to the glycemic measurements acquired in real time in the interstitial liquid.

CGM systems thus advantageously integrate pre-programmed alerts to warn the wearer of a risk of future hypoglycemia or hyperglycemia. The alerts transmitted to the bearer are preferably discrete alerts; these are, for example, slight vibrations of the case of the electronic watch, or even notifications displayed on the screen of the electronic watch. Alternatively, notifications can be displayed on the screen of a mobile device, such as a smartphone, having a connection with the CGM system.

One advantage of monitoring and the transmission of alerts is to draw the wearer's attention to a risk of hypoglycaemia or hyperglycaemia, even if the wearer does not take the initiative to regularly consult himself. even blood sugar tracking information on his wearable device.

As such, the American patent application published under the number US 2015/335272 A1 describes a CGM system which comprises an electronic watch with blood glucose test strips, and which comprises a mobile phone having a Bluetooth wireless connection with the watch. When an excursion below a hypoglycemia threshold is expected in the next few hours for the wearer of the watch, an alert is sent to the mobile phone, and can also be sent to pre-registered contacts.

However, this state-of-the-art system has drawbacks, especially for night-time use.

A first drawback is that the health alerts emitted by the CGM system do not take into account the more or less significant variations acceptable for blood sugar levels depending on the state of sleep. For many diabetic patients, blood glucose is likely to drop sharply and reach low levels overnight, without however creating a significant health risk (and therefore without the need to alert the wearer during the sleep). The lack of consideration of the sleep profile causes unnecessary alerts to be issued.

A second drawback is that the CGM system is quick to emit an alert (therefore a notification and/or an audible signal that wakes the wearer) while the wearer is in a sleeping state, or even in a deep sleep state. The alert is issued when an excursion below the threshold is planned for a future period of less than a certain number of hours, regardless of the wearer's activity status. Warnings harmful to the health and/or well-being of the wearer are therefore issued.

It would be preferable that such unwanted alerts be avoided, not only because it is undesirable to unnecessarily wake up a patient whose general state of health may already be degraded, but also to ensure that the user continues to use its CGM system.

Existing CGM systems are therefore likely to multiply useless and/or harmful alerts, and to wake up the wearer without significant and proven health risk. There is a risk that the wearer of the CGM system may be inconsistent in using the CGM system, or even cut the power supply to the CGM system before falling asleep.

GENERAL DESCRIPTION OF THE INVENTION

In view of the foregoing, state-of-the-art body monitoring devices are not entirely satisfactory.

There is a need for a method for monitoring the concentration of bodily analyte (for example blood sugar) allowing the wearer of a monitoring device to be alerted to a predictable excursion of his analyte concentration outside a range of normality, the excursion possibly being planned during his sleep, while limiting as far as possible the alerts that are undesirable for the wearer or irrelevant given his condition. In particular, we want to take into account the awake state or not of the wearer.

In particular, a glycemic monitoring method is sought which can be carried out by an electronic watch or an electronic bracelet of a CGM system, independently generating health alerts if necessary, while dismissing unnecessary alerts which would wake up the wearer or come out of a deep sleep.

Preferably, a method and a device are sought which require few or no adjustments on the part of the wearer. The monitoring device sought is capable of operating without necessarily asking the wearer to specify at what time he is going to bed, and/or at what time he plans to wake up.

More generally, a secondary need is for the desired monitoring device to adapt to the wearer's activity profile and sleep habits, with as little stress as possible on the wearer.

In order to meet these needs, the invention relates, according to a first aspect, to a method for monitoring the concentration of body analyte in an individual, implemented by a processing unit of a body monitoring device associated with a sensor sleep. The method includes steps of:
obtaining a state of sleep of the individual, from a measurement acquired by the sleep sensor,
determination of relevance of a health alert, the health alert being determined as relevant under the following two cumulative conditions:
- if a predicted concentration of bodily analyte associated with the individual at a future instant is located outside of a range of normality;
- and if the sleep state obtained from the individual is a state conducive to receiving the health alert,
and, in the event that the health alert is determined to be relevant, issuing an alert signal by the body monitor.

It is thus evaluated whether a state of sleep of the wearer of the body monitoring device allows the transmission of the alert, based on a measurement from an additional sleep sensor, before the transmission of the alert signal by the device of bodily monitoring.

A first advantage is to eliminate undesirable alerts for the wearer taking into account his state of sleep. In particular, unnecessary alerts which would be likely to wake the wearer (and/or unnecessary alerts which would bring the wearer out of a state of deep sleep) are deleted. This promotes regular use of the body monitoring device by the wearer, including during the night.

A second benefit is that the body-monitoring device remains capable of issuing health alerts autonomously. Although an additional measure illustrating the state of sleep is used, the wearer is not necessarily asked to determine this additional measure.

Optional and non-limiting characteristics of the monitoring method of the invention are the following, taken alone or in any of the technically possible combinations:

- the sleep state obtained from the individual is a predictive sleep state at the future time;

- the method comprises an additional step of calculating an alert instant prior to said future instant, a predictive sleep state of the individual at the alert instant being considered more conducive to receiving the health alert than the predictive sleep state of the individual at the future instant.

An advantage of this additional feature is to ensure that the alert signal is emitted at a favorable time for the wearer (the alert time), while being emitted early enough to allow the wearer to take corrective action before an excursion beyond the range of normality;

- the predictive sleep state at the future instant is a deep sleep state and/or an asleep state;

- the predicted sleep state at the time of the alert is a light sleep state or an awake state;

- the individual's sleep state is determined according to a behavioral model associated with the individual;

- the behavioral model is obtained by a behavioral analysis module integrated into the body analysis device, or is obtained from a remote behavior database.

An advantage of these additional features is that it is not necessary to ask the wearer to specify what time he is going to bed, and/or what time he plans to wake up. Indeed, the behavioral model can make it possible to predict this information directly from the measurements provided by the sleep sensor;

- the behavioral model associated with the individual comprises a probability distribution of the duration of sleep of the individual, and/or a probability distribution of the hour of falling asleep of the individual, and/or a probability distribution of the individual's waking time, and/or a probability distribution of the individual's deep sleep duration, and/or a probability distribution of the individual's light sleep duration;

- the behavioral model associated with the individual comprises correlation data, said correlation data matching a plurality of measurement ranges for the measurement acquired by the sleep sensor and a plurality of sleep states;

- the correlation data make it possible to deduce from the measurement acquired by the sleep sensor that the individual is in a pre-sleep state;

- the predictive concentration of bodily analyte is obtained for a plurality of instants, the range of normality being variable as a function of said instants;

- the range of normality associated with the future instant is determined according to the predicted sleep state at the future instant;

- the method includes obtaining an actual sleep state at a present time, and includes issuing the alert signal at the present time if the actual sleep state at the present time is conducive to health alert;

- the method comprises an additional step of acquisition, by the body monitoring device, of the predictive concentration of body analyte associated with the individual at the future instant;

- the predicted concentration of body analyte is obtained independently of the individual's sleep state;

- the predicted concentration of body analyte is determined according to a concentration measurement obtained by an additional sensor;

- the additional sensor is embedded in the body monitoring device;

- the method comprises the detection of a concentration measurement error at the level of the additional sensor, the alert signal being provided to notify the concentration measurement error to the individual or to a third party.

An advantage of a method according to the latter case is that instead of immediately sending an alert signal to the wearer of the device, concerning a malfunction of the additional sensor (for example of the blood glucose sensor), it is possible to obtain a predictive concentration in bodily analyte decorrelated from the dysfunctional measurement of the sensor and to determine the relevance of the alert signal according to this predictive concentration.

- the alert signal includes an audible signal and/or a visual signal and/or a tactile signal;

- the alert signal is sent by the body monitoring device to the individual.

In this variant, the individual is notified himself, which is particularly advantageous in cases where the individual can take suitable corrective measures;

- the alert signal includes a communication signal to a third-party device.

In this variant, the third-party device can in particular emit an alert signal for another individual distinct from the wearer of the body monitoring device, which is advantageous if the individual may need outside help. The third-party device can also be a device used to alert the individual, for example a luminous device in the room where the individual is located;

- the third-party device includes a personal mobile device and/or a home device and/or an emergency service server;

- the measurement acquired by the sleep sensor comprises a heart rate measurement, and/or an accelerometer measurement, and/or an impedance measurement by electrocardiogram, and/or a gyrometer measurement, and/or a visual measurement sleepiness of the individual, and/or a measurement of the respiratory rate;

- the predictive body analyte concentration comprises a predictive blood glucose concentration, and/or a predictive blood lactate concentration.

According to a second aspect, the invention relates to a body monitoring device, said device being adapted to be worn by an individual and being configured to implement a monitoring method as defined above, the device comprising:
an alert relevance determining module, configured to compare a predictive body analyte concentration associated with the individual against a range of normality, and further configured to obtain a sleep state of the individual, the alert relevance determination being further configured to receive a sleep measurement from a sleep sensor,
an alert signal output module, configured to output an alert signal if the alert relevance determination module determines that a health alert is relevant.

Optional and non-limiting characteristics of the body monitoring device according to the second aspect of the invention are the following, taken alone or in any of the technically possible combinations:

- the sleep sensor is integrated into the monitoring device;

- the sleep sensor comprises a heart rate sensor, and/or an accelerometer, and/or a gyrometer, and/or a respiratory rate sensor, and/or an electrocardiograph sensor, and/or an imaging device;

- the device further comprises an additional sensor configured to acquire a concentration measurement of bodily analyte in real time;

- the additional sensor comprises an electrochemical sensor with one or more needles;

- the device further comprises a behavioral analysis module, comprising a memory in which a behavioral model associated with the individual is stored, the behavioral analysis module being preferably configured to determine a predictive sleep state of the individual at a future time based on the sleep measurement, and/or a communication interface configured to receive behavior analysis information from a remote behavior database;

- the device comprises an electronic watch and/or an electronic bracelet and/or a tracker and/or a radio frequency transmitter.

According to a third aspect, the invention relates to a computer program product comprising code instructions which, when said instructions are executed by a processing unit, lead the processing unit to implement a method as defined below. Before.

According to a fourth aspect, the invention relates to computer-readable storage means on which are recorded code instructions which, when said instructions are executed by a processing unit, lead the processing unit to implement a method such as defined above.

GENERAL DESCRIPTION OF FIGURES

Other characteristics, objects and advantages of the invention will emerge from the description which follows, which is purely illustrative and not limiting, which must be read in conjunction with the appended drawings, among which:

FIG. 1 is a representative diagram of a body monitoring system integrating a body monitoring device according to an exemplary embodiment.

Figure 2 illustrates steps of a blood glucose monitoring method according to a first embodiment, which can be implemented by the system of Figure 1.

Figure 3 is a first example of a graph of the sleep state M of a watch wearer and of the predictive glycemia C of said watch wearer as a function of time T, in which the determined sleep state is an awake state or a sleeping state.

FIG. 4 is a second example of a graph of the sleep state M of a watch wearer and of the predictive glycemia C of said watch wearer as a function of time T, in which the determined sleep state is an awake state , a light sleep state or a deep sleep state.

Figure 5 illustrates steps of a blood glucose monitoring method according to a second embodiment, which can be implemented by the system of Figure 1.

FIG. 6 is a third example of a graph of the sleep state M of a watch wearer and of the predictive glycaemia C of said watch wearer as a function of time T, illustrating the adjustment of a range of normality N of blood sugar based on an awake state or a sleeping state of the wearer.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

The following description concerns the example of a CGM system comprising an electronic watch for monitoring the glycemia levels (blood glucose concentration) of an individual. The person skilled in the art will however understand that the method and the device of the invention would be applicable to any type of monitoring device, for example to an electronic bracelet, or even to a fixed device which would not be worn by the individual. The monitoring device can integrate components transported by the individual and components not transported by the individual.

The method and the device of the invention apply just as advantageously to the concentration of a bodily analyte other than glycaemia, or to any other bodily physical quantity to be monitored. Possible examples are lactate concentration or hydration, for example in an interstitial fluid of the individual.

In all of the appended figures and throughout the description below, similar elements bear identical alphanumeric references.

Body monitoring system

There is shown in Figure 1 a body monitoring system according to an example. The system of Figure 1 comprises in particular a body monitoring device (here an electronic watch 1) intended to be worn by an individual. In this example, the body monitoring system is a CGM (for “Continuous Glucose Monitoring”) system.

The system of Figure 1 is preferably intended for daily use at home, but this system can also be set up in a hospital setting.

The watch 1 comprises an electronic box comprising a plurality of electronic components for monitoring blood sugar and determining health alerts. As will be seen below, the housing preferably comprises a removable capsule comprising a blood glucose sensor with one or more needles, the capsule being mounted on the housing.

The watch case 1 further comprises data processing means 9, preferably one or more processors, as well as one or more memories 80 such as hard memories, flash memories, etc. For example, the data processing means 9 and the memory 80 are integrated into one or more microcontrollers of an electronic card arranged in the casing. The data processing means 9 are preprogrammed with code instructions. The execution of these instructions allows the data processing means 9 to implement a body monitoring method, for example a method according to the example of Figure 2 described below or according to the example of Figure 5 described below.

Advantageously, the data processing means 9 are further provided with a network communication interface, for example Wi-Fi, 3G, 4G, 5G, Bluetooth, etc.

The watch 1 further comprises a bracelet, preferably adjustable, for attachment to a limb of the individual. A display device, preferably a screen, is advantageously integrated into the watch to display blood glucose monitoring data and/or sleep measurement data and/or health alert data.

Preferably, watch 1 also includes a battery for powering the components of the watch. The battery is preferably rechargeable. Optionally, the system of Figure 1 may further comprise an electrical charging station (not shown) for charging the battery of watch 1.

The monitoring device could comprise, as an alternative or in combination with the electronic watch, another type of device worn by the individual whose blood sugar level is to be monitored, such as an electronic bracelet, a tracker, etc.

The watch 1 very advantageously incorporates a sleep sensor 4. The sleep sensor 4 is configured to preferably acquire in real time a physical measurement illustrating the state of sleep M of the wearer, referred to here as a sleep measurement.

An advantage is that the watch 1 can then operate independently, not only for determining the wearer's predictive glycaemia, but also for determining the state of sleep (present or future) of the wearer.

The sleep sensor 4 preferably comprises a heart rate sensor, and/or an accelerometer, and/or a gyrometer, and/or a respiratory rate sensor, and/or an electrocardiograph sensor, and/or an imaging device, and /or a temperature sensor (optionally an optical temperature sensor), and/or a sound level meter, and/or a sensor acquiring a blood oxygen level measurement (arterial oxygen saturation or pulsed oxygen saturation).

In the present example, the sleep measurement is provided by an accelerometer coupled to a gyrometer and/or to a heart rate sensor.

As will be seen below, the sleep state M is here determined in real time. Preferably, the sleep state M is determined in connection with a behavioral model of the sleep of the individual. The watch 1 can include for this purpose a behavioral analysis module 8, configured to determine the predictive sleep state M of the wearer (at one or more future instants) according to the measurement provided by the sleep sensor(s) 4 The behavioral model can be preprogrammed in the memory 80 of the watch 1 to be executed by the data processing means 9.

Alternatively, the behavioral model can be stored in a remote database (not shown), for example a database based on a cloud or “cloud” infrastructure. It is preferably a secure private database, managed for example by a health service provider. However, the database storing the behavioral analysis information could, alternatively, be a free or “open data” database. The watch 1 then comprises a communication interface, to receive the behavioral analysis information from the cloud. The watch can operate in “cloud-to-cloud data exchange” mode with said database.

In both cases, the sleep behavioral model of the individual is preferably implemented by an algorithm obtained by machine learning methods, for example an algorithm based on neural networks. The parameters of the algorithm for the implementation of the behavioral model were preferably obtained at the end of a training phase, before a practical use of the algorithm to calculate a sleep state M.

As an alternative or in combination, the watch 1 could be connected in a network to an external device supplying the measurement of sleep, or supplementing said measurement, for example to a video camera installed in the room of the individual and configured to visually determine the individual's sleep state (e.g. via eyelid position, facial detection, etc.). Other examples of external devices, which may or may not be coupled to the sensors of the watch, are the following: motion sensors positioned for example under the wearer's mattress and/or under a pillow of the wearer's bed, and/or camera( s) video, and/or Doppler camera(s), and/or infrared camera(s), and/or thermal camera(s).

Advantageously, the watch 1 further comprises a blood glucose sensor 5 configured to acquire, preferably in a completely autonomous manner, a measurement illustrating the concentration C of the body analyte of the individual (here, a blood glucose level of the interstitial liquid of the epidermis of the individual).

The blood glucose sensor 5 preferably comprises an electrochemical sensor with one or more needles. The blood glucose sensor 5 is for example in accordance with the description of international application WO 2019/141743 A1 in the name of the Applicant.

According to this example, the blood glucose sensor is integrated into a capsule configured to removably engage with the case of the watch 1, and comprises a plurality of micro-needles provided to be inserted through the surface layers of the epidermis of the individual, in an interstitial fluid of the individual. An electrochemical measurement of the blood sugar is carried out by the network of micro-needles, in association with the processing means of the watch 1. An advantage of this type of measurement is to be minimally invasive for the wearer, while translating reliable the individual's blood glucose level.

Alternatively, the predictive glycaemia C could be acquired in part or totally by an additional component connected in a network with the watch 1, then transmitted to the watch 1 via a communication interface. Optionally, the body monitoring system comprises for example an insulin pump (not illustrated in FIG. 1) having a direct connection or via the cloud with the watch 1. The insulin pump can provide real-time data the wearer's drug intake, useful for determining their predictive blood sugar levels C.

It should be noted that, optionally, an automatic learning model can also be used to determine the predictive concentration of body analyte (here the predictive glycaemia C) of the wearer. The data of said model can then be stored in the memory 80 of the watch 1, or even acquired from the cloud.

The watch 1 comprises an alert relevance determination module 6. This module is configured to obtain the predictive glycemia C of the wearer, for example from the memory 80, and to compare said predictive glycemia to a range of normality N. The module alert relevance determination 6 is further configured to receive a sleep measurement from the sleep sensor 4, and to obtain the wearer's predictive sleep state M. The sleep state is for example stored in the memory 80.

In addition, the watch 1 comprises an alert signal emission module 7. This module is configured to emit an alert signal, in the event that a health alert concerning the wearer's blood sugar level is determined to be relevant (determination preferably performed by the alert relevance determination module 6, during the implementation of the monitoring method).

The system of Figure 1 further comprises, optionally and advantageously, an electronic device belonging to an emergency contact of the individual (here a mobile phone 2), and / or an emergency server 3 belonging for example to an emergency service (hospital, rescue, fire department, etc.).

It will be understood, however, that the use of the CGM system should enable the wearer to monitor his blood sugar alone, so as to avoid having a third party and/or an emergency service having to take charge of the individual. The CGM system must, for example, allow the wearer to "re-sugar", that is to say to ingest foods containing glucose or to take a dose of glucose, or on the contrary to lower his glycemia, upstream of a predictive excursion of the wearer beyond a range of normal blood sugar levels.

As will be seen below, the alert signal SIG emitted by the alert signal emission module 7 can be intended for the wearer, and/or intended for a third individual, for example the owner of the mobile phone 2 or an emergency service operator. The owner of the mobile telephone 2 is for example a person present in the bearer's home or a person present in an emergency service such as a hospital service, able to take emergency measures.

Preferably, the watch 1 is connected to the mobile telephone 2 and to the emergency server 3, via one or more wireless communication networks (Internet, 3G, 4G, 5G, Bluetooth etc.). In the present example, the communication network is an Internet network I. The watch 1 is here connected via the mobile 2 to a cloud infrastructure service on the Internet network. The issuance of health alerts is preferably managed at least partially by said cloud service. Alternatively, watch 1 can be connected directly to the Internet if it has an integrated communication interface (for example Wi-Fi or 4G or 5G).

The monitoring system in Figure 1 can implement real-time blood glucose monitoring of the individual (painlessly and with low hygienic risk) as described below. It is preferably a stand-alone CGM system, which does not require additional equipment for blood glucose determination.

Blood Glucose Monitoring Method – Example 1

Figure 2 illustrates successive steps of a blood glucose monitoring method, according to one embodiment. The method 20 is for example implemented by the body monitoring device (here the electronic watch 1) as described above.

The method 20 comprises the following general steps:
- obtaining the sleep state M of the wearer, from a measurement acquired by the sleep sensor 4 of the watch 1,
- determination of the relevance of a health alert, the health alert being determined as relevant (and therefore intended to be sent to the wearer of the watch 1 and/or possibly to a third party individual or service) under the following two cumulative conditions: the predictive glycaemia C of the individual at a future instant is located outside a range of normality N, and the sleep state obtained is a state considered favorable for receiving the health alert,
- finally, if the health alert is determined as relevant in the previous step, emission of an alert signal by the watch 1.

The future instant at which the predictive glycaemia C passes outside the normality range N will be called “excursion instant” throughout the description below, and will bear the numerical reference t2.

We therefore take into account the sleep state of the wearer between the instant of implementation of the method and the instant of excursion t2, to establish the relevance of the health alert. An associated objective is to avoid unwanted alerts for the wearer and which would not be absolutely necessary, for example alerts which would wake up the wearer.

The details of several examples of implementation are described below.

• Obtaining predictive blood glucose

At an optional step 100, the body monitoring device (here the watch 1) is put into service. For example, the wearer puts the watch 1 back on his wrist, and/or recalibrates the blood sugar sensor 5. Preferably, once the watch 1 has been put in place and supplied with current, the following steps 200 to 500 are implemented in continued. The blood sugar level is constantly monitored, for example at regular intervals.

At an optional step 200, the body monitoring device acquires the predictive concentration of body analyte of the wearer, here the predictive glycemia C. The predictive concentration of body analyte is obtained for a plurality of future instants, according to the glycemia measured in real time and possibly other factors, according to any method known to those skilled in the art. Preferably, a predictive glycaemia curve C is obtained continuously, over an entire future time range.

The predictive glycaemia C may in particular depend on a behavioral glycemic model of the wearer. The wearer's behavioral glycemic model may take into account the blood glucose measurement observed in real time, and other factors such as the wearer's food intake, the wearer's drug intake (such as insulin), a level of wearer's activity, stress level, past blood sugar levels, etc. The predictive glycemia C can also depend on the sleep state of the wearer, obtained in real time or predicted for the future. The behavioral glycemic model can for example take into account the duration and/or the quality of sleep, during the current day and/or during past days, to predict the glycemic evolution.

Optionally, the behavioral glycemic model can integrate data from the mobile telephone and/or from an additional electronic device and/or from a preferably secure remote database (for example in “data exchange cloud- to-cloud”).

By way of example, the glycemic model data may include food intake data that has been entered on a dedicated mobile application or detected by a connected device (for example by automatic scanning of food packaging), and/or may understand drug intake data from a dedicated mobile application or a medical device such as an insulin pump (not illustrated in the appended figures).

For example, FIG. 3 shows a predictive glycemia C of the wearer of the watch 1 as a function of time T. The predictive glycemia C is obtained over a time range which extends from a present instant t0 (9 p.m.) until at a final instant t3 (7 a.m. the next day). Blood sugar tends to drop overnight.

Alternatively, several predictive glycaemia probability curves can be established, with more or less optimistic hypotheses of glycaemia evolution.

Alternatively or in combination with blood sugar, the analyte measured in step 200 may comprise wearer hydration and/or blood lactate concentration.

In the present example, the blood sugar sensor 5 is, as indicated above, a micro-needle sensor carrying out an electrochemical measurement in the interstitial liquid of the epidermis of the individual. In normal operation, the micro-needles remain permanently in the epidermis of the individual. They are only moved when changing the removable capsule. An advantage is that the individual is not obliged to carry out any manual operation (for example a puncture) to obtain the blood sugar.

More generally, the predictive glycaemia C is preferably established from a measurement supplied by a glycaemia sensor 5 incorporated in the monitoring device. Alternatively, the monitoring device receives the predictive blood glucose C from another separate device.

In certain cases, it is desirable that the predictive glycaemia C not be obtained from a measurement of the glycaemia sensor 5. In particular, in the case of a glycaemia sensor with one or more needles, it may happen that the needle(s) are not inserted correctly into the epidermis, or that the watch case has momentarily separated from the removable capsule which incorporates the sensor. In this case, provision may be made for the watch 1 to transmit an error message to the wearer or to a third party, to take account of the fact that the wearer could be in hypoglycaemia or hyperglycaemia during the period when the blood glucose measurement is dysfunctional.

Preferably, the method below is then implemented to determine whether the error message concerning the malfunction of the blood glucose measurement must be sent immediately, or whether it is necessary to postpone its sending (with regard to the state of wearer's sleep). The predictive glycaemia C is then determined without referring to the dysfunctional glycaemia measurement, for example by extrapolating from previous non-dysfunctional measurements, and/or by using other data (food intake, drug intake, etc.).

From the predictive glycaemia C, it is possible to determine an excursion instant t2 where the glycaemia becomes abnormal, for example by reference to one or more predetermined glycaemia thresholds. In the examples below, the predictive glycaemia C is notably compared to the limits of one or more "normality ranges" of the glycaemia. It will be understood that it is necessary to alert the wearer before this instant of excursion t2, so that the wearer takes corrective measures (for example to resugar) allowing to straighten the glycemia and to avoid the excursion.

A possible rule would be to send an alert signal to the wearer (for example sound and/or light) a certain time before the excursion time, for example 30 minutes before the excursion time.

However, in the present method, it is desired to adapt the instant at which the alert signal is received. We allow ourselves to adjust the moment when the wearer (or a third party) receives the alert signal. The adjustment of this moment of reception depends on a sleep state of the wearer.

• Determination of a sleep state

With reference to the graph of Figure 3, a first example of determining a sleep state of the wearer of the watch is described. The sleep state can be determined according to this first example in method 20.

The sleep state is determined from at least one sleep measurement. The sleep measurement is typically a physical measurement provided by the sleep sensor 4. Here, the sleep measurement is provided by an accelerometer coupled to a gyrometer and/or to a heart rate sensor. For example, wearer movement data is associated with wearer body temperature data to determine an awake state or a sleepy state of the wearer. This physical measurement can be associated with measurements provided by other devices, for example motion sensors positioned for example under the wearer's mattress and/or video cameras, as indicated above.

The CGM system is able to determine the sleep state of the wearer at the present time, here at time t0. Additionally, the CGM system is preferably capable of obtaining predictive sleep states M of the wearer at a plurality of future times.

The predictive sleep state M is advantageously obtained for a plurality of future instants, here over a time range which extends from the present instant t0 (9 p.m.) until the final instant t3 (7 a.m. the next day). Preferably, a curve of the predictive sleep state M is obtained continuously, over the entire time range.

In the example of Figure 3, the predictive sleep state varies mainly between two sleep state levels: a level M1 corresponding to the awake state of the wearer, and a level M2 corresponding to the sleeping state of the wearer .

In said example, the wearer is in an awake state from 9 p.m. to 10 p.m. Then, a falling asleep phase extends between about 10 p.m. and about 10:30 p.m. From 10 p.m., the wearer settles into bed, his level of movement decreases and his heart rate decreases. At 10:15 p.m., it is considered that watch 1 is able to detect that the wearer is falling asleep. At 10:30 p.m., at the end of this falling asleep phase, the wearer reaches the asleep state (until around 6:30 a.m., an awake state being regained around 7 a.m.).

Very advantageously, the predictive sleep state of the wearer can be determined according to a behavioral model (managed for example by the behavioral analysis module 8 of the watch 1). The behavioral model is very preferably a model specific to the wearer, possibly trained using behavioral data specific to the wearer. The behavioral data of the individual can be recorded directly in his body monitoring device, or even recorded securely in a remote server. The remote server (not shown) belongs for example to a health service provider.

By combining the sleep state of the wearer measured in real time with other factors, the behavioral model can predict the evolution of the sleep state. The behavioral model associated with the wearer comprises for example a probability distribution of sleep duration of the wearer, and/or a probability distribution of time of falling asleep of the wearer, and/or a probability distribution of time of awakening of the wearer . Thus, the behavioral model can predict the future sleep state from the sleep habits of the wearer. The behavioral model aims in particular to translate recurring patterns observed in the sleep habits of the wearer.

It is also possible to integrate circumstantial data for determining the state of sleep, such as an alarm time programmed for a wake-up signal (in particular in the case where the watch 1 also performs a wake-up function), and/ or calendar data accessible in a database concerning activities of the wearer, and/or expected times of falling asleep or waking entered in advance.

Preferably, the behavioral model for determining sleep states only takes into account sleep of long duration, for example of duration greater than or equal to 2 hours. Thus, the behavioral model advantageously does not take into account naps or other brief drowsiness. An advantage is that the switching of the watch 1 to a sleep mode (regarding, for example, the normal blood sugar range, as will be seen below) is carried out for a long sleep, which avoids warning in advance the wearer of an excursion that could be treated downstream from the siesta.

It will be noted that the behavioral model of the wearer can in particular vary according to the periods, and for example according to the day of the week. For example, if sleep measurements over previous weeks have shown that the wearer tends to get up early every other Sunday, and to get up two hours later the other Sunday, the wearer's behavioral model may reflect this trend. The behavioral model can also integrate sleep data (anonymised) from an entire population, in order to predict, for example, the wearer's sleep state according to parameters such as the wearer's age or gender, the number of steps taken in the day, etc. The behavioral model can optionally integrate predictive activity data of the wearer, which makes it possible to deduce predictions on his sleep state. For example, the behavioral model can integrate carrier diary data, carrier appointments, non-working days/holidays, etc.

The behavioral model may further assist in interpreting the measurements acquired by the sleep sensor 4. For example, the behavioral model includes correlation data, mapping a plurality of ranges for the measurement acquired by the sleep sensor 4 and the different states of sleep.

For exemple :
- a high range of detected heart rate and/or acceleration can be associated with the awake state,
- and a low range of detected heart rate and/or acceleration can be associated with the sleeping state.

Advantageously, the correlation data can allow the detection in advance of a phase of falling asleep and/or pre-falling asleep of the wearer. As an alternative or in combination, the correlation data can comprise a signal for detecting a position of the wearer, for example a position in his bed or out of bed. This signal can come from one or more motion sensors positioned for example under the wearer's mattress, and/or from one or more video or Doppler camera(s), etc.

Additionally, the behavioral model may advantageously include a sleep pattern, in which a decrease in heart rate and/or a near-cancellation of acceleration is associated with falling asleep. Falling asleep can last longer or shorter depending on the wearer's profile. Thus, once the heart rate and/or acceleration begins to decrease, the monitoring device is able to predict when the wearer will reach the sleeping state.

Very preferably, the different sleep states M determined are independent of the predictive glycaemia C. The sleep states here depend solely on the sleep measurements provided by the sensors and/or on the behavioral model.

Returning to the method 20 of Figure 2, at a step 300a, the predictive sleep state M of the wearer of the watch 1 is determined during a period which precedes the instant of excursion t2. It is thus sought to establish whether the bearer will be willing to receive an alert during said period. For example, we observe the predictive sleep state M between time [t2 minus 30 minutes] and time t2.

In the example of Figure 3, the excursion time t2 occurs at 2 o'clock. From this moment of excursion, the normality range N′ is exceeded. The wearer is predicted to be at level M2 of the sleep state (asleep state) between time [t2 minus 30 minutes] and time t2. As will be seen below, it is preferred to alert the wearer at an earlier time.

Referring to the graph of Figure 4 , there is illustrated a second example of determining a sleep state of the wearer of the watch 1. During the method 20, the sleep state can be determined in accordance with the second example.

The predictive glycaemia C has here been determined in accordance with the preceding description. With regard to the predictive sleep state M, the general principles described above apply. But unlike the first example in Figure 3, there are three levels of sleep in this second example: a level M'1 corresponding to the awake state, a level M'2 corresponding to the state of light sleep and a level M '3 corresponding to the state of deep sleep.

The falling asleep phase (between 10:30 p.m. and 11:00 p.m. in this second example in Figure 4) thus corresponds to the transition of the wearer between the awake state and the light sleep state. This sleep onset phase is modeled as an affine function (smooth transition between the awake state and the light sleep state). Conversely, the different passages between the state of light sleep and the state of heavy sleep (here at 1h20, 3h, 4h40) are here modeled by a staircase function.

It will be understood that any number of sleep states greater than or equal to 2 can be defined for the modeling of the sleep of the individual. For example, you can choose any number of sleep states from among: awake, falling asleep, light slow sleep, deep slow sleep, paradoxical sleep.

To take account of the alternative modeling of the sleep states in the second example of FIG. 4, if the determination of the predictive sleep state M depends on a behavioral model, the latter may advantageously comprise a probability distribution of the duration the individual's deep sleep duration, and/or a probability distribution of the individual's light sleep duration.

One advantage is to have a more precise image of the state of sleep of the wearer of watch 1, during the period which precedes the instant of excursion t2.

• Determining the Relevance of a Blood Glucose Health Alert

It is determined in a step 400, taking into account the excursion time t2, when a health alert is relevant. It is recalled that watch 1 is programmed to trigger a health alert under the following two cumulative conditions:
- the predictive glycaemia C associated with the wearer at the future time is located outside the normal range,
- the sleep state M of the wearer is a state deemed conducive to receiving the alert.

At the time of excursion t2, the predictive glycaemia C here exceeds the normality range.

In Figure 3 and in Figure 4, a distinction is made between a range of normality of wakefulness N and a range of normality of sleep N'.

The upper limits of the two ranges (hyperglycemia threshold) are identical here. It should be noted that alternatively, the hyperglycemia thresholds associated with the range of normality of awakening N and the range of normality of sleep N' could be distinct.

On the other hand, the lower limit of the normal wakefulness range N (wake hypoglycaemia threshold) is strictly greater than the lower limit of the normal sleep range N′ (sleep hypoglycaemia threshold). This rule reflects the fact that variability in a low glycemic region may be permitted during sleep. Blood glucose values between the awake hypoglycemia threshold and the sleep hypoglycemia threshold are considered acceptable during sleep. Alternatively, these two hypoglycemia thresholds could be identical.

Preferably, the lower limit of the range of normality of awakening is higher by at least 10 milligrams per deciliter of blood compared to the lower limit of the range of normality of sleep. In the present example, the lower limit of the range of normality of wakefulness N is equal to 80 milligrams per deciliter of blood, and the lower limit of the range of normality of sleep N' is equal to 65 milligrams per deciliter of blood. It will be noted that one and/or the other of these lower glycemia limits could be adjustable by the user or by a health practitioner, for example with a user interface of the watch 1.

Here, the glycemia normality range is adjusted according to the predictive sleep state M. Thus, if the sleep measurement makes it possible to detect that a sleep phase of the wearer is in progress, then the normality range can be adjusted accordingly.

It will be noted that the range of normality could also be adjusted more finely, for example according to a state of light sleep or deep sleep of the wearer.

Possibly, the range of normality is “moving” according to the state of sleep; for example, the lower limit of the normality range (hypoglycemia threshold) can be shifted downwards if the planned excursion time t2 is close to the planned wakefulness time t3. Thus, the health alert can be postponed until the expected time of awakening t3.

With regard to the second condition examined in step 400 (the sleep state M obtained from the wearer is a state deemed suitable for receiving the alert), this determination depends on the model adopted for the sleep of the wearer:
- in Figure 3, the waking state (level M1) is considered conducive to receiving the alert, while the end of the sleep phase and the asleep state (level M2) are not conducive;
- in Figure 4, the state of wakefulness (level M'1) as well as the state of light sleep (level M'2) are deemed favorable, whereas the end of the sleep phase and the state asleep (level M'3) are not conducive.

Any other rule is possible for the determination of sleep states favorable or unfavorable to the emission of an alert.

Thus, in step 400, if the predictive sleep state M during the period which precedes the instant of excursion t2 (for example between the instant [t2 minus 30 minutes] and the instant t2) n' is not conducive to receiving the health alert, then watch 1 does not program a health alert during this period.

Conversely, if the state of sleep before the excursion time t2 is conducive to receiving the health alert, then the health alert is programmed directly before the excursion time t2. For example, if the excursion is scheduled less than 30 minutes after falling asleep, then it is relevant to issue the alert 30 minutes before the excursion.

It should be noted that exceptions to the principles described above are possible. For example, if the predictive glycemia C at the end of the night is below a critical level for hypoglycemia or above a critical level for hyperglycemia, the wearer can be considered to waking up during the night because of his hypoglycaemia/hyperglycaemia. It is then possible to transmit a health alert despite the sleeping state of the wearer. It will be understood that a balance must be achieved between avoiding unwanted health alerts for the wearer, and the assurance of alerting the wearer wisely to avoid complications.

In a scenario where the blood glucose sensor 5 is temporarily dysfunctional, the alert signal may include the error message that a problem has been detected at the level of the sensor, to invite the wearer or a third party to put it back into operation. the sensor. It is then advantageous to take into account the predictive glycaemia C of the wearer, to assess whether there is a risk of hypoglycaemia or hyperglycaemia during the estimated period when the wearer goes to sleep and where there will be no reliable blood sugar measurement. . If there is no imminent risk that the blood sugar will exceed critical thresholds, it is possible to wait for a moment when the state of sleep is deemed favorable before sending this error message.

Preferably, if it has been determined in step 400 that the predictive sleep state M during the period preceding the instant of excursion t2 is not conducive to receiving the alert, it is determined at a additional step 410 a time prior to the excursion time t2 and suitable for issuing the health alert. This earlier instant will be denoted "alert instant" in the description below, and will bear the numerical reference t1.

Indeed, it would be dangerous not to inform the wearer of the planned excursion.

Advantageously, the alert time t1 is a time when the predictive sleep state M is deemed suitable for receiving the alert. Preferably, the alert time t1 is equal to the time prior to t2 and closest to t2, such that the sleep state at t1 is deemed conducive to receiving the alert.

In the example of Figure 3, the alert time t1 is located at the beginning of the sleep phase (at approximately 10:15 p.m.) of the wearer. Indeed, at the end of the falling asleep phase, the wearer is in a sleeping state and we do not want to interrupt his sleep. In this case, a health alert that occurs during the individual's sleep phase would be considered irrelevant. By setting an alert time t1 during the falling asleep phase, the wearer can return to a completely awake state at 10:15 p.m., resugar and then resume falling asleep.

In the example of Figure 4, the alert time t1 is near the end of the first phase of light sleep. Indeed, the instant of excursion t2 is planned during the deep sleep phase which immediately follows this first phase of light sleep.

Thus, an advantage of the sleep modeling adopted in this last example is to be able to identify an alert instant t1 closer to the excursion instant t2, without interrupting the deep sleep phase which is important for recovery. In this case, a health alert that occurs during the individual's deep sleep phase would be considered irrelevant.

It will be noted that, optionally, the predictive concentration of bodily analyte can depend on a number of alert levels equal to at least three. For example, four alert levels are defined (critical hypoglycemia, mild hypoglycemia, mild hyperglycemia, critical hyperglycemia). It can then be determined that the health alert is relevant and urgent, without necessarily taking into account the predictive sleep state M of the wearer, if predictive glycaemia C exceeds one of the critical levels.

For example, an emergency contact owner of the mobile telephone 2 and/or an emergency service 3 is notified, in the event of an excursion of the predictive glycaemia C of the wearer beyond a criticality range (formed by the critical hypoglycemia level and critical hyperglycemia level, not shown in Figure 3 and Figure 4). In the latter case, the time between the alert instant t1 and the excursion instant t2 can be adapted so that the emergency contact or the emergency service can intervene.

If the Predictive Glucose C stays within the normal range along the entire curve, then no health alert is transmitted.

• Transmission of an alert signal

Returning to the method 20 of Figure 2, the body monitoring device finally issues the health alert if applicable, at a step 500. The health alert is for example issued at the alert time t1 determined at the end of step 410, or directly before the excursion time t2.

The health alert corresponds to the SIG alert signal illustrated in Figures 3 and 4. This alert should allow the wearer or the wearer's entourage to take corrective measures, in order to avoid an excursion of the blood sugar level out of range. the normal range.

For example, the alert signal SIG comprises an audible signal and/or a visual signal and/or a tactile signal, emitted by the watch 1 to the wearer. In the present example, the SIG alert signal combines an audible signal (to attract the wearer's attention) and, preferably, a visual signal on the screen of the watch 1 (to inform the wearer of his predictive blood sugar level) .

One advantage is to draw the wearer's attention to the risk of hypoglycaemia or hyperglycaemia, even if the wearer does not take the initiative to regularly consult the blood sugar monitoring information on his monitoring device ( i.e. on the watch 1) or on another device connected to the monitoring device.

Alternatively or in combination, the SIG alert signal comprises a communication signal to a third-party device, for example the mobile telephone 2 of the emergency contact (which is for example a person present at the same home as the wearer) and/ or server 3 of an emergency service. This solution may be particularly relevant if the wearer has limited autonomy, and may need help to resuscitate.

It is also possible to provide for connecting the watch 1 to other electronic devices present in the environment of the wearer or of a third party, to carry out the alert intended for the wearer or the third party. For example, the SIG alert signal may include a command for a lighting device in the wearer's bedroom, such as one or more remotely controllable connected lights. The luminous device can adapt its luminous intensity so as to warn the wearer. If the wearer is in a light sleep phase, a soft sound and/or light signal may be sufficient.

Blood Glucose Monitoring Method – Example 2

Figure 5 illustrates successive steps of a method 50 for monitoring blood sugar, according to an alternative embodiment. The method 50 can also be implemented by the body monitoring device (electronic watch 1) as described above. The optional steps 100 and 200, as well as the alert signal transmission step 500, are identical to the description above associated with the method of Figure 2.

In step 300b, the watch 1 acquires the wearer's predictive sleep state M over a series of future instants comprised between the present instant t0 and the expected waking instant t3.

Next, in step 400, the predictive glycaemia C is compared to the range of normality (ranges N and N′) during the various future instants comprised between the present instant t0 and the predicted instant of awakening t3, in order determine the relevance of a health alert.

In this example, the knowledge of predictive sleep states is only used to adapt the normal range of blood glucose, and is not used to determine the right time to generate a SIG alert signal.

Finally, if the wearer's sleep state M at the present time t0 is conducive to receiving the health alert, and if an excursion is scheduled at a future excursion time t2 beyond the normal range, then the watch 1 implements in step 500 the transmission of the alert signal at the present time t0.

There is shown in Figure 6 an example of determining the relevance of a health alert, in accordance with the method 50.

It is assumed here that the sleep states previously described in relation to Figure 3 (asleep state reached around 11 p.m., return to the waking state around 7 a.m.) correspond to a habitual sleep pattern of the wearer. Thus, usually, watch 1 compares the predictive blood glucose level to a range of normality of wakefulness N, except between 11 p.m. and 7 a.m. when watch 1 usually uses the normality range of sleep N'.

The predictive glycaemia curve C is here identical to the predictive glycaemia curves of Figures 3 and 4. On the other hand, the state of sleep M between the present instant t0 and the expected waking instant t3 differs on the example of Figure 6, compared to Figures 3 and 4. In other words, the wearer's sleep pattern in Figure 6 is distinct from the wearer's usual sleep pattern.

Given the wearer's usual sleep pattern (asleep state reached around 11:00 p.m., return to the waking state around 7:00 a.m.), the watch 1 can initially determine a normality range N′ (sleep normality range ) between 11 p.m. and 2 a.m. This determination is illustrated by the hatched rectangle R in Figure 6. Using the normality range N', a health alert is not necessary, because the predictive blood glucose curve C remains contained within the rectangle R and it does not there is no hypoglycemia.

However, the real sleep states of the wearer correspond here to a sleeping state reached around 2 a.m., and to a return to the waking state around 7 a.m. Using the sleep measurement from the sleep sensor 4, the watch 1 here readjusts the normality range in real time.

Thus, watch 1 then determines a range of normality N (range of normality of awakening) between 11 p.m. and 2 a.m. The readjustment of the normality range has the effect of creating a new excursion instant t2, here at around 11:15 p.m.

It will be understood that, more generally, if the state of sleep at a given instant varies with respect to a predictive sleep state previously calculated or with respect to a habitual sleep pattern of the wearer, then a position of the predictive glycemia within or outside the normal range may also vary. This situation can occur in particular if the wearer goes to bed and/or gets up earlier or later than usual (case of Figure 6), or even if the wearer gets up unexpectedly.

Finally, the sleep state of the wearer at the present time t0 of readjustment of the range of normality (at approximately 10:45 p.m. in Figure 6) is an awake state. The waking state is considered conducive to receiving the health alert. The SIG alert signal is therefore emitted directly at the present time t0. In this example, the body monitoring device therefore takes into account the wearer's usual sleep pattern and also real-time sleep measurements, to determine the relevance of the health alerts.

Claims (25)

Method for monitoring the body analyte concentration in an individual, the method being implemented by a processing unit of a body monitoring device (1) associated with a sleep sensor (4), and comprising the steps of:
obtaining (300a, 300b) a sleep state of the individual, from a measurement acquired by the sleep sensor (4),
determination (400) of relevance of a health alert, the health alert being determined as relevant under the following two cumulative conditions:
- a predictive concentration (C) of bodily analyte associated with the individual at a future instant (t2) is located outside of a range of normality (N);
- the sleep state obtained from the individual is a state conducive to receiving the health alert,
and, if the health alert is determined to be relevant, issuing (500) an alert signal by the body monitoring device (1). A monitoring method according to claim 1, wherein the obtained sleep state of the individual is a predictive sleep state (M) at the future time (t2). Monitoring method according to claim 2, comprising an additional step of calculating (410) an alert instant (t1) prior to said future instant (t2), a predictive sleep state of the individual at the instant of alert (t1) being considered more conducive to receiving the health alert than the predictive sleep state (S) of the individual at the future instant (t2). A monitoring method according to any of claims 2 or 3, wherein the predictive sleep state (M) at the future time (t2) is a deep sleep state and/or an asleep state. A monitoring method according to any one of claims 3 or 4, wherein the predictive sleep state (M) at the alert time (t1) is a light sleep state or an awake state. A method of monitoring according to any one of claims 1 to 5, in which the sleep state of the individual is determined as a function of a behavioral model associated with the individual, the behavioral model being obtained by a module (8 ) behavior analysis integrated into the body analysis device (1), or being obtained from a remote behavior database. A method of monitoring according to claim 6, wherein the behavioral model associated with the individual comprises a probability distribution of the individual's sleep duration, and/or a probability distribution of the individual's sleep time, and/or an individual's waking time probability distribution, and/or an individual's deep sleep duration probability distribution, and/or an individual's light sleep duration probability distribution . A monitoring method according to any of claims 6 or 7, wherein the behavioral model associated with the individual comprises correlation data, said correlation data mapping a plurality of measurement ranges for the measurement acquired by the sensor (4) and a plurality of sleep states. Monitoring method according to any one of Claims 1 to 8, in which the predictive concentration (C) of bodily analyte is obtained for a plurality of instants, the range of normality (N) being variable as a function of said instants. A monitoring method according to claim 9 in combination with claim 2, wherein the range of normality (N) associated with the future time is determined as a function of the predicted sleep state at the future time (t2). A monitoring method according to any of claims 1 to 10, the method comprising obtaining (300b) an actual sleep state at a present time (t0), and comprising transmitting (500) the alert signal at the present time (t0) if the actual sleep state at the present time (t0) is conducive to the health alert. Monitoring method according to any one of claims 1 to 11, comprising an additional step of acquisition (200), by the body monitoring device (1), of the predictive concentration (C) of body analyte associated with the individual at the future instant (t2),
said predictive concentration (C) of body analyte being preferably obtained independently of the sleep state (M) of the individual. Monitoring method according to claim 12, in which the predictive concentration (C) of bodily analyte is determined as a function of a concentration measurement obtained by an additional sensor (5), the additional sensor (5) preferably being embedded in the body monitoring device (1). Monitoring method according to claim 13, comprising the detection of a concentration measurement error at the additional sensor (5), the alert signal being provided to notify the concentration measurement error to the individual or a third. Monitoring method according to any one of Claims 1 to 14, in which the alert signal comprises an audible signal and/or a visual signal and/or a tactile signal, emitted by the body monitoring device (1) to the destination of the individual. A monitoring method according to any one of claims 1 to 15, wherein the alert signal comprises a communication signal to a third party device, the third party device comprising a personal mobile device (2) and/or a home appliance and/or or an emergency service server (3). Monitoring method according to any one of claims 1 to 16, in which the measurement acquired by the sleep sensor (4) comprises a measurement of heart rate, and/or an accelerometer measurement, and/or a measurement of impedance by electrocardiogram, and/or a gyrometer measurement, and/or a visual measurement of the sleepiness of the individual, and/or a measurement of the respiratory rate. A monitoring method according to any of claims 1 to 17, wherein the predictive body analyte concentration comprises a predictive blood glucose concentration, and/or a predictive blood lactate concentration. Body monitoring device (1), said device (1) being adapted to be worn by an individual and being configured to implement a monitoring method according to any one of claims 1 to 18, the device (1) comprising:
an alert relevance determination module (6), configured to compare a predictive concentration (C) of bodily analyte associated with the individual with a range of normality (N), the relevance determination module (6) alert being further configured to receive a sleep measurement from a sleep sensor (4),
and an alert signal output module (7), configured to output an alert signal if the alert relevance determination module (6) determines that a health alert is relevant. Monitoring device (1) according to claim 19, the sleep sensor (4) being integrated into the monitoring device (1), the sleep sensor (4) preferably comprising a heart rate sensor, and/or an accelerometer, and/or a gyrometer, and/or a respiratory rate sensor, and/or an electrocardiograph sensor, and/or an imaging device. Monitoring device (1) according to any one of claims 19 or 20, the device (1) further comprising an additional sensor (5) configured to acquire a concentration measurement (C) of body analyte in real time, the sensor additional (5) preferably comprising an electrochemical needle sensor. A monitoring device (1) according to any one of claims 19 to 21, the device (1) further comprising:
- a behavioral analysis module (8), comprising a memory in which a behavioral model associated with the individual is stored, the behavioral analysis module (8) being preferably configured to determine a predictive sleep state (M) of the individual at a future time (t2) based on the sleep measurement, and/or
- a communication interface configured to receive behavior analysis information from a remote behavior database. Monitoring device (1) according to any one of Claims 19 to 22, the device (1) comprising an electronic watch and/or an electronic bracelet. Computer program product comprising code instructions which, when said instructions are executed by a processing unit, cause the processing unit to implement a monitoring method according to any one of claims 1 to 18. Computer-readable storage means on which are recorded code instructions which, when said instructions are executed by a processing unit, cause the processing unit to implement a monitoring method according to any one of claims 1 to 18.