Computerized method and system for determining a nutritional recommendation.

Abstract

The computerized process for determining a personalized nutritional recommendation for a user uses temporal data on the user's metabolite concentrations obtained from in vivo measurements taken regularly over time in the user's interstitial fluid by at least one wearable sensor worn by the user, including for glucose, non-esterified fatty acids, glycerol, and / or ketone bodies. A computerized recommendation module selects a recommendation suitable for the user based on the temporal data by applying rules that associate recommendations from the list with temporal metabolite concentration data. Figure for the abstract: Figure 5

Description

Title of the invention: Computerized method and system for determining a nutritional recommendation. FIELD OF THE INVENTION

[0001] The present invention relates to computerized methods and systems for determining a nutritional recommendation. TECHNOLOGICAL BACKGROUND

[0002] In the field of nutrition, a person can regularly consult a specialized nutritionist to obtain recommendations on their nutrition, with the aim of achieving one or more objectives, such as losing weight, or not gaining fat mass.

[0003] There is, however, a difficulty in that the specialized nutritionist has access to little information concerning the person who consults him. This little information is essentially declarative, so that the nutritionist is dependent on the goodwill of the person who communicates the information to him. Despite this proactive approach of consulting a specialized nutritionist, this person may, consciously or unconsciously, not provide very accurate information on the reality of his nutritional practice.

[0004] Furthermore, given that the appointments are spaced out over time, it is difficult for the nutritionist to assess whether the recommendation has actually been followed by the person, how seriously, and for how long.

[0005] Furthermore, given that nutritional metabolisms are very complex, and potentially very different from one person to another, and can be very dependent on exogenous factors such as stress or a sleep problem for example, it is very difficult for the nutritionist to provide recommendations that are truly adapted to the person.

[0006] Patches or wearables have recently been proposed for real-time monitoring of blood glucose levels. A person places the patch on their skin, and the sensor makes a repeated estimate, for example several times per hour, of the blood glucose level. Such sensors have been developed primarily for diabetic patients.

[0007] For diabetic patients, such sensors can be used, and the measured data can be processed, to recommend an amount of insulin or glucagon to be injected. However, this process does not lead to a change in the patient's diet.

[0008] However, the technology has become more widespread, and such sensors can be now used, with varying degrees of accuracy, by everyone or under prescription.

[0009] The invention thus aims to propose a nutritional coach which makes it possible to influence a user's nutritional diet effectively and over the long term. Summary of the invention

[0010] Thus, the invention relates to a computerized method for determining a personalized nutritional recommendation for a user in which, having temporal data on the concentration of the user's metabolites obtained by in vivo measurements carried out regularly over time in a non-surgical manner in the interstitial fluid of the user by at least one portable sensor worn by the user, comprising at least temporal data on the concentration of at least two of the metabolites from the following list: {glucose, non-esterified fatty acids, glycerol, ketone bodies] in the interstitial fluid of the user, a computerized recommendation module selects at least one recommendation adapted to the user from a list of recommendations comprising several recommendations, as a function at least of said temporal data on the concentration of metabolites,using at least said temporal metabolite concentration data, by applying rules associating the recommendations from the list of recommendations with temporal metabolite concentration data.

[0011] Thanks to these provisions, it is possible to frequently determine a recommendation that is truly adapted to the user, based on numerous, good-quality data.

[0012] According to the embodiments, at least one group of metabolites from the following list is used:

[0013] - glucose and non-esterified fatty acids,

[0014] - glucose and glycerol,

[0015] - glucose and ketone bodies,

[0016] - non-esterified fatty acids and glycerol,

[0017] - non-esterified fatty acids and ketone bodies,

[0018] - glycerol and ketone bodies,

[0019] - glucose, unesterified fatty acids and glycerol,

[0020] - glucose, unesterified fatty acids and ketone bodies,

[0021] - glucose, glycerol and ketone bodies,

[0022] - non-esterified fatty acids, glycerol and ketone bodies,

[0023] glucose, unesterified fatty acids, glycerol and ketone bodies.

[0024] According to different aspects, it is possible to provide one and / or the other of the characteristics below taken alone or in combination.

[0025] According to one embodiment, a computerized characterization module determines a composite index of the user from said temporal data of concentration of metabolites of the user, and the computerized recommendation module selects said at least one recommendation adapted to the user by using said composite index, by application of rules associating the recommendations of the list of recommendations with composite indices.

[0026] According to one embodiment, the computerized recommendation module determines a relevance score of at least a plurality of recommendations in the list of recommendations, and selects said recommendation adapted to the user according to said relevance scores.

[0027] According to one embodiment, the computerized recommendation module determines the relevance score further based on a goal to be achieved defined by the user, in which said rules associate the recommendations from the list of recommendations with temporal data of metabolite concentration and with goals to be achieved.

[0028] According to one embodiment, the computerized recommendation module determines the relevance score of a recommendation by determining a prediction of a value of a control parameter by simulating an absence of recommendation follow-up, and a prediction of said value of the control parameter by simulating an implementation of said recommendation from at least said temporal metabolite concentration data.

[0029] According to one embodiment, the computerized recommendation module selects said at least one suitable recommendation from, in addition, a history of recommendations for the user.

[0030] According to one embodiment, the computerized recommendation module selects the recommendation adapted to the user based in addition on at least data relating to the identity, tastes, and / or constraints of the user entered by the user via a human-machine interface.

[0031] According to one embodiment, the computerized recommendation module selects the recommendation adapted to the user based further on diet, health, physical exercise and / or sleep data, as well as times associated with this data, entered by the user via a human-machine interface.

[0032] According to one embodiment, said at least one, in particular at least two, suitable recommendations are transmitted to the user, and the user selects a recommendation from among said at least one, in particular at least two, suitable recommendations.

[0033] According to one embodiment, the computerized method further comprises a phase of observation of the user using said portable sensor prior to the selection of said at least one suitable recommendation.

[0034] According to another aspect, the invention relates to a set of computer programs comprising instructions adapted for implementing the above method, when the set of computer programs is executed by one or more computers,

[0035] According to another aspect, the invention relates to a computerized system for determining a personalized nutritional recommendation for a user, comprising: . at least one portable sensor worn by the user and adapted to perform in vivo measurements regularly over time in the interstitial fluid of the user to produce temporal data of concentration of metabolites of the user, comprising at least temporal data of concentration of at least two of the metabolites from the following list: {glucose, non-esterified fatty acids, glycerol, ketone bodies] in the interstitial fluid of the user, . a computerized recommendation module adapted to select at least one recommendation adapted to the user from a list of recommendations comprising several recommendations, based at least on said temporal metabolite concentration data, by applying rules associating the recommendations from the list of recommendations with temporal metabolite concentration data, Brief description of the drawings

[0036] Embodiments of the invention will be described below with reference to the drawings, briefly described below:

[0037] [Fig-1] schematically represents a side view of a detection system 1.

[0038] [Fig.2] schematically represents in top view an electronic system of the detection system 1.

[0039] [Fig.3] schematically represents a portable communicating device.

[0040] [Fig.4] schematically represents a server.

[0041] [Fig.5] schematically represents a display screen of the portable device com communicating.

[0042] [Fig.6] schematically represents a fat storage / release curve for the user depending on time.

[0043] In the drawings, like references designate identical or similar objects. DETAILED DESCRIPTION

[0044] The invention relates to a computerized system. As shown in [Fig.l], according to an exemplary embodiment, the computerized system comprises a detection system 1. The detection system 1 is of the type which is worn (the Anglicism “wearable”, which does not does not have a satisfactory French translation, is most suitable for characterizing the detection system 1) on the user's skin 2. The detection system 1 can be applied to any suitable location on the user's skin, for example on the forearm or the stomach. The detection system 1 is assembled to the user's skin 2, non-surgically, so that it can remain in position for at least a few days. The assembly to the user's skin 2 is for example carried out by gluing. The detection system 1 then comprises a lower surface 3, intended to be turned towards the user's skin 2, and comprising a layer of glue 4 or other adhesive adhering to the user's skin 2.

[0045] The detection system 1 can thus take the form of a “patch” applied to the skin 2 of the user. [Fig. 1] represents the detection system 1 in the configuration worn by the user. The detection system 1 can be provided with a peelable film (not shown) assembled to the adhesive layer 4, and which can be removed in order to assemble the detection system 1 to the skin.

[0046] The detection system 1 also comprises microneedles 5. The microneedles 5 have a generally elongated shape, and protrude from the lower surface 3 of the detection system 1. The microneedles 5 may be arranged parallel to each other and distributed over the lower surface 3. Thus, when the detection system 1 is worn by the user, as shown in [Fig.l], the microneedles 5 penetrate the skin 2 of the user. Typically, the length of the microneedles 5 is of the order of 0.15 to 1.5 millimeters (mm), so that they are in contact with the interstitial fluid of the user's skin.

[0047] At least a portion of the microneedles 5 is adapted to measure a concentration of an analyte of the interstitial fluid by an electrochemical method. These analytes of the interstitial fluid are also called metabolites. A microneedle 5 may typically comprise a ligand complementary to the analyte to be measured, and a transducer. The ligand comprises a chemical species complementary to the analyte to be measured, such that the chemical reaction between the analyte to be measured and the ligand generates a modification of a physicochemical signal. The transducer converts this signal into an electrical signal.

[0048] In the context of the present invention, the detection system 1 is adapted to detect several distinct analytes. This can be achieved by using several microneedles 5, the microneedles 5 in question using different ligands and transducers, each adapted to an analyte to be detected. According to an exemplary embodiment, one of the analytes detected by the detection system 1 is the glucose of the interstitial fluid. For example, the detection of glucose is carried out by a glucose oxidase reacting with the glucose of the interstitial fluid to generate hydrogen peroxide H2O2, which is electrochemically detectable. According to an exemplary embodiment, one of the analytes detected by detection system 1 are non-esterified fatty acids, generally referred to by the English acronym "NEFA", for "non-esterified fatty acids"). For example, the detection of non-esterified fatty acids is achieved using coenzyme A, adenosine triphosphate (ATP), acyl-coenzyme A synthetase and acyl-coenzyme A oxidase, which react with NEFAs to generate hydrogen peroxide H2O2, which is electrochemically detectable.

[0049] According to an exemplary embodiment, one of the analytes detected by the detection system 1 is glycerol. For example, the detection of glycerol is carried out using a glycerol kinase + glycerol-3P-oxidase + ATP, or a glycerol oxidase, which react with glycerol to generate hydrogen peroxide H2O2, which is detectable electrochemically.

[0050] According to another exemplary embodiment, one of the analytes detected by the detection system 1 is ketone bodies, in particular acetoacetate, [3-hydroxybutyrate] and / or acetone.

[0051] This analyte is for example detected via enzymes and co-factors which interact with it to generate a nicotinamide adenine dinucleotide hydride (NADH), which is detectable electrochemically.

[0052] The detection system 1 thus comprises an electronic system 6. The electronic system 6, as shown in more detail in [Fig. 2], may for example comprise an electronic card carrying the various electrical components of the electronic system. The electronic system 6 comprises a microprocessor 7 adapted to receive the electrical signal from the microneedles 5. In addition, the microprocessor 7 may be adapted to carry out signal processing of the electrical signal from the microneedles.

[0053] The electronic system 6 may also include a memory 8 for storing the electrical signals received from the micro-needles and / or processed by the microprocessor 7.

[0054] The electronic system 6 may comprise a power supply 9 adapted to supply electrical energy to the microprocessor 7 and any components of the electronic system that require it. If necessary, the power supply 9 may be exchanged and / or recharged. However, since the detection system may only be used for a few days, this is not mandatory.

[0055] The electronic system 6 also comprises a communication module 10. For example, a wireless internet communication module is provided by radio waves, in particular short-range. For example, it is provided that the communication module 10 implements a Bluetooth communication protocol available on the priority date of the present patent application.

[0056] As seen in [Fig.l], the computerized system also includes a portable communicating device 11. The portable communicating device 11 is for example a so-called “smart” mobile phone (generally referred to by the Anglicism “smartphone”), a tablet, a laptop, a connected watch, or other. As shown in [Fig. 3], the portable communicating device 11 comprises a microprocessor 12, a memory 13, and a power supply 14 adapted to supply electrical energy to the microprocessor 12 and any other components of the portable communicating device 11 that need it. If necessary, the power supply 14 can be exchanged and / or recharged via a charger (not shown).

[0057] The portable communicating device 11 also comprises a first communication module 15. The first communication module 15 is complementary to the communication module 10 of the detection system. In the example presented, a wireless internet communication module by radio waves is provided. For example, it is provided that the communication module 15 implements the Bluetooth communication protocol mentioned above.

[0058] The portable communicating device 11 also comprises a second communication module 16. For example, a wireless internet communication module by radio waves is provided. This second communication module 16 has a greater range than that of the first communication module 15. For example, it is provided that the second communication module 16 implements a wifi communication protocol available on the priority date of the present patent application. Alternatively or additionally, the second communication module 16 can implement wired internet communication.

[0059] The portable communicating device 11 also comprises a human-machine interface 17 adapted to allow the user to interact with the processor 12, that is to say to enter information which is transmitted to the processor 12, and for the processor 12 to be able to communicate information to the user. For the present invention, the human-machine interface 17 of the portable communicating device 11 comprises at least one display 18, of the screen type. The screen may comprise a touch screen. Other means of restoring and entering information are possible, such as for example a loudspeaker, a microphone and a voice recognition software module or other.

[0060] The memory 13 is adapted to store information received from the detection system 1, from the server 19, and / or processed by the microprocessor 12. The memory 13 can thus store temporal information relating to the glucose concentration over time g(t), and temporal information relating to the NEFA concentration over time n(t). The temporal information relating to the analyte concentration over time is generally designated by the function a(t). Thus, the function a(t) can be a composite function of g(t), n(t), or even other time functions relating to different analytes. The time value is given by a digital clock of the portable communicating device 11. The measurement time step of the different analytes can be different.

[0061] The memory 13 may also store a computer program to be executed by the microprocessor 12. The computer program in question may also be called an “application”. The computer program in question may allow for managing different data processing, communication and user interaction functionalities. As will be described below, the computer program may comprise several functional computerized modules. The division, made in the description below, of the computer program into modules, is a functional division which does not presage the actual writing of the computer program code.

[0062] The application can be programmed to enter into memory various information relating to the user, for example by means of a declarative system implemented via the human-machine interface 17. The user can thus enter various information into memory, for example by means of questionnaires.

[0063] The information provided by the user may, for example, include: • Its weight, • His estimated stress level, • One or more emotions, • His meals (meal time, estimated amount of energy ingested, composition of plate), • His sleep phases (estimated time of falling asleep, time of waking up), • His physical exercise phases (estimated start time, end time, duration, intensity, type),

[0064] This information can thus be entered by the user repeatedly.

[0065] Referring again to [Fig. 1], the computer system also comprises a computer server 19 (also shown in more detail in [Fig. 4]). The computer server 19 comprises a processor 20, a memory 21, and a power supply 22 adapted to supply electrical power to the processor 20 and any other components of the computer server that require it.

[0066] The computer server 19 also comprises a communication module 23. The communication module 23 is complementary to the second communication module 16 of the portable communicating device 11.

[0067] The computer server 19 comprises a computerized characterization module 26. The computerized characterization module 26 is adapted to determine a composite index of the user, representative of the storage of fat by the user, such as for example an absolute value of storage of triglycerides in the adipose tissue, a variation or an acceleration of this value. The composite index is determined by applying a rule to the information relating to the user stored in the memory 21. The information relating to the user in question is all or part of the information listed above. The rule is also stored in the memory 21. The rule can be common to all users, or, alternatively, a personalized rule can be provided for each user, or for a predefined group of users, in a manner which will be described later.

[0068] The composite index ic(t) of the user can vary over time. Thus, the computerized characterization module 26 is applied repeatedly over time, using the information relating to the user enriched over time stored in the memory 21.

[0069] According to an exemplary embodiment, the rule comprises a physiological model determining a composite index representative of fat storage, from input data. There are numerous public physiological models that can be implemented by the characterization module 26. The physiological model will for example model the competing mechanisms of lipolysis, i.e., the destocking of triglycerides from adipose tissue, and synthesis, i.e., the storage of triglycerides within adipocytes. Below, the physiological model presented in “Mathematical modeling of hepatic lipid metabolism,” Pratt et al., Mathematical Biosciences 262 (2015), 167-181, is presented as an example. According to an embodiment of the invention implementing such a physiological model, the measurements of glucose, glycerol, and NEFA in the interstitial fluid are used as input data.

[0070] The physiological model can, for example, be a compartmentalized model, which uses variables representative of each compartment, as well as models of transport mechanisms between the compartments. For example, four compartments can be used, namely the liver, adipose tissue, skeletal muscle and blood plasma.

[0071] The model can for example use the following variables:

[0072] For the liver: concentrations of glucose (GL), glycogen (YL), G-6-P (PL), pyruvate (RL), free fatty acids (AL), stored / reserve volume of triglycerides (TL, SL);

[0073] For muscle: concentrations of glucose (GM), glycogen (YM), G-6-P (PM), pyruvate (RM), free fatty acids (AM), triglycerides (TM), notional AMP (P);

[0074] For plasma: concentrations of glucose (GB), non-esterified fatty acids (ANB), endogenous triglycerides (TLB), exogenous triglycerides (TCB), insulin (I);

[0075] For adipose tissue: the concentrations of glucose (FA), free fatty acids (Aa), triglycerides (Ta), glycerol (LA).

[0076] One can for example use

[0077] This model notably implements the following equations representative of transports:

[0078] (dl / dt) = kn+k22 erf [(GB-v) / cc] - kd I (1)

[0079] aL (dGL / dt) = SG(t) - kglGL + kgl2GB - vlgGl / (kLG+GL) - [l / (l+crepPL)] vLHGL / (kLH+GL )+k61PL(2)

[0080] aL (dYL / dt) = ½ kylIPL ​​(l+tanh[(lmax-YL) / c0] - |3L [YL / (YL+y0)] / (l+kdlI) (3)

[0081] aL (dPL / dt) = -1 / 2 kylIPL ​​(l+tanh[(lmax-YL) / c0] + |3L [YL / (YL+y0)] / (l+kdlI) - kpIPL + k^L a + PôRiAl+kpôI) + vlgGl / (kL +GL) vGl [l / (l+crepPL)] / (kLH+GL) (4)

[0082] aL (dRL / dt) = kpp RM + kpIPL - [36RL / ( l+kp6I) - calIRL + pB (5)

[0083] ci| (dAL / dt) = 3kciTGB + kbiANB + 3 kr TLB + kaiIRL — 3 v6A| / (k6+A|) + 3vioTL / (kio+TL) - 3 VsA^ks+AJ - kvA^l+ksI) (6)

[0084] aL (dSL / dt) = v6AL / (k6+AL) - k9aSL (7)

[0085] aL (dT^dt) = v8AL / (k8+AL) - F(I) v9TL / (k9+TL) - v10TL / (k10+TL) (8)

[0086] aM (dGM / dt) = (l+kgiI)(kgmGB — kgm2GM) - vmhGm [l / (l+crepPM)] / (kMH+GM) (9)

[0087] aM (dYM / dt) = 1 / 2 kymIPM (l+tanh[(mmax-YM) / c0] - |3M [YM / (YM+y0)] / (l+kdyI) (10)

[0088] aM (dPM / dt) = - / 2 kymIPM (l+tanh[(mmax-YM) / c0] + |3M [YM / (YM+y0)] / (l+kdyI) - k6pIPM + vmhGm / (kMH+GM) [l / (l+crepPM)] (11)

[0089] aM (dRM / dt) = k6p I PM - kppRM - p3RM IP (12)

[0090] Qm (dAM / dt) = - 3msIAM + 3mL. + 3kcmTGB+kbmANB + 3 ktTLB - p4AMP (13)

[0091] aM (dTM / dt) = msIAM - me (14)

[0092] dP / dt = pamp - p4AMP - p3RMIP (15)

[0093] aA (dTA / dt) = kaaIAAGA - |3f / (l+kftP) (16)

[0094] aA (dAA / dt) = - 3 kaaIAAGA + 3ka(l+kaiI)(l-klp)TCB + 3kbaTLB + knaANB (17)

[0095] aA (dLA / dt) = [3f / (l+kftP) - k^ ( 18)

[0096] aA (dGA / dt) = dBA (l+kgaI)(GB-GA) - kaaIAAGA (19)

[0097] dTCB / dt = SF(t) - kcmTCB-kclTCB - ka (l+kaiI)TCB (20)

[0098] dANB / dt = -kbmANB — kwANB — knaANB + 3[>f / (l+kflP) + 3kAkip(l+kaiI)TCB (21)

[0099] dTLB / dt = F(I) v9TL / (k9+TL) + k9aSL - krTLB - ktTLB - kbaTLB (22)

[0100] dGB / dt = kglGL - kgl2GB - (l+kgiI)(kgmGB - kgm2GM) - dBA(l+kgaI)(GB-GA) - p! (23)

[0101] The model parameters aA, aB, aL, aM, [36, [3f, [3i, [3m, pamp, pb, pe, ps, pb p3, p4, c0, ce, dBA, kio, kn, ki2, ki3, kM, k22, k3, kb, kbi, kbp, k7, k8, k9, k9a, ka, kaa, kai, kab kba, kbb kbm, kci, k cm, kd, kdi, kdy, kft can be determined from known values ​​in the scientific literature, and / or from a prior clinical trial. If necessary, several possible pre-recorded values ​​may be available for one or more of these parameters, and the value corresponding to the user may be determined based on user-provided responses to a questionnaire.

[0102] Alternatively or additionally, the value of certain variables can be determined during a calibration phase, where the user is subjected to predefined conditions (meals, sleep, exercise, etc.), where the analytes of the interstitial fluid are detected as described above, and values ​​of the parameters are determined which make it possible to determine a model which best corresponds to the input data and the detections.

[0103] According to an alternative embodiment, a physiological model is used which also takes into account as input data the measurements of the concentration of ketone bodies in the interstitial fluid.

[0104] As a variant of the above model, which uses the measured concentrations of glucose, NEFAs and glycerol, or even, in the variant presented, of ketone bodies, in the interstitial fluid, a physiological model is used which takes into account as input data one of the following combinations of concentrations measured in the interstitial fluid:

[0105] - glucose and NEFAs,

[0106] - glucose and glycerol,

[0107] - glucose and ketone bodies,

[0108] - NEFAs and glycerol,

[0109] - NEFAs and ketone bodies,

[0110] - glycerol and ketone bodies,

[0111] - glucose, NEFAs and ketone bodies,

[0112] - glucose, glycerol and ketone bodies,

[0113] - NEFAs, glycerol and ketone bodies.

[0114] The computerized characterization module 26 is thus adapted to repeatedly determine over time the composite index ic(t).

[0115] Where appropriate, several models are available at the server 19, and a particular model is applied for a given user. The selection of the model suited to the user can be made once and for all when the system is initialized. The user can select the model suited to him by answering a few simple questions via the human-machine interface. For example, there is a model for adults and a model for children, and the model is selected based on the user's age. Other distributions or divisions are possible.

[0116] According to an exemplary embodiment, the communication module 23 is adapted to transmit to the portable communicating device 11 repeatedly over time a value of the composite index ic(t). This communication is done at a predetermined frequency, or at each calculation of a new value of the composite index ic(t), or even when using the on-board computer program. on the portable communicating device 11, for example.

[0117] As shown in [Fig. 5], the portable communicating device 11 is adapted to display on its display 18 a curve 28 representing the composite index ic(t) (on the ordinate) over time (on the abscissa). The displayed time scale may be predetermined, for example the last 24 hours. The display 18 may also present a reference value for the curve ic(t), for example in the form of a horizontal straight line 29 such as for example a predetermined reference value, an initial value, an average value over the displayed time interval, or other.

[0118] The display 18 can also present the events, for example meals and phases of sleep and physical exercise, declared or detected, or the day-night cycle, presented by informative contents 30 arranged at the corresponding instant.

[0119] The part of the curve ic(t) 28 appearing below the reference line 29 corresponds to a period of fat destocking, while the part of the curve ic(t) 28 appearing above the reference line 29 corresponds to a period of fat storage.

[0120] The display may also include a menu 39 allowing access to various functions such as setting user data or preferences, access to questionnaires, declaration forms, and history. Menu 39 may, for example, include a button allowing a change in the time interval (limit and / or duration) of the displayed curve to be controlled.

[0121] The computerized system also comprises a computerized recommendation module 31. The computerized recommendation module 31 is adapted to process the input data, as presented above, and to determine, from this data, a personalized recommendation for the user.

[0122] The computerized recommendation module 31 may take as input the data listed above. It may also take as input the values ​​ic(t), which are obtained from these input data, or other interpreted data established from the input data, as presented below in the discussion relating to the computerized interpretation module 33.

[0123] The computerized recommendation module 31 has access to a recommendation database 32.

[0124] The recommendation database 32 may include several recommendations. For example, the recommendations are classified by categories. For example, the recommendations may each be classified into one or more categories among nutrition, sleep, health, organization, physical activity, psychology. Other categories may be considered. The following list of potential recommendations is not exhaustive, and may be updated regularly in the recommendation database 32. The recommendations may concern the nutrition, or more generally the user's lifestyle.

[0125] Among the recommendations relating to nutrition, we can for example list: • Lower the amount of a particular type of food, • Increase the amount of a particular type of food, • Replace one type of food with another, • Establish a regular practice limiting fat loss, • Change your diet with regard to the carbohydrate / fat ratio, the amount of carbohydrates and / or the amount of fats, • Change the type of carbohydrate or fat, • Change meal rhythm (number, time, spacing, volume, etc.), • Adjust diet according to physical activities, • Consume micronutrients, • Adjust your hydration;

[0126] Among the recommendations relating to sleep, we can for example list: • Modify a pre-night activity;

[0127] Among the health recommendations, we can for example list: • Take rest; • Try breathing exercises;

[0128] Among the recommendations relating to the organization, we can for example list: • Change power source; • Use recommended recipes; • Use a recommended menu; • Use a recommended shopping list, • Schedule time slots for shopping or having it delivered, cooking, etc.

[0129] Among the recommendations relating to physical activity, we can for example list: • Increase activity between physical exercise sessions, • Take up a new physical activity, • Adapt the duration and intensity of a commonly practiced physical exercise session, • Adapt the weekly number of physical exercise sessions, • Adapt the schedule of commonly practiced physical exercise sessions,

[0130] Among the recommendations relating to psychology, we can for example list: • Try to determine whether meals are emotional or hunger-driven meals, • Consult a nutritionist psychologist,

[0131] The list of recommendations arranged in the database 32 may be a general list. Alternatively, it may be a list personalized for the user. A computerized selection module 34 is adapted to select, from a general list of recommendations, a personalized list of recommendations adapted to the user, based on the identity of the user, his tastes and / or his constraints, entered via the human-machine interface 17, as will be described later. For example, if a recommendation from the general list is to consume a certain food, and the user declares in his tastes that he does not like this food, this recommendation is not kept in the personalized list of recommendations.

[0132] The computerized recommendation module 31 is adapted to determine at least one recommendation (for example a single recommendation, at most five recommendations) adapted to the user, in the list of recommendations available in the recommendation database 32.

[0133] For this, the computerized recommendation module 31 is adapted to determine the relevance of a recommendation for the user.

[0134] For example, the computerized recommendation module 31 comprises a set of rules determining a level of relevance of each recommendation based on user data.

[0135] The rules may use one or more user-related data.

[0136] Each rule can be identified by an identifier ri, where the index i varies from 1 to n, n representing the number of recommendations, so that the relevance level p can be defined from a relevance function pi of the rule ri for the user uk , where the index k varies from 1 to m, n representing the number of users, in the following way: p = pi (ri, uk). The relevance level p is for example an adimensional parameter reported on a scale from 0 to 1.

[0137] According to one example, a function pi for a recommendation i can be defined as follows: • p = 0 if P > Pd; and • p = 1 if P < Pd,

[0138] where P represents the user's weight, and Pd represents a predetermined weight.

[0139] Such a rule allows to exclude from recommendations an inappropriate rule if the user's weight exceeds a predetermined weight. Such a function can be applied for recommendations relating to certain physical exercises.

[0140] The computerized recommendation module 31 is thus implemented to determine a predetermined number of rules adapted to the user, from the data relating to the user.

[0141] The data relating to the user can be used in raw form by the computerized recommendation module 31. This is the case, for example, of the user's weight, as described above.

[0142] Alternatively or additionally, the user-related data may be interpreted by a computerized interpretation module 33 from the user-related data.

[0143] The computerized interpretation module 33 provides interpreted data relating to the user, determined from the raw data. An example of implementation is for example described below in relation to [Fig.6].

[0144] [Fig.6] schematically represents the curve ic(t) determined as described above- above for the user for a predetermined time interval. For example, the predetermined time interval can be written [tO-Tp; tO], where tO is a predetermined instant, for example the present instant, or the instant for which the last measurement ic(t) is available, or midnight the day before, and Tp is a predetermined duration, for example 3 hours or 1 day.

[0145] In itself, the curve ic(t) over the time interval [tO-Tp; tO] is already interpreted data.

[0146] Other interpreted data can be determined from the curve ic(t), such as for example the maximum or minimum amplitude, the time of the maximum or minimum amplitude with respect to the time interval [tO-Tp; tO] and / or with respect to the present time; a statistical quantity associated with the field ic(t), such as for example the mean value, a weighted mean value, a standard deviation; a maximum speed or acceleration of the curve; an inflection point; an area above or below a reference value, or a difference between these areas, ...

[0147] For example, data interpreted from the field ic(t) over the time interval [tO-Tp; tO] corresponds to a difference E between the area above a reference value and the area below this reference value over the time interval considered, representative of the net storage of fat over the time interval considered.

[0148] Other examples of determining interpreted data Ei from user data are possible. For example, the measured concentration of ketone bodies in the interstitial fluid can be used as input to determine interpreted data relating to a meal or physical exercise.

[0149] The relevance functions pi are for example determined during an upstream learning phase, which will be described later.

[0150] Thus, the computerized recommendation module 31 determines one or more recommendations relevant to the user on the basis of the results of the relevance functions, and presents these recommendations to the user via the human-machine interface 17. The chosen recommendation is then selected by the user.

[0151] Alternatively, the computerized recommendation module 31 provides a list of recommendations to the user before determining their relevance and, if the user previously chooses a recommendation from this list, the computerized recommendation module 31 informs the user of the relevance score associated with this recommendation for him, if applicable, with regard to the recommendations automatically determined by the computerized recommendation module 31 and their relevance score. The user is then free to select the recommendation chosen from the automatically determined recommendations and the one previously chosen by the user.

[0152] Ultimately, the recommendation chosen by the user is saved in memory.

[0153] The computerized recommendation module 31 is for example provided in the portable communicating device 11.

[0154] Typically, the computerized recommendation module 31 is based on user data extending over a period of at least several days, typically at least one week. Such a duration seems useful to be able to determine interpreted user data relevant for the implementation of a recommendation.

[0155] The computerized interpretation module 33 is thus adapted to repeatedly determine a value for an interpreted parameter. The temporal evolution of this interpreted parameter can be taken into account by the computerized recommendation module. Thus, in particular, to determine the value of an interpreted parameter at time t, the computerized interpretation module 33 can use the user's data history, and compare the recent data to the user's interpreted data history, in order to identify past data resembling the current data, and having already been interpreted. The notion of "resemblance" can be determined by calculating the distance between two data fields.For example, if fat storage associated with a meal of a particular type (e.g., "breakfast") has been identified in the past, the computerized interpretation module seeking to interpret a present signal compares this signal with a past signal associated with fat storage associated with this type of meal and, in the event of a positive comparison, the present signal is interpreted as fat storage associated with this type of meal. The comparison is positive if the distance between the two signals is less than a certain predetermined threshold.

[0156] The computerized interpretation module 33 may use only the user's data. Alternatively or additionally, the computerized interpretation module 33 may use data relating to other users, stored in the database 25. Indeed, certain events may be similar for different users, and it is possible to drastically increase the volume of comparison data available to implement an interpretation.

[0157] In summary, the computerized interpretation module 33 includes a history of interpreted data comprising raw data, and interpreted data associated in the past with this raw data, and determines the interpreted data by comparing the instantaneous raw data of the user with the raw data of the history. When the comparison of the instantaneous raw data and the raw data of the history is less than a predetermined threshold, the instantaneous raw data is associated with the interpreted data associated in the history with this raw data of the history.

[0158] The computerized interpretation module 33 can also interact with the user to label a piece of data. If the computerized interpretation module 33 considers that it has a low level of confidence in the interpretation of a piece of data, it can control the interaction with the user, by providing the raw data, the interpretation proposal, and a possibility of confirmation / rejection to the user via the human-machine interface 17. For example, a pop-up window can be provided on the display 18, equipped with controllable “Yes” / ”No” buttons. The low level of confidence can be determined when the distance between the signal and the interpreted data to which it is compared is within a predetermined interval (outside this interval, either the distance is small, and the signal is interpreted without submission to the user, or the distance is large, and the signal is considered to be unrelated to the interpreted signal).

[0159] The parameter is determined repeatedly, for example once a day. It may be associated with a part of the day, for example, namely morning, midday, afternoon, evening or night. Typically, the parameter may be associated with meals, or a type of meal. Thus, for example, a parameter may be determined for a predetermined period of time following breakfast each day, or following each meal.

[0160] The system further comprises a computerized interrogation module 37 adapted to trigger an interface with the user in order to obtain information from the user. The triggering may be based on time, and / or on user data, and / or based on an interaction of the user with the communicating portable device 11. For example, the triggering takes place every morning at 8 a.m., or is based on an analysis of user data. In this case, the computerized module 37 is used to assist the computerized interpretation module 33. Thus, if the computerized interpretation module 33 is only able to interpret the measured data with a confidence level below a certain threshold, it can select one or more questions depending on the confidence level in the interpretation of the data, which will be displayed by the computerized interrogation module 37, so that the answer entered by the user is used to interpret the data.

[0161] The computerized interrogation module 37 controls the human-machine interface 17 in order to obtain information. For example, it is possible to provide for the display of windows on the display 18, allowing the user to enter information.

[0162] The information in question may relate, for example, to diet, physical exercise, emotions / stress and / or sleep.

[0163] For example, with respect to food, the following questions and the information entered in response to the questions may be used: • What did you just eat / drink (type of food / drink, quantity)? •

[0164] This question is for example triggered at a typical pre-programmed time following a meal, or in the event of detection of an instantaneous increase in fat storage in the user data.

[0165] For example, with respect to physical exercise, the following questions and the information entered in response to the questions may be used: • Have you been active today? • Are you tired ? •

[0166] These questions are, for example, triggered at the end of the day.

[0167] Regarding physical exercise, the following questions and the information entered in response to the questions can also be used: • What kind of physical exercise did you just do? • For how long? • At what intensity? • How do you feel? •

[0168] These questions are for example triggered in the event of detection of physical exercise.

[0169] For example, regarding emotions and stress, one can use the following questions, and the information entered in response to the questions: • How was your day ? • What is your level of satisfaction? • Have you been stressed today? • Are you injured or sick? • Are you feeling any particular emotion or stress right now?

[0170] These questions are, for example, triggered at the end of the day or when the user opens the application.

[0171] For example, with respect to sleep, the following questions and the information entered in response to the questions may be used: • Did you sleep well? • Has your sleep time been delayed compared to usual? • Did you sleep more or less than usual last night? •

[0172] These questions are for example triggered in the morning, or in case a data processing detects an unusual data during the previous night. The unusual nature of a data can be estimated by estimating a distance between a data and past data for the same period of time corresponding to previous days.

[0173] As shown in [Fig.4], the computer server 19 has access to a user database 25. The user database 25 includes data relating to different users of detection systems 1. Thus, the user database 25 stores, for different users u, temporal data of analyte concentration au(t), as well as information relating to the users u, such as listed above (weight, meals, sleep, etc.). Where appropriate, the user database 25 is created during a pre-clinical trial, or before the computerized system is put into service for the users.

[0174] The computerized system comprises a computerized evaluation module 35 adapted to determine a parameter for evaluating the impact I of the recommendation on the user.

[0175] The impact assessment parameter can be determined several times over time, so that the determined values ​​are entered in the I(t) field.

[0176] The impact evaluation parameter I(t) corresponds, for a control parameter, to a difference between on the one hand the value of this parameter at time t, and on the other hand a prediction of the value of this control parameter at time t, this prediction being made at a time t-sooner than t, in the hypothesis where the recommendation was not followed.

[0177] The predictive computerized module 36 is thus implemented for the instant t-soon, for example at the time of the user's selection of the recommendation, and determines, for the instant t, a prediction of the value of the control parameter in the lack of follow-up of the recommendation.

[0178] The predictive computerized module 36 is based, for example, on the history of data relating to the user already obtained at time t-t to determine the forecast.

[0179] For example, if the control parameter is the user's weight, and the user's history shows that, over the previous week, the weight remained stable, then the predictive computer module 36 predicts that the weight would remain stable in the absence of following the recommendation.

[0180] The computerized evaluation module 35 then compares the value of the impact evaluation parameter with the goal sought by the user. For example, if the goal sought by the user is weight loss, and the value of the impact evaluation parameter corresponds to weight loss, then the recommendation is determined to have had a positive impact. In other words, the recommendation is deemed effective.

[0181] If the recommendation is deemed ineffective, the recommendation may be deleted from the personalized list of recommendations by the computerized selection module 34.

[0182] If the recommendation is deemed effective for the desired purpose, this information can be transmitted to the database 25. Thus, the database 25 stores the examples of recommendations deemed effective. This information is associated with the user's information, because the same recommendation can be effective for one user and not for another, it is therefore important to link the recommendation and its impact to the user data.

[0183] When repeating the step of determining a recommendation, past recommendations for this user may be taken into account. Thus, a past recommendation may not be selected, even if it had a positive impact, to vary the recommendations for the same user.

[0184] In the foregoing description, it has been described that certain processors are adapted to implement certain operations. However, since the processors of the computer system are also adapted to communicate with each other, as described, it is possible, in other embodiments, that the distribution of tasks between the different processors is different.

[0185] The system just described can operate as follows.

[0186] Firstly, we have the database 25 with user data.

[0187] The system is installed for a given user. The user installs the “wearable” detection system 1, and installs the ad hoc computer programs on his “smartphone” 11. The “Bluetooth” connection between the “wearable” and the “smartphone” is established, and from time to time, the “smartphone” receives data from the detection system 1.

[0188] In the initialization phase, the user also provides information to the portable communicating device 11 via the human-machine interface 17. For example, the portable communicating device 11 executes a computer program displaying a questionnaire on the display 17, and allowing the user to enter information.

[0189] The memory 13 of the portable communicating device 11 thus stores information relating to the user. This information may relate to the identity of the user, his tastes, his constraints and his objectives. This is information which is likely to remain constant throughout the use of the system, apart perhaps from the objectives, which may vary, but over a long period.

[0190] Among the information relating to the user's identity, we can for example list: • Age • Gender, • Weight, • The size, • The user's current diet, • The characteristics of the user's regular physical exercise (type of activity, weekly hourly volume), • The activity relating to the user's occupation (type of job and associated constraints (physical effort, stress, etc.)), • The quality of sleep, • A medication history, • A medical history, • Stress, • A usual food, • Alcohol consumption, • The usual use of home cooking, restaurant cooking, industrial or artisanal takeaway dishes, etc. •

[0191] Among the information relating to the user's tastes, we can for example list: • His favorite and / or disliked dishes and / or drinks, • Favorite and / or disliked physical activities, • Preferred and / or disliked sleep management activities, • Favorite and / or disliked stress management activities, • Favorite and / or disliked stress management activities,

[0192] Among the information relating to user constraints, we can for example list: • Time constraints, • Organizational constraints, • Logistical constraints, • Health constraints, • Budgetary constraints, • Constraints on access to certain foods and / or physical activities, • The absence of constraints, • The freedom to change certain aspects relating to diet, physical exercise, sleep, emotional eating, etc., •

[0193] Among the information relating to the user's objectives, we can for example list: • The desire to lose fat mass, • The desire to stabilize one’s body, • The desire to gain weight without increasing the proportion of fat mass, •

[0194] The computerized selection module 34 selects, from the general list of recommendations, a personalized list of recommendations adapted to the user, based on the identity of the user, his tastes and / or his constraints. If necessary, this step can be repeated over time, in particular if the identity, tastes and / or constraints declared by the user change.

[0195] The system then implements a user observation period.

[0196] During this observation period, the measured data are stored. In addition, the user can use the human-machine interface 17 of his portable communicating device to declare relevant information, typically meals, periods of sleep, physical exercise and / or stress.

[0197] The observation period typically lasts several days, for example a week.

[0198] During the observation period, the computerized interpretation module 33 is used, so as to generate the interpreted data for the user, as presented above. Alternatively, the interpreted data are determined by further taking into account the information entered by the user, relating to the identity of the user, his tastes, and / or his constraints.

[0199] The generation of data interpreted by the computerized interpretation module continues at the end of the observation period.

[0200] At the end of the observation period, the system has sufficient data to implement the computerized recommendations module 31.

[0201] The computerized recommendation module 31 is used to generate a personalized recommendation to the user, as described above.

[0202] For the generation of personalized recommendations, the computerized recommendation module 31 takes into account the data relating to the user, in particular the physiological data, but also the declared data relating to the identity of the user, his tastes and / or his constraints.

[0203] Via the human-machine interface 17, the user validates the recommendation, or one of the recommendations made to him. This validation is recorded in memory 13.

[0204] The process described above, with an observation period followed by a recommendation, is then repeated.

[0205] The computerized evaluation module 35 determines an evaluation parameter I of the impact of the recommendation on the user.

[0206] If the recommendation is deemed ineffective, the recommendation may be deleted from the personalized list of recommendations by the computerized selection module 34.

[0207] If the recommendation is deemed effective for the desired purpose, this information can be transmitted to the database 25. Thus, the database 25 stores the examples of recommendations deemed effective. This information is associated with the user's information, because the same recommendation can be effective for one user and not for another, it is therefore important to link the recommendation and its impact to the user data.

[0208] When repeating the step of determining a recommendation, past recommendations for this user may be taken into account. Thus, a past recommendation may not be selected, even if it had a positive impact, to vary the recommendations for the same user.

[0209] Data relating to the user, whether raw data, interpreted data, data entered by the user, implemented recommendations, etc., are transmitted regularly to the database 25.

[0210] The server 19 comprises a computerized module for determining relevance functions 38. The computerized module for determining relevance functions 38 determines the relevance functions of each recommendation. For this, the computerized module for determining relevance functions 38 includes as input the history of each user, in particular, for each recommendation, the user data before the recommendation and the parameter for evaluating the impact of the recommendation. These data are used, for example by linear regression, to determine a relevance function of the recommendation based on the user input data. For example, if the data of many users show that, for overweight users, the recommendation to remove bread from the diet is beneficial, while this recommendation has little impact for users of normal weight, we can establish a relevance function of this recommendation as a function of the user's weight, which is an increasing function. This determination step can be implemented once and for all, during an upstream learning phase, when the system is put into service for all the first users, on the basis of data obtained during pre-clinical tests. If necessary, it can be regularly updated as user data is enriched.

[0211] According to one embodiment, the detection system 1 comprises several distinct detection devices of the type that are worn on the skin. For example, distinct detection devices can detect different metabolites.

[0212] According to one embodiment, such a nutritional coach does not make any medical diagnosis, in the sense that it is not a matter of determining a particular illness from which a user is suffering from based on the measured data. Such a nutritional coach does not implement any therapy, in the medical sense, in that the users of the nutritional coach do not suffer from an illness that the nutritional coach aims to cure, that the nutritional coach simply issues a recommendation that the user is free to follow or not, and that, if the user follows the recommendation, he or she is not “cured”. List of reference signs

[0213] 1: Detection system

[0214] 2: Skin

[0215] 3: lower surface

[0216] 4: layer of glue

[0217] 5: micro-needles

[0218] 6: electronic system

[0219] 7: microprocessor

[0220] 8: memory

[0221] 9: power supply

[0222] 10: communication module

[0223] 11: portable communicating device

[0224] 12: processor

[0225] 13: memory

[0226] 14: power supply

[0227] 15: first communication module

[0228] 16: second communication module

[0229] 17: man-machine interface

[0230] 18 display

[0231] 19 server

[0232] 20 processor

[0233] 21 memory

[0234] 22 power supply

[0235] 23 communication module

[0236] 25 Database

[0237] 26 computerized characterization module

[0238] 28 curve

[0239] 29 straight line

[0240] 30 information content

[0241] 31 computerized recommendation module

[0242] 32 Recommendation database

[0243] 33 computerized interpretation module

[0244] 34 computerized selection module

[0245] 35 computerized evaluation module

[0246] 36 computerized predictive module

[0247] 37 computerized interrogation module

[0248] 38 module computerized determination of relevance functions

[0249] 39 menu

Claims

Claims

1. A computerized method for determining a personalized nutritional recommendation of a user in which: - at least one portable sensor worn by the user provides temporal data of the user's metabolite concentration obtained by in vivo measurements carried out regularly over time in a non-surgical manner in the interstitial fluid of the user, comprising at least temporal data of the concentration of at least two of the metabolites from the following list: {glucose, non-esterified fatty acids, glycerol, ketone bodies} in the interstitial fluid of the user, said method being characterized in that: - a computerized characterization module (26) determines a composite index of the user from said temporal data of the user's metabolite concentration,and in that: - a computerized recommendation module selects at least one recommendation adapted to the user from a list of recommendations comprising several recommendations using said composite index, based at least on said temporal metabolite concentration data, using at least said temporal metabolite concentration data, by applying rules associating the recommendations from the list of recommendations with composite indices.,

2. A computerized method according to claim 1, wherein the computerized recommendation module determines a relevance score of at least a plurality of recommendations in the recommendation list, and selects said recommendation suitable for the user based on said relevance scores.

3. A computerized method according to claim 2, wherein the computerized recommendation module determines the relevance score further based on a user-defined goal to be achieved, wherein said rules associate the recommendations in the recommendation list with temporal metabolite concentration data and goals to be achieved.

4. A computerized method according to claim 2 or 3, wherein the computerized recommendation module determines the relevance score of a recommendation by determining a prediction of a value of a control parameter by simulating an absence of recommendation follow-up, and a prediction of said value of the control parameter by simulating an implementation of said recommendation from at least said temporal metabolite concentration data.

5. Computerized method according to one of claims 1 to 4, in which the computerized recommendation module selects said at least one suitable recommendation from further a history of recommendations for the user.

6. Computerized method according to one of claims 1 to 5, in which the computerized recommendation module selects the recommendation adapted to the user based in addition on at least data relating to the identity, tastes, and / or constraints of the user entered by the user via a human-machine interface.

7. A computerized method according to any one of claims 1 to 6, wherein the computerized recommendation module selects the recommendation suitable for the user based further on diet, health, exercise and / or sleep data, as well as times associated with these data, entered by the user via a human-machine interface.

8. Computerized method according to one of claims 1 to 7, in which said at least one, in particular at least two, suitable recommendations are transmitted to the user, and the user selects a recommendation from among said at least one, in particular at least two, suitable recommendations.

9. Computerized method according to one of claims 1 to 8, further comprising a phase of observation of the user using said portable sensor prior to the selection of said at least one suitable recommendation.

10. Set of computer programs comprising instructions adapted for, from temporal data of concentration of metabolites of the user obtained by in vivo measurements, carried out by at least one portable sensor worn by the user, regularly over time in a non-surgical manner in the interstitial fluid of the user, comprising at least temporal data of concentration of at least two of the metabolites from the following list: {glucose, non-esterified fatty acids, glycerol, ketone bodies] in the interstitial fluid of the user, the implementation of the method including at least the following steps: - a computerized characterization module (26) determines a composite index of the user from said temporal data of concentration of metabolites of the user, and: - a computerized recommendation module selects at least one recommendation adapted to the user from a list of recommendations comprising several recommendations using said composite index, based at least on said temporal metabolite concentration data, using at least said temporal metabolite concentration data, by applying rules associating the recommendations from the list of recommendations with composite indices, when the set of computer programs is executed by one or more computers.

11. A computerized system for determining a personalized nutritional recommendation for a user, comprising: . at least one portable sensor worn by the user and adapted to carry out in vivo measurements regularly over time in the interstitial fluid of the user to produce temporal data of concentration of metabolites of the user, comprising at least temporal data of concentration of at least two of the metabolites from the following list: {glucose, non-esterified fatty acids, glycerol, ketone bodies} in the interstitial fluid of the user, said computerized system being characterized in that it further comprises: . a computerized characterization module (26) adapted to determine a composite index of the user from said temporal data of concentration of metabolites of the user, and . a computerized recommendation module adapted to select at least one recommendation adapted to the user from a list of recommendations comprising several recommendations using said composite index, based at least on said temporal metabolite concentration data, by applying rules associating the recommendations from the list of recommendations with composite indices.

Images