COMPUTER-IMPLEMENTED METHOD

Abstract

A computer-implemented method for deriving a physiological rank indicative of a physiological status of a user, the computer-implemented method comprising acquiring, from a sensor on a wearable device worn by a user, data including bodily parameter data related to the user, and applying a model to the bodily parameter data to obtain physiological information related to the user, and deriving, from the physiological information, a physiological rank indicative of a physiological status of the user wearing the device, wherein the physiological rank is a given value on a physiological rank scale.

Claims

1. A computer-implemented method for deriving a physiological rank indicative of a physiological status of a user, the computer-implemented method comprising: acquiring, from a sensor on a wearable device worn by a user, data including bodily parameter data related to the user; and, applying a model to the bodily parameter data to obtain physiological information related to the user; and, deriving, from the physiological information, a physiological rank indicative of a physiological status of the user wearing the device, wherein the physiological rank is a given value on a physiological rank scale.

2. The computer-implemented method of claim 1 further comprising: acquiring other sensor information in addition to the physiological information; and/or, acquiring user input information.

3. The computer-implemented method of claim 2 further comprising: storing a notification data table, the notification data table associating notifications with stored physiological ranks; and, comparing the derived physiological rank with the stored physiological ranks and, based on this comparison, selecting a notification to output to the user; and, outputting the selected notification to the user.

4. The computer-implemented method of claim 3, wherein the notification data table further associates notification with stored other sensor information and/or stored user input information, and wherein the method further comprises: comparing acquired other sensor information and/or acquired user input information with the stored other sensor information and/or the stored user input information, wherein the selecting the notification to output to the user is further based upon this comparison.

5. The computer-implemented method of claim 1 wherein the physiological information is temperature information, and wherein the physiological rank is a temperature rank of the user.

6. The computer-implemented method of claim 5 wherein the temperature rank is an output of a temperature value or a temperature status.

7. The computer-implemented method of claim 5 wherein the bodily parameter data is a body tissue absorption spectrum.

8. The computer-implemented method of claim 5 further comprising: acquiring other sensor information in addition to the physiological information; and/or, acquiring user input information, wherein the other sensor information includes one or more of hydration information obtained from a hydration sensor, heart rate information obtained from a heart rate sensor, blood pressure information obtained from a blood pressure sensor, activity information obtained from an accelerometer, and climate information obtained from a climate sensor.

9. The computer-implemented method of claim 5, further comprising: acquiring other sensor information in addition to the physiological information; and/or, acquiring user input information, wherein the user input information includes one or more of cervical mucous status and date of menstruation.

10. A computer-implemented method for deriving a physiological rank indicative of a physiological status of a user, the computer-implemented method comprising: applying a model to bodily parameter data acquired from a user to obtain physiological information related to the user; and, deriving, from the physiological information, a physiological rank indicative of a physiological status of the user wearing the device, wherein the physiological rank is a value on a physiological rank scale.

11. The computer-implemented method of claim 5, further comprising a step of determining a hydration status of a user, the computer-implemented method further comprising: acquiring, from a sensor on a wearable device worn by a user, data including additional bodily parameter data related to the user; and, applying a model to the bodily parameter data to obtain hydration information related to the user; wherein, the model deriving, from the hydration information, a hydration rank indicative of a hydration status of the user, wherein the hydration rank is a given grade on a hydration rank scale.

12. The computer-implemented method of claim 11 wherein each hydration rank on the hydration ranks scale maps onto a respective output of a standard clinical point of care test.

13. The computer-implemented method of claim 12 wherein the standard clinical point of care test is a test of urine osmolality, urine specific gravity, fluid gain, fluid loss, increases or decreases in body weight or mass representing fluid gain or fluid loss, respectively, or serum osmolality.

14. The computer-implemented method of claim 11 wherein the hydration rank is a hydration index, and wherein the hydration index is a given value on a hydration index scale.

15. The computer-implemented method of claim 14 wherein each hydration index on the hydration index scale maps onto a respective output of a standard clinical point of care test.

16. The computer-implemented method of claim 14 wherein the hydration index scale is sub-divided into a plurality of sub-ranges of hydration index values, each of the plurality of sub-ranges corresponding to a different clinical hydration status of the user, and wherein the method further comprises: determining which sub-range of the plurality of sub-ranges the hydration index value falls within.

17. The computer-implemented method of claim 11 wherein the hydration rank is a clinical hydration status of the user.

18. The computer-implemented method of claim 17 further comprising: outputting the clinical hydration status of the user, or the hydration index.

19. The computer-implemented method of claim 11 wherein the bodily parameter data includes a water absorption spectrum.

20. The computer-implemented method of claim 11 wherein the model includes a regression model.

21. The computer-implemented method of claim 5, further comprising a determination of how well a user's body is regulating an analyte, the computer-implemented method further comprising: acquiring, from a sensor on a wearable device worn by the user, data including bodily parameter data related to the user; and, applying a model to the bodily parameter data to obtain analyte concentration information related to the user, wherein, the model derives, from the analyte concentration information, a health score indicative of how well the user's body is regulating the analyte, wherein the health score is a given grade on a health score scale.

22. A wearable device comprising a processor, the processor configured to carry out the computer-implemented method of claim 1.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0516] These and other features and advantages of the present invention will be appreciated and understood with reference to the specification, claims, and appended drawings wherein:

[0517] FIG. 1 shows a flowchart of example steps of the computer-implemented method;

[0518] FIG. 2 is a schematic diagram of an optical sensing module that may be configured to carry out the computer-implemented method;

[0519] FIG. 3 shows a flowchart of example steps of the computer-implemented method;

[0520] FIG. 4 is a depiction of a further example of steps of a computer-implemented method;

[0521] FIG. 5 depicts an example of a temperature scale;

[0522] FIG. 6 shows a plot of temperature against time, and shows time-correlated user-input information;

[0523] FIG. 7A is an example of an output of the computer-implemented invention in the form of a graphical user interface (GUI), for example on a mobile device;

[0524] FIG. 7B is an example of an output of the computer-implemented invention in the form of a graphical user interface (GUI), for example on a mobile device;

[0525] FIG. 8 shows a further flowchart of example steps of the computer-implemented method;

[0526] FIG. 9 depicts an example of hydration information in the form of a hydration rank scale;

[0527] FIG. 10 is a depiction of a further example of steps of a computer-implemented method;

[0528] FIG. 11 is an example of an output of the computer-implemented invention in the form of a graphical user interface (GUI), for example on a mobile device;

[0529] FIG. 12A depicts a further example of hydration information in the form of a hydration rank scale; and

[0530] FIG. 12B depicts a further example of hydration information in the form of a hydration rank scale.

[0531] FIG. 13 shows a flowchart of example steps of the computer-implemented method;

[0532] FIG. 14 is a depiction of a further example of steps of a computer-implemented method;

[0533] FIG. 15 is a depiction of a further example of steps of a computer-implemented method;

[0534] FIG. 16 shows plots of blood glucose levels against time for a non-diabetic person and for a diabetic person;

[0535] FIG. 17A is a depiction of a range of blood glucose levels and of zones defined by sub-ranges of the blood glucose levels for the use case of a healthy user;

[0536] FIG. 17B is a depiction of a range of blood glucose levels and of zones defined by sub-ranges of the blood glucose levels for the use case of a diabetic user;

[0537] FIG. 18A is a depiction of a range of blood glucose levels, and of a total range of blood glucose levels which may occur in different use cases;

[0538] FIG. 18B is a depiction of a range of blood glucose levels, and of a total range of blood glucose levels which may occur in different use cases;

[0539] FIG. 19 is an example of an output of the computer-implemented invention in the form of a graphical user interface (GUI), for example on a mobile device;

[0540] FIG. 20 shows a further flowchart of example steps of the computer-implemented method.

[0541] FIG. 21 is an example of an output of the computer-implemented invention in the form of a graphical user interface (GUI), for example on a mobile device;

[0542] FIG. 22 is a depiction of a further example of steps of a computer-implemented method.

DETAILED DESCRIPTION

[0543] The detailed description set forth below in connection with the appended drawings is intended as a description of exemplary embodiments of a computer-implemented method provided in accordance with the present invention and is not intended to represent the only forms in which the present invention may be constructed or utilized. The description sets forth the features of the present invention in connection with the illustrated embodiments. It is to be understood, however, that the same or equivalent functions and structures may be accomplished by different embodiments that are also intended to be encompassed within the spirit and scope of the invention. As denoted elsewhere herein, like element numbers are intended to indicate like elements or features.

[0544] One or more embodiments of the present invention provide a computer-implemented method for deriving a physiological rank indicative of a physiological status of a user. The computer-implemented method comprises acquiring from a sensor 1101 on a wearable device worn by a user, data including bodily parameter data, for example an optical measurement such as an absorption spectrum related to the user.

[0545] The method further comprises applying a model to the bodily parameter data to obtain physiological information related to the user. This information could take the form, for example of information from the spectrum about the physiological status of the user (e.g. from the location/height of the peak). The model derives, from the physiological information, a physiological rank indicative of a physiological status of the user, wherein the physiological rank is a given grade on a physiological rank scale.

[0546] FIG. 1 shows a flow chart 1 setting out steps 2, 3, 4, 5, 6, 7 of the computer-implemented method. Computer-implemented methods according to other embodiments of the present invention may include some, but not all, of the steps shown in FIG. 1. Computer-implemented methods according to other embodiments of the present invention may include additional steps to the steps shown in FIG. 1. The first step 2 of the computer-implemented method shown in FIG. 1 is acquiring data from a sensor on the wearable device.

[0547] An example of an optical sensing module 1101 will now be described with reference to FIG. 2. The optical sensing module is typically located on the wearable device which acquires the data including the bodily parameter data (e.g. absorption spectrum) related to the user.

[0548] The optical sensing module 1101 includes a transmitter photonic integrated circuit (PIC) 4 located on a substrate 12. The PIC 14 includes a plurality of lasers (not visible in FIG. 3), each laser of the plurality of lasers operating at a wavelength that is different from the wavelength of the others. The optical sensing module 1101 is configured to drive the plurality of lasers one at a time. Light from the plurality of lasers exits the PIC 14 and therefore the optical sensing module 101 via one or more optical output ports. A mirror 10 is present to take the light from the plane of the PIC 14 and translate it into a direction more suitable for interrogating the surface. The direction is orthogonal or substantially orthogonal to the plane of the PIC 14.

[0549] The plurality of lasers emit light in a wavelength band which is sensitive to changes in water concentration within the sub corneal interstitial space. The plurality of lasers may emit light in the infrared wavelength band. The plurality of lasers may emit light in the near-infrared wavelength band. The plurality of lasers may emit light in the short wavelength infrared wavelength band. A laser within the plurality of lasers may emit light at 970 nm, 1200 nm, 1450 nm, 1950 nm, 2766 nm, 2898 nm, or 6097 nm, which correspond to water absorption peaks.

[0550] In other embodiments, the optical sensing module 1101 may include LEDs in addition to or instead of the lasers.

[0551] In use, emitted light from the plurality of lasers is transmitted towards the skin 13 of a user.

[0552] Back-scattered light from the surface of the skin 13, and from within a volume below the surface of the skin, returns to the optical sensing module 1101.

[0553] A photodetector array comprising photodetector pixels 1106, which collect the backscattered light, forms part of the optical sensing module 1101. In the example shown in FIG. 2, the photodetector array is located on the substrate 12 but is not part of the PIC 4.

[0554] An ASIC or microcontroller 11 is located on the substrate 12 of the optical sensing module 1101.

[0555] The wearable device carries out the computer-implemented method according to the present invention on a processor (e.g., on a processor of the microcontroller 11 of the wearable device). In other embodiments, an external device such as a mobile phone carries out the computer-implemented method according to the present invention on a processor of the external device.

[0556] When the data is acquired from optical sensing module 1110, or from other optical sensing modules, the bodily parameter data is a body tissue absorption spectrum where the absorption is in the water band. The physiological information is a quantitative value associated with the absorption spectrum, for example a wavelength shift of a peak of the absorption spectrum, a height of a peak of the absorption spectrum, or a width of a peak of the absorption spectrum.

[0557] In this way, an indication of clinical physiological status of the user can be provided, as the physiological information is sensitive to concentration changes of water within the skin sub-corneal interstitial fluid. As water in the dermis diminishes, the concentration of solutes become higher, thereby changing the degree of water absorption.

[0558] Returning to the flow chart 1 shown in FIG. 1, the second and third steps 3, 4 of the computer-implemented method shown in FIG. 1 respectively include carrying out data regression analysis 3 and determining a physiological rank 4. These steps are shown in more detail in FIG. 3.

[0559] FIG. 3 shows that the second and third steps 3, 4 of the computer-implemented method shown in FIG. 1, include firstly receiving the spectral data 200. Subsequently, pre-processing steps 202, 204 are applied to the data. The first pre-processing step 202 validates the data to determine whether the data has been acquired from a human user, or to determine whether the data includes outlying data. The second pre-processing step 204 applies a baseline correction to the data. This step will be described in more detail below with reference to FIG. 5. Subsequently, the spectra is fed into an offline model 206 and a physiological rank is calculated 208. These steps will be described in more detail below with reference to FIG. 4.

[0560] FIG. 4 shows the pre-processing step 204 of applying a baseline correction to the data in more detail. FIG. 4 shows that applying the baseline correction 204 to the data 20 comprises subtracting baseline data 21 from the data 20. The baseline data 21 is the average data acquired from the user over a long time period, for example over 24 hours, 1 week, or 1 month. A delta spectrum 22 results from the baseline correction 21 pre-processing step.

[0561] FIG. 4 further shows the step 206 of applying an offline model to the delta spectrum 22 to derive the physiological rank. In the example shown in FIG. 4, the offline model derives, as the physiological rank, both a physiological index and a physiological status of the user. The offline model includes an index model 23 which derives the physiological index, and a status model 25 which derives the physiological status. In the example shown in FIG. 4, the physiological index is a temperature, and the physiological status is a temperature status. The offline model further includes a confidence model 24 which calculates a confidence value for the derived physiological index.

[0562] The status model 25 comprises a PCA model and a logistic regression model, where the PCA model is applied to the delta spectrum 22, and the logistic regression model is applied to the output of the PCA model.

[0563] The index model 23 is a PLS regression model. The PLS regression model is applied to the delta spectrum.

[0564] Further details 210 of how the offline model can generated can be understood with reference to FIG. 3. Each of the index model 23 and the status model 25 of the offline model are generated or trained using training data. The index model and/or the status model may be machine learning models.

[0565] A first step 212 outputs of standard clinical tests are received. If the physiological rank is a temperature rank, the standard clinical test may be a test which measure temperature of the user orally using an oral thermometer, for example. In a second step 214 a training dataset is received. A third step 216, is a pre-processing step in which the training dataset is validated to determine whether the training data has been acquired from a human user, or to determine whether the training dataset includes outlying data. In a fourth step 218, the training dataset is mapped to the outputs of the standard clinical tests. In a fifth step 220, a physiological index value is calculated for the training dataset. In a sixth step 222, the offline model is developed.

[0566] Thus, training data is collected which comprises a plurality of training datasets, each training dataset being a dataset which comprises bodily parameter information. Each training dataset is associated with a clinical label, where each clinical label is associated with a respective plurality of outputs of standard clinical tests. The clinical data which is used to derive the outputs of the standard clinical tests are acquired at a similar time to the acquisition of the corresponding training dataset, for example within the same 5 minute interval, 15 minute interval or 1 hour interval.

[0567] In this way, the offline model effectively maps the data to outputs of standard clinical tests.

[0568] The required range of temperatures which can be derived from data acquired by the sensor may vary depending on the use case. As shown in FIG. 5, the required range of temperatures for the use case of women's health may be 35.0° C. to 41.1° C. The required range of temperatures for the use case of infection/fever determination may be 36.0° C. to 40.0° C. The required range of temperatures for the use case of thermoregulation may be 35.0° C. to 40.0° C. The required range of temperatures of the use case of critical care may be 28.0° C. to 43.0° C.

[0569] Returning to the computer-implemented method 1 shown in FIG. 1, the fourth step 5 of the computer-implemented method is integrating the data with other sensor information, user-input information or learnt metrics.

[0570] At this step, 5 the computer-implemented method comprises acquiring other sensor information. The other sensor information may include, for example, one or more of body temperature information obtained from a temperature sensor, heart rate information obtained from a heart rate sensor, blood oxygen saturation information obtained from a blood oxygen saturation sensor, respiratory rate information obtained from a respiratory rate sensor, hydration information obtained from a hydration sensor, accelerometer and motion information obtained from an accelerometer or a motion sensor, heart rate variability information obtained from a heart rate sensor, alcohol concentration, sleep/wake information obtained from a sleep sensor, blood pressure information obtained from a blood pressure sensor, analyte concentration information (wherein the analyte may be a metabolic fuel such as glucose, lactate, or ethanol) obtained from an analyte concentration sensor, and climate information obtained from a climate sensor (e.g., a thermometer).

[0571] User input information may be acquired from a user input into the wearable device or from a user input into an external device such as a mobile phone. The user input information may include, for example, one or more of weight information, height information, activity information, diet information, fluid intake information, sodium intake information, illness information, intoxication information, blood pressure information, sleep duration information, sleep quality information, gender information, age information, heart rate information, cervical mucous information and date of menstruation information.

[0572] The computer-implemented method may further comprise a step of acquiring a learnt basal physiological rank or basal bodily parameter data of the user. The basal physiological rank or basal bodily parameter data of the user may be learnt in a calibration period of the computer-implemented method using a machine learning model. The training data for the machine learning model may comprise a plurality of training basal datasets comprising bodily parameter data acquired from the wearable device, and a respective plurality of context labels. Each training basal dataset may be associated with a respective one of the context labels. Each of the context labels may include training other sensor information and/or training user input information. The training other sensor information or user input information is acquired at the same time as or at a similar time to the acquisition of the corresponding training basal dataset, for example within the same 5 minute interval, 15 minute interval or 1 hour interval.

[0573] Returning to the computer implemented method 1 shown in FIG. 1 the fifth step 6 of the computer-implemented method is determining a user prompt.

[0574] The user prompt which is determined may include a notification, time-correlated hydration information, or an indication that the user's hydration status is outside of a pre-determined basal hydration range. How each of these user prompts is determined will now be described.

[0575] To determine a notification, the computer-implemented method comprises storing a notification data table which associates notification with stored physiological ranks. The computer-implemented method comprises comparing the derived physiological rank with the stored physiological ranks respectively and, based on this comparison, selecting a notification to output to the user.

[0576] For example, the derived physiological rank may indicate that the user is above their basal temperature. The recommendation associated with this physiological rank may be for the user to drink water.

[0577] The notification data table may further associate notifications with stored other sensor information and/or user input information. The computer-implemented method may comprise comparing acquired other sensor information and/or acquired user input information with stored other sensor information and/or stored user input information respectively. Selecting the notification to output to the user may be further based upon this comparison.

[0578] For example, if a temperature is derived which is higher than a basal temperature, and user input/other sensor information indicates that the user is hydrated and has not conducted physical activity, the selected notification may indicate to the user that they may have an illness or infection.

[0579] To determine a notification, the computer-implemented method may comprise applying a user-notification machine learning model to the health score.

[0580] The user-notification machine learning model is trained using training data comprising a plurality of physiological ranks and a respective plurality of context labels. Each of the context labels may comprise user input information and/or other sensor information corresponding to the respective physiological ranks.

[0581] For example, the user-notification machine learning model may select a notification to output to the user based upon a derived temperature of the user and a user input indicative of where the user is in their menstrual cycle, such as their dates of menstruation or their cervical mucous state. As shown in FIG. 6, a user's basal temperature during the luteal phase 27 of their menstrual cycle is 0.2-0.3° C. higher than their basal temperature during the follicular 26 phase of their menstrual cycle. Further, as shown in FIG. 6, a user's basal temperature when they are pregnant 28 may be approximately 0.2-0.3° C. higher than the basal temperature of the user when they are not pregnant 26.

[0582] To determine time-correlated physiological information, the computer-implemented method may comprise a step of storing the physiological rank and storing the time at which the physiological rank is derived. The computer-implemented method thus comprises obtaining time-correlated physiological rank information from previously derived stored physiological ranks and their corresponding stored times.

[0583] The computer-implemented method may further comprise a step of storing the other sensor information and/or user input information. The computer-implemented method may thus comprise obtaining time-correlated other sensor information and/or time correlated user input information from previously derived stored other sensor information and/or user input information and their corresponding stored times.

[0584] To determine whether the user's physiological status is outside of a pre-determined basal physiological range, the computer-implemented method may comprise a step of applying a basal physiological model to the bodily parameter data, or the derived physiological rank.

[0585] In one or more embodiments, the basal physiological model takes the learnt basal bodily parameter data or learnt physiological rank as an input. The basal physiological model may derive whether the derived bodily parameter data, or derived physiological rank is more than a pre-determined threshold away from a user's basal bodily parameter data or basal physiological rank respectively.

[0586] Returning to the method 1 shown in FIG. 2, the sixth step 7 of the computer-implemented method shown in FIG. 1 is delivery of an output to the user.

[0587] The output may be displayed to the user on the wearable device, or on an external device such as a mobile phone.

[0588] FIGS. 7A and 7B show examples of an outputs 31, 34 which may displayed to a user. The output 31 in FIG. 7A includes a notification to the user that they may have ovulated 32, and the user's temperature in degrees Celsius 65. The output 34 in FIG. 7B includes a notification to the user that their temperature is slightly elevated 35, and the user's temperature in degrees Celsius and degrees Fahrenheit 36.

[0589] The output further may further include time-correlated temperature information 29 along with time-correlated user input information 30 and/or other sensor information 30 as shown in FIG. 6.

[0590] FIG. 8 shows a flow chart 40 for a method which includes embodiments of the computer-implemented method of the present invention. At a first step 41, the user wears a wearable device such that it is in contact with their skin. At a second step 42, data is acquired from sensors of the wearable device. At a third step 43, other sensor information, user input data, and learnt parameters are acquired. At a fourth step 44, an output is provided to the user. At a fifth step 45, the user makes a choice based on the output. At a sixth step 46, data is re-acquired from the sensors. The third step 43 to the sixth step 46 are carried out in a loop. At a seventh step 47, which follows the fifth step, time-correlated information and learnt parameters are acquired.

[0591] One or more embodiments of the present invention provide a computer-implemented method for determining a hydration status of a user. The computer-implemented method comprises acquiring from a sensor 1101 on a wearable device worn by a user, data including bodily parameter data, for example an optical measurement such as an absorption spectrum related to the user.

[0592] The method further comprises applying a model to the bodily parameter data to obtain hydration information related to the user. This information could take the form, for example of information from the spectrum about how hydrated the user is (e.g. from the location/height of the peak). The model derives, from the hydration information, a hydration rank 1 indicative of a hydration status of the user, wherein the hydration rank 1 is a given grade on a hydration rank scale 51.

[0593] FIG. 9 shows an example of a hydration rank scale 51 according to an embodiment of the present invention. In the embodiment shown in FIG. 9, the hydration rank scale 51 is a scale of hydration indices 10. Each of the hydration indices 50 are integer numbers, and the scale runs from −5 to +5. The model derives a hydration index 50 on this hydration index scale 51.

[0594] As shown in FIG. 9, each hydration index 50 on the hydration index scale 51 is indicative of a clinical hydration status of the user. The hydration index 12 scale is sub-divided into a plurality of sub-ranges 52 of hydration index values 1, each of the plurality of sub-ranges 52 corresponding to a different clinical hydration status of the user. These clinical hydration statuses include severe dehydration, moderate dehydration, mild dehydration, euvolemia, mild overhydration, moderate overhydration and severe overhydration. The sub-range 52 including hydration index −5 corresponds to severe dehydration, the sub-range including hydration indices −3 and −4 correspond to moderate dehydration, the sub-range 52 including hydration indices −1 and −2 correspond to mild dehydration, the sub-range 52 including hydration index 0 corresponds to euvolemia, the sub-range 52 including hydration indices +1 and +2 corresponds to mild overhydration, the sub-range including hydration indices +3 and +4 corresponds to moderate overhydration, and the sub-range 52 including hydration index +5 corresponds to severe overhydration.

[0595] As further shown in FIG. 9, each hydration index 50 on the hydration index scale 51 maps onto respective ranges of outputs of a plurality of standard clinical point of care tests 53a/b/c/d. In the example shown in FIG. 9, the standard clinical point of care tests 53a/b/c/d are tests of urine osmolality 53c, urine specific gravity 53d, fluid loss 53a and serum osmolality 53b.

[0596] Each of these standard clinical point of care tests 53a/b/c/d defines ranges of outputs of the standard clinical point of care test, wherein each of the ranges corresponds to a respective clinical hydration status.

[0597] In the example shown in FIG. 11, a hydration index value of −5, which indicates severe dehydration, corresponds to USG values of over 1.030, which also indicates severe dehydration. A sub-range including hydration index values of −3 or −4, which indicate moderate dehydration, corresponds to USG values of between 1.020 and 1.030, which also indicate moderate dehydration. A sub-range including hydration index values of −1 or −2, which indicate mild dehydration, corresponds to USG values of approximately 1.020, which also indicates mild dehydration. A sub-range including hydration index values of −2, −1, 0, 1 or 2, which indicate a hydration status between mild dehydration and mild overhydration, corresponds to USG values of between 1.005 and 1.020, which also indicate a hydration status between mild dehydration and mild overhydration. A hydration index value of 0, which indicates euvolemia corresponds to USG values of approximately 1.010, which also indicates euvolemia. A sub-range including hydration index values of +1 or +2, which indicate mild overhydration, corresponds to a USG value of approximately 1.005, which also indicates mild overhydration. A sub-range including hydration index values of +1, +2, +3, or +4, which indicates a hydration status between mild and moderate overhydration, corresponds to USG values of between 1.002 and 1.005, which also indicates a hydration status between mild and moderate overhydration. A hydration index value of +5, which indicates severe overhydration, corresponds to USG values of less than 1.002, which also indicates severe overhydration.

[0598] In other embodiments, the hydration rank scale 51 may be a scale of clinical hydration statuses. In this case, the model derives a clinical hydration status on this clinical hydration status scale. In such embodiments, the clinical hydration statuses (severe dehydration, moderate dehydration, mild dehydration, euvolemia, mild overhydration, moderate overhydration and severe overhydration) shown in FIG. 9, which map onto respective ranges of outputs of the plurality of standard clinical point of care tests, 16a/b/c/d, are directly derived by the model.

[0599] FIG. 1 shows a flow chart 1 setting out steps 2, 3, 4, 5, 6, 7 of the computer-implemented method. Computer-implemented methods according to other embodiments of the present invention may include some, but not all, of the steps shown in FIG. 1. Computer-implemented methods according to other embodiments of the present invention may include additional steps to the steps shown in FIG. 1. The first step 2 of the computer-implemented method shown in FIG. 1 is acquiring data from a sensor on the wearable device.

[0600] An example of an optical sensing module 1101 will now be described with reference to FIG. 2. The optical sensing module is typically located on the wearable device which acquires the data including the bodily parameter data (e.g. absorption spectrum) related to the user.

[0601] The optical sensing module 1101 includes a transmitter photonic integrated circuit (PIC) 4 located on a substrate 12. The PIC 14 includes a plurality of lasers (not visible in FIG. 2), each laser of the plurality of lasers operating at a wavelength that is different from the wavelength of the others. The optical sensing module 1101 is configured to drive the plurality of lasers one at a time. Light from the plurality of lasers exits the PIC 14 and therefore the optical sensing module 101 via one or more optical output ports. A mirror 10 is present to take the light from the plane of the PIC 14 and translate it into a direction more suitable for interrogating the surface. The direction is orthogonal or substantially orthogonal to the plane of the PIC 4.

[0602] The plurality of lasers emit light in a wavelength band which is sensitive to changes in water concentration within the interstitial space. The plurality of lasers may emit light in the infrared wavelength band. The plurality of lasers may emit light in the near-infrared wavelength band. The plurality of lasers may emit light in the short wavelength infrared wavelength band. A laser within the plurality of lasers may emit light at 970 nm, 1200 nm, 1450 nm, 1950 nm, 2766 nm, 2898 nm, or 6097 nm, which correspond to water absorption peaks.

[0603] In other embodiments, the optical sensing module 1101 may include LEDs in addition to or instead of the lasers.

[0604] In use, emitted light from the plurality of lasers is transmitted towards the skin 13 of a user.

[0605] Back-scattered light from the surface of the skin 13, and from within a volume below the surface of the skin, returns to the optical sensing module 1101.

[0606] A photodetector array comprising photodetector pixels 1106, which collect the backscattered light, forms part of the optical sensing module 1101. In the example shown in FIG. 2, the photodetector array is located on the substrate 12 but is not part of the PIC 4.

[0607] An ASIC or microcontroller 11 is located on the substrate 12 of the optical sensing module 1101.

[0608] The wearable device carries out the computer-implemented method according to the present invention on a processor (e.g., on a processor of the microcontroller 11 of the wearable device). In other embodiments, an external device such as a mobile phone carries out the computer-implemented method according to the present invention on a processor of the external device.

[0609] When the data is acquired from optical sensing module 1110, or from other optical sensing modules, the bodily parameter data is a body tissue absorption spectrum where the absorption is in the water band. The hydration information is a quantitative value associated with the absorption spectrum, for example a wavelength shift of a peak of the absorption spectrum, a height of a peak of the absorption spectrum, or a width of a peak of the absorption spectrum.

[0610] In this way, an indication of clinical hydration status of the user can be provided, as the hydration information is sensitive to concentration changes of water within the skin sub-corneal interstitial fluid. As water in the dermis diminishes, the concentration of solutes become higher, thereby changing the degree of water absorption.

[0611] Returning to the flow chart 1 shown in FIG. 1, the second and third steps 3, 4 of the computer-implemented method shown in FIG. 1 respectively include carrying out data regression analysis 3 and determining a hydration index value 4. These steps are shown in more detail in FIG. 3.

[0612] FIG. 3 shows that the second and third steps 3, 4 of the computer-implemented method shown in FIG. 1, include firstly receiving the spectral data 200. Subsequently, pre-processing steps 202, 204 are applied to the data. The first pre-processing step 202 validates the data to determine whether the data has been acquired from a human user, or to determine whether the data includes outlying data. The second pre-processing step 204 applies a baseline correction to the data. This step will be described in more detail below with reference to FIG. 10. Subsequently, the spectra is fed into an offline model 206 and a hydration index value is calculated 208. These steps will be described in more detail below with reference to FIG. 10.

[0613] FIG. 10 shows the pre-processing step 204 of applying a baseline correction to the data in more detail. FIG. 10 shows that applying the baseline correction 204 to the data 20 comprises subtracting baseline data 21 from the data 20. The baseline data 21 is the average data acquired from the user over a long time period, for example over 24 hours, 1 week, or 1 month. A delta spectrum 22 results from the baseline correction 21 pre-processing step.

[0614] FIG. 10 further shows the step 206 of applying an offline model to the delta spectrum 22 to derive the hydration rank 1. In the example shown in FIG. 10, the offline model derives, as the hydration rank 1, both a hydration index and a hydration status of the user. The offline model includes an index model 23 which derives the hydration index, and a status model 25 which derives the hydration status.

[0615] The status model 25 comprises a PCA model and a logistic regression model, where the PCA model is applied to the delta spectrum 22, and the logistic regression model is applied to the output of the PCA model.

[0616] The index model 23 is a PLS regression model. The PLS regression model is applied to the delta spectrum.

[0617] Further details 210 of how the offline model can generated can be understood with reference to FIG. 3. Each of the index model 23 and the status model 25 of the offline model are generated or trained using training data. The index model and/or the status model may be machine learning models.

[0618] A first step 212 outputs of standard clinical point of care tests are received, and in a second step 214 a training dataset is received. A third step 216, is a pre-processing step in which the training dataset is validated to determine whether the training data has been acquired from a human user, or to determine whether the training dataset includes outlying data. In a fourth step 218, the training dataset is mapped to the outputs of the standard clinical point of care tests. In a fifth step 220, a hydration index value is calculated for the training dataset. In a sixth step 222, the offline model is developed.

[0619] Thus, training data is collected which comprises a plurality of training datasets, each training dataset being a dataset which comprises bodily parameter information. Each training dataset is associated with a clinical label, where each clinical label is associated with a respective plurality of outputs of standard clinical point of care tests. The clinical data which is used to derive the outputs of the standard clinical point of care tests are acquired at a similar time to the acquisition of the corresponding training dataset, for example within the same 5 minute interval, 15 minute interval or 1 hour interval.

[0620] In this way, the offline model effectively maps the data to outputs of standard point of care tests.

[0621] Returning to the computer-implemented method 1 shown in FIG. 1, the fourth step 5 of the computer-implemented method is integrating the data with other sensor information, user-input information or learnt metrics.

[0622] At this step, 5 the computer-implemented method comprises acquiring other sensor information. The other sensor information may include, for example, body temperature obtained from a temperature sensor, activity information obtained from an accelerometer, heart rate information obtained from a heart rate sensor and blood pressure information obtained from a blood pressure sensor.

[0623] User input information may be acquired from a user input into the wearable device or from a user input into an external device such as a mobile phone. The user input information may include, for example, weight information, activity information, diet information, fluid intake information, illness information and intoxication information.

[0624] The computer-implemented method may further comprise a step of acquiring a learnt basal hydration rank or basal bodily parameter data of the user. The basal hydration rank or basal bodily parameter data of the user may be learnt in a calibration period of the computer-implemented method using a machine learning model. The training data for the machine learning model may comprise a plurality of training basal datasets comprising bodily parameter data acquired from the wearable device, and a respective plurality of context labels. Each training basal dataset may be associated with a respective one of the context labels. Each of the context labels may include training other sensor information and/or training user input information. The training other sensor information or user input information is acquired at the same time as or at a similar time to the acquisition of the corresponding training basal dataset, for example within the same 5 minute interval, 15 minute interval or 1 hour interval.

[0625] Returning to the computer implemented method 1 shown in FIG. 1, the fifth step 6 of the computer-implemented method is determining a user prompt.

[0626] The user prompt which is determined may include a cause of a hydration status, a recommendation, a type of rehydration fluid to consume, a volume of rehydration fluid for the user to consume, a rehydration schedule for the user to follow, time-correlated hydration information, an indication that the user's hydration status is outside of a pre-determined basal hydration range, or an indication that the user remains dehydrated or overhydrated after a reassessment time. How each of these user prompts is determined will now be described.

[0627] To determine the cause of a hydration status, the computer-implemented method may comprise storing a hydration status cause data table which associates causes of a hydration status with stored other sensor information and/or stored user input information. The computer-implemented method may comprise, when a hydration rank is determined which indicates that the user's clinical hydration status is a pre-determined clinical hydration status, comparing acquired other sensor information and/or user input information with stored other sensor information and/or user input information respectively, and based on this comparison, selecting a cause of a clinical hydration status.

[0628] The pre-determined clinical hydration status may be dehydration or overhydration.

[0629] For example, the derived hydration index may indicate that the user is dehydrated, and the other sensor information may include temperature information which indicates that the user has a high temperature. The cause of dehydration associated with this temperature information may be, for example, that the user is ill, or that the user is overheated.

[0630] To determine a recommendation, the computer-implemented method comprises storing a recommendation data table which associates recommendations with stored hydration ranks. The computer-implemented method comprises comparing the derived hydration rank with the stored hydration ranks respectively and, based on this comparison, selecting a recommendation to output to the user.

[0631] For example, the derived hydration rank may indicate that the user is dehydrated. The recommendation associated with this hydration rank may be for the user to drink water.

[0632] The recommendation data table may further associate recommendations with stored other sensor information and/or user input information. The computer-implemented method may comprise comparing acquired other sensor information and/or acquired user input information with stored other sensor information and/or stored user input information respectively. Selecting the recommendation to output to the user may be further based upon this comparison.

[0633] For example, the derived hydration rank may indicate that the user is dehydrated, and activity information obtained from an accelerometer may indicate that the user is doing/has just finished doing exercise. The recommendation associated with this activity information for this hydration rank may be to rest, or to consume electrolyte fluid.

[0634] To determine a type of rehydration fluid to consume, the computer-implemented method comprises storing a rehydration fluid type data table. The rehydration fluid type data table associates types of rehydration fluids with stored other sensor information and/or stored user input information.

[0635] The computer-implemented method comprises, when a hydration rank is derived which indicates that the clinical hydration status of the user is dehydration, comparing acquired other sensor information and/or acquired user input information with stored other sensor information and/or stored user input information respectively and, based on this comparison, selecting a type of rehydration fluid.

[0636] For example, the derived hydration rank may indicate that the user is dehydrated, and activity information obtained from an accelerometer may indicate that the user is doing/has just finished doing exercise. The rehydration fluid selected in this case may be an electrolyte fluid.

[0637] To determine a volume of rehydration fluid for the user to consume, the computer-implemented method comprises storing a rehydration fluid volume data table which associates volumes of rehydration fluid with stored hydration ranks.

[0638] The computer-implemented method comprises, when a hydration rank is derived which indicates that the clinical hydration status if the user is dehydration, comparing a derived hydration rank with stored hydration ranks and, based on this comparison, selecting a volume of rehydration fluid.

[0639] The rehydration fluid volume data table may further associate volumes of rehydration fluid with stored other sensor information and/or user input information.

[0640] The computer-implemented method may comprise, when a hydration rank is derived which indicates that the clinical hydration status of the user is dehydration, comparing acquired other sensor information with stored other sensor information and, based on this comparison, selecting a volume of rehydration fluid. Alternatively, or in addition, the computer-implemented method may comprise, when a hydration rank is derived which indicates that the clinical hydration status of the user is dehydration, comparing acquired user input information with stored user input information and, based on this comparison, selecting a volume of rehydration fluid.

[0641] To determine a rehydration schedule for the user to follow, the computer-implemented method may comprise storing a rehydration schedule data table which associates a schedule by which rehydration fluid should be consumed with stored hydration ranks.

[0642] The computer-implemented method may comprise, when a hydration rank is derived which indicates that the clinical hydration status of the user is dehydration, comparing a derived hydration rank with stored hydration ranks and based on this comparison, selecting a schedule by which rehydration fluid should be consumed.

[0643] The rehydration fluid volume data table may further associate a schedule by which rehydration fluid should be consumed with stored user input information and/or other sensor information.

[0644] The computer-implemented method may comprise, when a hydration rank is derived which indicates that the clinical hydration status of the user is dehydration, comparing acquired other sensor information or user input information with stored other sensor information or user input information respectively and, based on this comparison, selecting a schedule by which rehydration fluid should be consumed.

[0645] For example, the hydration rank may indicate that the user is moderately dehydrated and temperature information acquired from a temperature sensor may indicate that the user is overheated. The selected schedule by which rehydration fluid should be consumed may be selected based on these factors. The schedule may include a type of rehydration fluid to consume, an overall volume of rehydration fluid to consume, sub-volumes of the overall volume of rehydration fluid to consume and times at which the sub-volumes of rehydration fluid should be consumed.

[0646] To determine time-correlated hydration information, the computer-implemented method may comprise a step of storing the hydration rank and storing the time at which the hydration rank is derived. The computer-implemented method thus comprises obtaining time-correlated hydration rank information from previously derived stored hydration ranks and their corresponding stored times.

[0647] The computer-implemented method may further comprise a step of storing the other sensor information and/or user input information. The computer-implemented method may thus comprise obtaining time-correlated other sensor information and/or time correlated user input information from previously derived stored other sensor information and/or user input information and their corresponding stored times.

[0648] To determine whether the user's hydration status is outside of a pre-determined basal hydration range, the computer-implemented method may comprise a step of applying a basal hydration model to the bodily parameter data, or the derived hydration rank.

[0649] In one or more embodiments, the basal hydration model takes the learnt basal bodily parameter data or learnt hydration rank as an input. The basal hydration model may derive whether the derived bodily parameter data, or derived hydration rank is more than a pre-determined threshold away from a user's basal bodily parameter data or basal hydration rank respectively.

[0650] To determine whether the user remains dehydrated or overhydrated after a reassessment time, the computer-implemented method comprises re-deriving the hydration rank after a reassessment time. The reassessment time may be a pre-determined reassessment time, a user-input reassessment time, or a reassessment time based upon the initially derived hydration rank.

[0651] Returning to the method 1 shown in FIG. 1, the sixth step 7 of the computer-implemented method shown in FIG. 1 is delivery of an output to the user.

[0652] The output may be displayed to the user on the wearable device, or on an external device such as a mobile phone.

[0653] FIG. 11 shows an example of an output being displayed to a user on a mobile device which is a mobile phone 54. The output includes a derived hydration index 50 of −4, and a derived clinical hydration status 55 of mildly dehydrated.

[0654] The output further includes a recommended rehydration schedule 56 which is to drink 500 ml of electrolyte fluid at a current time, 300 ml of water after half an hour, and another 300 ml of water after an hour.

[0655] The output also includes time-correlated hydration information 57 which shows that the user's hydration index has fallen from 0 to −4 over a period of time.

[0656] FIG. 11 also shows that a user input may be received on a portion 58 of the user interface of the mobile device which displays “Start Rehydration Timer” to the user to cause the computer-implemented method to re-derive the hydration status or the hydration index of the user after a pre-determined reassessment time. The computer-implemented method may comprise, when a hydration status or a hydration index, derived after the reassessment time, indicates that the user remains dehydrated after the reassessment time, outputting an alert to the user.

[0657] FIG. 8 shows a flow chart 40 for a method which includes embodiments of the computer-implemented method of the present invention. At a first step 41, the user wears a wearable device such that it is in contact with their skin. At a second step 42, data is acquired from sensors of the wearable device. At a third step 43, other sensor information, user input data, and learnt parameters are acquired. At a fourth step 44, an output is provided to the user. At a fifth step 45, the user makes a choice based on the output. At a sixth step 46, data is re-acquired from the sensors. The third step 43 to the sixth step 46 are carried out in a loop. At a seventh step 47, which follows the fifth step, time-correlated information and learnt parameters are acquired.

[0658] FIGS. 16 and 16 show examples of different hydration index scales 12 to that shown in FIG. 19. As shown in FIGS. 9, 16A and 16B, the hydration index scale 51, and how its hydration indices 50 map onto output of standard clinical point of care tests 16a/b/c/d, may depend on the use case for the computer-implemented method. For example, FIG. 19 shows a hydration index scale 51 running from −5 to +5 which may be used for critical care patients. FIG. 16A shows a hydration index scale 51 running from −9 to 0 which may be used for both healthy and vulnerable users. FIG. 16B shows a hydration index scale 51 running from −5 to +5 which may be used for healthy, high-risk users such as athletes. Populations that may benefit from the present invention include elderly populations, endurance athletes, those travelling to altitude or hot climates, and those in occupations with a high risk of dehydration and overheating.

[0659] One or more embodiments of the present invention provide a computer-implemented method for determining how well a user's body is regulating an analyte. The computer-implemented method comprises acquiring from a sensor 1101 on a wearable device worn by a user, data including bodily parameter data related to the user, for example an optical measurement such as a diffuse reflectance spectrum related to the user.

[0660] The method further comprises applying a model to the bodily parameter data to obtain analyte concentration information related to the user. The information could take the form, for example, of information from the spectrum about the levels of an analyte such as glucose in the user's blood (e.g., from the location/height of the peak). The model derives, from the analyte concentration information, a health score indicative of how well the user's body is regulating the analyte, wherein the health score is a given grade on a health score scale.

[0661] FIG. 1 shows a flow chart 1 setting out steps 2, 3, 4, 5, 6, 7 of the computer-implemented method. Computer-implemented methods according to other embodiments of the present invention may include some, but not all, of the steps shown in FIG. 1. Computer-implemented methods according to other embodiments of the present invention may include additional steps to the steps shown in FIG. 1. The first step 2 of the computer-implemented method shown in FIG. 1 is acquiring data from a sensor on the wearable device.

[0662] An example of an optical sensing module 1101 will now be described with reference to FIG. 2. The optical sensing module 1101 is typically located on the wearable device which acquires the data including the bodily parameter data (e.g. absorption spectrum) related to the user.

[0663] The optical sensing module 1101 includes a transmitter photonic integrated circuit (PIC) 4 located on a substrate 12. The PIC 14 includes a plurality of lasers (not visible in FIG. 2) each laser of the plurality of lasers operating at a wavelength that is different from the wavelength of the others. The optical sensing module 1101 is configured to drive the plurality of lasers one at a time. Light from the plurality of lasers exits the PIC 14 and therefore the optical sensing module 101 via one or more optical output ports. A mirror 10 is present to take the light from the plane of the PIC 14 and translate it into a direction more suitable for interrogating the surface. The direction is orthogonal or substantially orthogonal to the plane of the PIC 4.

[0664] The plurality of lasers emit light in a wavelength band which is sensitive to changes in sensitive to changes in concentrations of the analyte for which the analyte concentration value is determined. The plurality of lasers may emit light in the infrared wavelength band. The plurality of lasers may emit light in the near-infrared wavelength band. The plurality of lasers may emit light in the short wavelength infrared wavelength band.

[0665] In other embodiments, the optical sensing module 1101 may include LEDs in addition to or instead of the lasers.

[0666] In use, emitted light from the plurality of lasers is transmitted towards the skin 13 of a user.

[0667] Back-scattered light from the surface of the skin 13, and from within a volume below the surface of the skin at a depth at which sufficient blood is present for sensing the presence of the analyte, returns to the optical sensing module 1101.

[0668] A photodetector array comprising photodetector pixels 1106, which collect the backscattered light, forms part of the optical sensing module 1101. In the example shown in FIG. 2, the photodetector array is located on the substrate 12 but is not part of the PIC 4.

[0669] An ASIC or microcontroller 11 is located on the substrate 12 of the optical sensing module 1101.

[0670] The wearable device carries out the computer-implemented method according to the present invention on a processor (e.g., on a processor of the microcontroller 11 of the wearable device). In other embodiments, an external device such as a mobile phone carries out the computer-implemented method according to the present invention on a processor of the external device.

[0671] When the data is acquired from optical sensing module 1110, or from other optical sensing modules, the bodily parameter data is an infrared diffuse reflectance spectrum. The analyte concentration information is a quantitative value associated with the infrared diffuse reflectance spectrum, for example a wavelength shift of a peak of the reflectance spectrum, a height of a peak of the reflectance spectrum, or a width of a peak of the reflectance spectrum.

[0672] In this way, an indication of how well the user's body is regulating the analyte in the blood can be provided, as the analyte concentration information is sensitive to concentration changes of the analyte in the blood.

[0673] Returning to the flow chart 1 shown in FIG. 1 the second and third steps 3, 4 of the computer-implemented method shown in FIG. 1 respectively include carrying out data regression analysis 3 and determining a health score 4. These steps are shown in more detail in FIG. 13.

[0674] FIG. 13 shows that the second and third steps 3, 4 of the computer-implemented method shown in FIG. 1, include firstly receiving the data 300. Subsequently, pre-processing steps 302, 304 are applied to the data. The first pre-processing step 302 validates the data to determine whether the data has been acquired from a human user, or to determine whether the data includes outlying data. The second pre-processing step 304 applies a baseline correction to the data. Subsequently, the spectra is fed into an offline model 306 and a health score is calculated 3208. These steps 300, 302, 304, 306, 308 are further depicted in FIG. 4.

[0675] FIG. 14 shows the steps of receiving the spectral data 300, of validating the data 302 and of applying a baseline correction to the data 304.

[0676] FIG. 14 further shows the step 3061 of applying a PLS regression model to the data output by the pre-processing step 304. The PLS regression model forms a part of the offline model. The PLS regression model outputs an analyte concentration value.

[0677] FIG. 14 further shows that a step 3062 of applying post-processing to outputs of the PLS regression model is carried out, and that subsequently the health score is calculated. The post-processing step 3062 and the calculation of the health score is shown in more detail in FIG. 15.

[0678] In the embodiment depicted in FIG. 15 the offline model is applied to a plurality of datasets acquired from the sensor on the wearable device at different times, and a respective plurality of analyte concentration values are derived by the PLS regression model.

[0679] FIG. 15 shows that the first post-processing step 3062a is smoothing the time profile of the respective plurality of analyte concentration values. The second post-processing step 3062b, is truncating the time profile of the plurality of analyte concentration values to a desired time range.

[0680] After the post-processing steps 3062a, 3062b are carried out, the health score is derived 308. As shown in FIG. 15, there are multiple methods by which the health score can be derived. These methods include calculating a duration in which a user's analyte concentration values remain within a given sub-range, calculating a mean analyte concentration value, and calculating a variability of a user's analyte concentration values. These methods will now be discussed in more detail.

[0681] To calculate the health score from a duration in which the user's analyte concentration values remain within a given sub-range, the computer-implemented method includes the following steps.

[0682] The computer-implemented method includes, for each derived analyte concentration value in the plurality of derived analyte concentration values, determining a sub-range (or zone) of analyte concentration values which the derived analyte concentration value falls within. The computer-implemented method further includes acquiring a plurality of time stamps indicative of the respective plurality of times at which each of the datasets in the plurality of datasets are acquired. The computer-implemented method further includes calculating, using the plurality of sub-ranges and the plurality of time stamps, a duration of time in which a user's analyte concentration remains within a given sub-range. This duration is used to derive the health score.

[0683] For example, if a user's analyte concentration remains within a sub-range (or zone) which is indicative of high analyte concentration for a long time, a health score may be derived which indicates that a user's body is sub-optimally regulating the analyte. If a user's analyte concentration remains within a sub-range (or zone) which is indicative of an ideal analyte concentration for a long time, a health score may be derived which indicates that a user's body is optimally regulating the analyte. This concept is demonstrated in FIG. 16.

[0684] FIG. 16 is a graph of blood glucose levels against time. The blood glucose levels are divided into zones 60a/b/c. Zone 1 60a indicates ideal blood glucose levels during sleep and fasting. In a 24 hour period, the ideal duration that a user's blood glucose levels are in zone 1 60a is approximately greater than 19 hours. Zone 2 60b indicates ideal blood glucose levels within a 2 hour period after a meal. In a 24 hour period, the ideal duration that a user's blood glucose levels are in zone 2 60b is less than approximately less than 5 hours. Zone 3 60c indicates hyperglycemia (high blood glucose levels). In a 24 hour period, the ideal duration that a user's blood glucose levels are in zone 3 60c may be less than 30 minutes.

[0685] FIG. 16 shows a plot of the blood glucose levels of a non-diabetic person (the lower line 61), and a plot of the blood glucose levels of a diabetic person (the upper line 62). As shown in FIG. 4, the blood glucose levels of the diabetic person spend a longer duration in zones 2 and 3 than the blood glucose levels of the non-diabetic person. The blood glucose levels of the non-diabetic person spend a longer duration in zone 1 than the blood glucose levels of the diabetic person.

[0686] The sub-ranges of the blood glucose levels which define the zones, and the number of zones used in the computer-implemented method may vary based on the use case and the utility of the computer-implemented method. FIG. 17A and FIG. 17B show the number of zones and the sub-ranges of the blood glucose levels which define the zones for non-diabetic users and diabetic users respectively. FIG. 17A shows 2 zones 63a/b, which can be applied when the computer-implemented method is utilized only in fasting periods. FIG. 17A further shows 5 zones 64a/b/c/d/e, which can be applied when the computer-implemented method is utilized in both fasting periods and meal periods, or which can be applied when the computer-implemented method is utilized over a 24-hour period. FIG. 17B shows 6 zones 65a/b/c/d/e/f.

[0687] FIG. 18A and FIG. 18B show total ranges of glucose levels which may occur in different use cases of prediabetic users, healthy users, health and fitness users or public service users, users with uncontrolled diabetes, users with controlled diabetes, and users with gestational diabetes. It may be useful to be able to obtain data from the sensor across these total ranges in each of these use cases.

[0688] To calculate the health score from a mean analyte concentration value, the computer-implemented method includes calculating, using the plurality of analyte concentration values, a mean of the plurality of analyte concentration values.

[0689] To calculate the health score from the variability of the user's analyte concentration values, the computer-implemented method includes calculating, using the plurality of analyte concentration values, a variability of the plurality of analyte concentration values. The variability may be a standard deviation, a coefficient of variation, a mean amplitude glucose excursion, a J-index, a mean absolute difference, or a mean absolute glucose.

[0690] A higher variability may indicate that the user's body is not regulating the analyte as well, and thus a higher variability may result in a lower health score. A lower variability may indicate that the user's body is regulating the analyte better, and thus a lower variability may result in a higher health score. This concept is demonstrated in FIG. 19.

[0691] FIG. 19 shows three graphs 66, 67, 68 of blood glucose levels against time. The top graph 66 shows blood glucose levels which are well regulated, the middle graph 67 shows blood glucose levels which are moderately well regulated, and the bottom graph 68 shows blood glucose levels which are badly regulated. The variability of the blood glucose levels increases as the blood glucose levels become less well regulated (i.e., the variability is smallest in the top graph, and is greatest in the bottom graph).

[0692] FIG. 15 further shows that the computer-implemented method includes outputting various outputs, which may include the health score, a mean analyte concentration and time-correlated information. The outputs are discussed in more detail below with reference to FIG. 19.

[0693] Details of how the offline model can generated can be understood with reference to FIG. 13. At a first step, 312 outputs of standard clinical analyte concentration tests are received, and other sensor information and/or user input information are received. In a second step 314 training datasets, each training dataset comprising bodily parameter data, which are diffuse reflectance spectra, are received. A third step 316 is a pre-processing step in which each training dataset is validated to determine whether the training dataset has been acquired from a human user, or to determine whether the training dataset includes outlying data. In a fourth step 318, the training datasets are mapped to the outputs of the standard clinical analyte concentration tests. In a fifth step 320, a health score is calculated for the training datasets using the outputs of the standard clinical analyte concentration tests. In a sixth step 322, the offline model is developed using a PCA model and a regression model. In a seventh step 324, trends between the health score and the other sensor information and/or user input information are investigated. The seventh step 324 is discussed in more detail below.

[0694] Thus, training data is collected which comprises a plurality of training datasets, each training dataset being a dataset which comprises bodily parameter information. Each training dataset is associated with a clinical label, where each clinical label is associated with a respective output of the standard clinical analyte concentration test outputs. The clinical data which is used to derive the standard clinical analyte concentration test outputs are acquired at a similar time to the acquisition of the corresponding training dataset, for example within the same 5 minute interval, 15 minute interval or 1 hour interval.

[0695] In this way, the offline model effectively maps the data to outputs of standard clinical analyte concentration tests.

[0696] Returning to the computer-implemented method 1 shown in FIG. 1, the fourth step 5 of the computer-implemented method is integrating the data with other sensor information, user-input information or learnt metrics.

[0697] At this step, 5 the computer-implemented method comprises acquiring other sensor information and/or user input information.

[0698] The other sensor information may include, for example, body temperature information obtained from a temperature sensor, heart rate information obtained from a heart rate sensor, blood oxygen saturation information obtained from a blood oxygen saturation sensor, respiratory rate information obtained from a respiratory rate sensor, hydration information obtained from a hydration sensor, accelerometer and motion information obtained from an accelerometer or a motion sensor, heart rate variability information obtained from a heart rate sensor, alcohol concentration, sleep/wake information obtained from a sleep sensor, and blood pressure information obtained from a blood pressure sensor.

[0699] The user input information may be acquired from a user input into the wearable device or from a user input into an external device such as a mobile phone. The user input information may include, for example, weight information, height information, activity information, diet information, fluid intake information, sodium intake information, illness information, intoxication information, blood pressure information, sleep duration information, sleep quality information, gender information, age information and heart rate information.

[0700] Returning to the computer implemented method 1 shown in FIG. 1, the fifth step 6 of the computer-implemented method is determining a user prompt. The user prompt which is determined may include a recommendation, time-correlated health score information and/or time correlated analyte concentration value information.

[0701] To determine a recommendation, the computer-implemented method may comprise storing a recommendation data table which associates recommendations with stored parameters. In this case, the computer-implemented method comprises comparing a derived parameter value with the stored parameter values respectively and, based on this comparison, selecting a recommendation to output to the user.

[0702] The parameter values may be health scores, durations of time in which a user's analyte concentration remains within a given sub-range, rates at which a user's analyte concentration changes, analyte concentration values, or peak analyte concentration values of a plurality of analyte concentration values.

[0703] For example, the derived parameter value may indicate that the user's blood glucose level is too high. The recommendation associated with this parameter value may be for the user to conduct activity.

[0704] The recommendation data table may further associate recommendations with stored other sensor information and/or user input information. The computer-implemented method may comprise comparing acquired other sensor information and/or acquired user input information with stored other sensor information and/or stored user input information respectively. Selecting the recommendation to output to the user may be further based upon this comparison.

[0705] To determine a recommendation, the computer-implemented method may comprise applying a user-recommendation machine learning model to the health score.

[0706] The user-recommendation machine learning model is trained using training data comprising a plurality of health scores and a respective plurality of context labels. Each of the context labels may comprise user input information and/or other sensor information corresponding to the respective health score. As shown in FIG. 19, the user input information and/or other sensor information may be collected at step 1 212 of generating the offline model. Applying the context labels to the health score may be carried out at step 7 324 of generating the offline model.

[0707] To determine time-correlated health score information, the computer-implemented method may comprise a step of storing the health score and storing the time at which the health score is derived. The computer-implemented method thus comprises obtaining time-correlated health score information from previously derived stored health scores and their corresponding stored times.

[0708] To determine time-correlated analyte concentration value information, the computer-implemented method may comprise a step of storing the derived analyte concentration value and storing the time at which the analyte concentration value is derived. The computer-implemented method thus comprises obtaining, from previously derived stored analyte concentration values and their corresponding stored times, time-correlated analyte concentration value information.

[0709] The computer-implemented method may further comprise a step of storing the other sensor information and/or user input information. The computer-implemented method may thus comprise obtaining time-correlated other sensor information and/or time correlated user input information from previously derived stored other sensor information and/or user input information and their corresponding stored times.

[0710] Returning to the method 1 shown in FIG. 1 the sixth step 7 of the computer-implemented method shown in FIG. 1 is delivery of an output to the user.

[0711] The output may be displayed to the user on the wearable device, or on an external device such as a mobile phone.

[0712] FIG. 20 shows an example of an output being displayed to a user on a mobile device which is a mobile phone 54. The output includes a derived health score 70 which is 90%.

[0713] The output further includes a bar chart 71 showing the duration that the user's analyte concentration values spent in each zone over a period of approximately 10 hours.

[0714] The output also includes time-correlated health score information 72, which shows the variation of a user's health score over a period of 12 days.

[0715] FIG. 8 shows a flow chart 40 for a method which includes embodiments of the computer-implemented method of the present invention. At a first step 41, the user wears a wearable device such that it is in contact with their skin. At a second step 42, data is acquired from sensors of the wearable device. At a third step 43, other sensor information, user input data, and learnt parameters are acquired. At a fourth step 44, an output is provided to the user. At a fifth step 45, the user makes a choice based on the output. At a sixth step 46, data is re-acquired from the sensors. The third step 43 to the sixth step 46 are carried out in a loop. At a seventh step 47, which follows the fifth step, time-correlated information and learnt parameters are acquired.

[0716] According to an embodiment of the invention, the computer-implemented method may comprise deriving a plurality of physiological ranks. For example, the computer-implemented method may comprise deriving a temperature, a hydration rank and a health score. FIG. 21 shows an example of an output which may be output to a user. The output includes the user's health score 70 hydration index 50, hydration status 52, temperature 36, and a notification 34 that the user's temperature is elevated.

[0717] As shown in FIG. 22, the model which derives the physiological rank may include a short-wavelength infrared (SWIR) algorithm 91. The SWIR algorithm may take SWIR data 93 which includes bodily parameter data as an input and may output physiological information. As shown in FIG. 1, the SWIR algorithm may further take as inputs, the outputs of other algorithms 95a/b. In this way, the SWIR algorithm may be calibrated. The SWIR algorithm and the other algorithms may further take as an input user entered demographic information 97. In this way, the SWIR algorithm and the other algorithms may be calibrated.

[0718] As further shown in FIG. 22, each model may include a step 99 of deriving the physiological rank and/or recommendations or notifications etc. The model may further include the step of storing time-correlated physiological rank information 101.

[0719] As further shown in FIG. 22, derived physiological ranks and/or physiological information from each of the models may be combined. For example, the computer-implemented method may comprise outputting a recommendation based upon both the derived temperature and the derived hydration index. Further, the computer-implemented method may comprise deriving an overall rank based on the combination of the derived temperature and the derived hydration index.

[0720] Although exemplary embodiments of a computer-implemented method have been specifically described and illustrated herein, many modifications and variations will be apparent to those skilled in the art. Accordingly, it is to be understood that a computer-implemented method constructed according to principles of this invention may be embodied other than as specifically described herein. The invention is also defined in the following claims, and equivalents thereof.