In an example, the optoelectronic circuit board 300 may be fixed to the bottom wall 102, 202 of the back case 100, 200 by adhesive, by coupling elements or any other suitable method.

Once the optoelectronic circuit board 300 is assembled within the inner space of the housing of a back case 100, 200 according to any of the disclosed examples, the resulting back case may be ready to be attached to or to replace an existing back case of a wrist watch or other timekeeping device. When the proposed back case is to be attached to an existing back case a thin cover may be placed between the case and the watch body for protection, e.g. against moisture accumulation.

The assembled back case, i.e. the back case having the optoelectronic circuit board within the housing, may have a thickness determined by the thickness of the height of the side wall. In an example, the battery may have a thickness of about 1.9 mm, the charging coil may have a thickness of about 0.4 mm and the side wall a height of about 0.8-2 mm.

In an example, the optoelectronic circuit board 300 may comprise more than two paired radiation sources e.g. near infrared, short wave infrared, infrared, mid-infrared, etc. for measuring tissue constituents such as water, lipids e.g. yellow and brown fat, and glucose. Additional and related biomarkers, e.g. hydration, lipid metabolic activity, blood sugar levels and glucose metabolic activity, may also be then calculated.

The back case and the optoelectronic circuit board may be manufactured e.g. by CNC machining, milling, moulding, press fitting, 3D printing, etc. Any suitable material such as metal e.g. stainless steel, titanium, aluminium, gold, silver; plastic, composite materials, carbon fibre or any combination may be used to manufacture a back case and/or an optoelectronic circuit board according to any of the disclosed examples.

FIG. 4 depicts the bottom view of a watchband or a timekeeping band 400 to be attached around a user wrist in an adjustable manner. The watchband 400 may comprise a flexible layer 401 e.g. made of fabric, plastic, leather, metal, etc., that may have a first 410 and a second portions 420 to be coupled to the watch body 1 of a timekeeping device. The watchband 400 may further comprise coupling elements 402 for coupling the watch face 1 to the flexible layer 401.

The first portion 410 may comprise adjusting elements 411, e.g. perforations, in order to enable the user to adapt the watchband to the size of user wrist. The second portion 420 may comprise a buckle 440 arranged at an end 421 of the flexible layer. The buckle 440 may comprise a frame 441 and a pivoting prong 442 coupled to the frame. The pivoting prong 442 may be introduced in a specific adjusting element to adjust the watchband to the user wrist.

In order to ensure a proper counterforce when fastening the watchband 400 i.e. when pulling up the buckle, around the wrist of the user, the watchband may further comprise a solid plate 430 e.g. made of metal, which may be arranged below the flexible layer.

The watchband 400 may replace the existing fastening system of a timekeeping device thereby ensuring a tighten fixation which aids to maintain the sensor-body contact pressure and improves the signal while avoiding vasoconstriction and/or user discomfort.

Additionally, the watchband 400 may further comprise a tightening mechanism 500 that may be arranged in the frame of the buckle 440. The use of the tightening mechanism 500 enables obtaining a predefined fixation on the wrist and thereby improving the sensor-skin contact and the accuracy of measurements.

FIG. 5 shows a tightening mechanism 500 that may be arranged at a side of the buckle. The tightening mechanism 500 may comprise a ratchet gear 510 having a plurality of locking teeth arranged at the periphery, over a predetermined angular range and facing a pivoting strut 520 that may engage the teeth. The mechanism 500 may further comprise an unlocking mechanism 530, e.g. a knob, and compression spring (not shown) for holding the unlocking mechanism 530 against the pivoting strut against the ratchet gear 510.

In order to unlock the mechanism, the user may press the unlocking mechanism and release the pivoting strut.

The buckle can freely move in the opposite direction to locking thereby allowing the watch band to form a circular shape around the wrist when no measurement is taken.

In an example, the watchband 400 may comprise two tightening mechanisms 500 arranged at opposite sides of the buckle.

FIG. 6 depicts a system 600 which may comprise an illumination module 601 comprising at least two paired radiation sources according to any of the disclosed examples for illuminating user skin, a capture module 602 comprising at least one radiation detector for detecting the light exiting from the user according to any of the disclosed examples and a calculation module 603 for calculating an absorption signal. The system 600 may further comprise a feature extraction module 604 for extracting features from the absorption signal and a communication module 605 for managing the communications between the calculation and feature extraction modules. In an example, the communication module may further manage communications between the system 600 and an external communication device 610.

FIG. 7 depicts a system 700 for non-invasive measurements of biomarkers which may comprise a back case 701 according to any of the disclosed examples and wherein the back case may comprise an optoelectronic circuit board 702 according to any of the disclosed examples. The system 700 may comprise an interface module 703 and a mobile device 704 for controlling signal acquisition procedure, acquisition mode and acquisition parameters. In an example, the system 700 may further comprise a detachable device for obtaining measurements in a different body position. The detachable device may be adhered to user body e.g. in a removable manner. In an example, the detachable device may be a sticker.

FIG. 8A depicts a block diagram of a method 800 for non-invasive measurements of biomarkers which may be implemented with a back case comprising an optoelectronic circuit according to any of the disclosed examples. Firstly, the at least two paired radiation sources having different wave lengths λ1, λ2 may be provided.

In an example, the first wave length λ1 may achieve deeper penetration than the second wave length λ2 which may only achieve superficial penetration. The first wave length λ1 may be, e.g. red light, infrared (IR) or near infrared (NIR), and when comparing with second wave length λ2, it may be less sensitive to oxyhemoglobin and therefore more stable over time, and less sensitive to temperature. The second wave length λ2, e.g. green light, may capture large intensity variations as it may have greater absorption of oxyhemoglobin and deoxyhemoglobin and so the cardiac cycle may be better reflected. By using such different wave lengths a better signal to noise ratio (SNR) may therefore be achieved. In addition, the second wave length λ2 may be more sensitive to temperature and vasodilatation/constriction.

Once a back case is attached or coupled to a timekeeping device and the latter fastened e.g. around user wrist, the user skin may be illuminated, in block 801, with at least two paired radiation sources having different wave lengths. The acquired signal and the skin penetration may therefore be improved. In an example, the skin may be interchangeably illuminated by the two paired radiation sources. In an example, the skin may be firstly illuminated by an IR or NIR radiation wave length to get robust PPG signal in reflectance mode.

Then, in block 802, the radiation exiting from the skin, i.e. the reflected radiation, may be captured, e.g. by a radiation detector arranged on the optoelectronic circuit board of the back case, and an absorption signal, e.g. a PPG signal, may be generated, in block 803, by the radiation detector. Then, in block 804, the signal epochs or epochs of the PPG signal may be extracted. Signal epochs are signal segments or signal frame lengths of a predefined length which may be long enough to facilitate robust statistical information and short enough to capture non-stationary information. In an example, the length of each epoch may be of about 8 seconds to accommodate 8-10 heart cycles within.

Pulses and/or pulse morphology of the PPG signal may then, in block 805, be identified and the PPG signal may then be filtered, in block 806, to get ideal artefact free PPG signal. In an example, the filtering may comprise rejecting whole pulses.

The filtering may then be repeated for all pulses of every signal epoch until each signal epoch is within a predetermined reliability range.

Then, in block 807, filtered pulsed, i.e. non-removed pulses, may be collated thereby forming an acceptable (artefact free or ideal) derived PPG signal. The derived PPG signal (PPGd) may then be used for calculating the biomarkers (later on disclosed).

In an example, the method 800 may further comprise assessing SNR_PPGof the PPG signal to verify whether SNR_PPGis within an acceptable range of values. The SNR_PPGof the PPG signal may indicate the level of the clinically useful signal to the level of background noise. In an example, the SNR_PPGmay be the ratio of filtered PPG signal variance to the unfiltered raw signal, i.e. noise, variance.

In an example, the method may further comprise assessing the PPG signal quality which may be performed e.g. by Kurtosis, Skewness or any other suitable method. The Skewness (S_PPG) is defined as:

S
PPG

=

E
⁢

(

x
-

μ
x

)

σ
x
3

where x is the raw, i.e. unfiltered, PPG signal, μ_xand σ_xare the empirical estimates of the mean and standard deviation of x, and E is the expected value operator.

In an alternative example, Perfusion index (PI_PPG) of the PPG signal may be used to calculate PPG signal quality. The PI_PPGis the ratio of the pulsatile blood flow (AC) to the non-pulsatile or static blood in peripheral tissue (DC), and represents the represents the difference of the amount of light absorbed through the pulse of when light is transmitted through the tissue. The PI_PPGmay be calculated as:

PI
PPG

=

AC
DC

×
100

=

(

emmission
⁢

peak

-

absortion
⁢

peak

)

μ
x

×
100

Wherein μ_xis the raw unfiltered signal mean, and emission and absorption peaks are calculated on the filtered signal.

In case the SNR is out of the predefined range of values, the skin may be further illuminated with a predefined wave length radiation e.g. 916-960 nm. IR/NIR wavelength, until the maximum possible gain limits of the illumination driver is reached. The SNR_PPG, PI_PPGor S_PPGmay then be re-evaluated and, in the event that a non-working signal persists and no further gain increase may be achieved, the skin may be illuminated by alternative radiation having a wave length of e.g. 510-522 nm, corresponding to green colour wavelength. The amplification gain can be also increased in order to ensure better SNR.

In an example, temperature may also be closely monitored, since cold temperature conditions may significantly reduce PPG signal quality, whereas warm temperature conditions may improve the quality of the PPG signal e.g. up to four times. Warm temperature conditions ensure reliable measurements related to peripheral extraction and circulation.

FIG. 8B schematically illustrates a detailed implementation of the method of FIG. 8A. In block 8010 the method may be initiated when a condition is met (e.g. user selection, a predetermined time interval is reached, a time of the day etc.). In block 8020 the skin may be illuminated using e.g. IR/NIR wavelengths. In block 8040 a PPG signal may be captured and epochs may be extracted. The SNR may be assessed based on the absorption properties of the skin and the tightness of the strap around the wrist of the user. In decision box 8042, it may be determined if the SNR assessed is acceptable. If the SNR is determined to be not acceptable then in decision box 8044 it may be determined if the gain or frequency of the signal may be increased. If the gain can be increased, then in block 8060 the amplifier gain is increased and the method moves back to block 8040. If in decision box 8044 it is determined that the gain or frequency cannot be further increased, then the skin may be illuminated using a green wavelength in box 8080 and the process moves back to box 8040. Now, if after the aforementioned feedback control loop involving optimizing timing, intensity, gain or switching wavelength, the SNR is determined to be acceptable in decision box 8042. Then in box 8100 the distinct pulses' morphology may be identified to meet a predefined PPG signal template considered as ideal. By using a plurality of template matching algorithms, such as the Common Spatial Patterns (CSP) technique, template subtraction, Principal/Independent Component Analysis (PCA/ICA), simple Correlation or alternative, that may compare the actual signal with the template signal a similarity may be identified. A dynamic time warping may be used to stretch each beat to match the ideal running template. This process may be repeated for every pulse in each signal epoch until an ideal PPG derived signal (PPGd) is captured in box 8140. The process may then end in box 8150. As mentioned, the epoch is defined as the signal frame length able to accommodate about 8 to 10 heart cycles, long enough to capture robust statistical signal information, yet short enough to capture nonstationary information. Decision box 8105 may determine if pulse is acceptable and in box 8120 suboptimal pulses may be rejected.

FIG. 9 depicts a block diagram of a method 900 for extracting a feature output set which may be used to calculate biomarkers. Biomarkers may include Heart Rate (HR), Heart Rate Variability (HRV), Blood Oxygen Level (SpO₂), Cardiac Output (CO). Blood Pressure (BP), Respiratory Rate (RR), Arterial Stiffness (AS), Blood Flow Volume (BFV), Temperature and Activity tracking, etc.

Firstly, a PPGd signal e.g. the signal calculated from method 800, comprising a plurality of epochs having pulses may be provided, in block 901. The pulses of the PPGd signal may, in block 908, be separated.

In addition, the falling edge ePW(i) of each pulse (i denotes each pulse) may be extracted, in block 910, and the beginning and end tails of each ePW(i) may be discarded, in block 912, to eliminate edge discrepancies. To that end an empirically defined threshold value, e.g. equal to a length segment of 0.2*HR seconds, may be used.

The first derivative of each ePW(i) may then be calculated, in block 914, and filtered, in block 916, e.g. by low pass filter, in order to smooth out high frequency content. In an example, a moving average filtering with windows size of 31 samples may be used for filtering the first derivative.

The filtered first derivative may then be used to find the zero crossings (zCx), in block 920, e.g. using a step by step zero-crossing adaptive/non-adaptive technique.

In the event the zero crossings are found, their location may be determined and a respective ePW(i) dicrotic notch/incisura position may be found and assigned, in block 924, and such value stored in a memory device. The algorithm may be moved to the subsequent pulse, in block 926. The process may be repeated for all the pulses of the PPGd signal. We use the terms dicrotic notch and incisura to refer descriptively to the two possible positions of the notch. In the first case, (if the position is matched with an upstroke) we name it to be a “dicrotic notch”, whether in the latter case, (if the position matches an inflection of the waveform) we will refer to it as an incisura.

On the contrary, in the event the zero crossings are not found, the peaks of first derivative of ePW(i) may be found, in block 918, and may be used to analyse each peak position in a descending order thereby to locating the peak corresponding to the incisura peak, in block 922. In an example, the selection of the peak may be based on position criteria with respect to the inter-beat time.

The output of block 922, may then be evaluated and used to determine the incisura position if it fulfils certain inclusion criteria, as follows: Disjoint pulses (908) are identified to be: upstrokes (upslope), or downstrokes (downslope) and inflections—based on the bending pattern of the pulse under study. The “dicrotic notch and/or incisura peak” is identified if one or more of the following criteria are met: a downstroke followed by a small upstroke and/or a downstroke followed by an inflection and an upstroke and/or a downstroke followed by an inflection and another downstroke. If more than one of these is identified, we select the one with the higher slope (negative/positive) curvature in the respected segment of question.

The algorithm may be moved to the subsequent pulse, in block 926.

In parallel to the separation of the pulses of the PPGd signal, the emission and absorption peaks of the PPGd signal may similarly be detected, in blocks 904 and 906, respectively, e.g. by using an algorithm based on geometric definition of signal trends and the statistical definition of peaks and valleys.

Additionally, the method 900 may further comprise calculating the second derivative of the PPGd signal, in block 914a, after separating the pulses of the PPGd signal. Then, the maximum and minimum peaks of the second derivative of the PPGd signal may be found, in block 916a, and the ratio of the maximum peak to the minimum peak may be calculated.

After processing all PPGd's pulses, all epoch's dicrotic notch/incisura may be retrieved, in block 928, and a feature output set may be calculated, in block 930, by adding the emission and absorption peaks latency and amplitude, the peak and valley latency of the PPGd's second derivative, as well as the computed epoch's HR (902) that may be calculated using e.g. dominant frequency extraction techniques or time domain detection techniques (i.e. autocorrelation).

The output feature set (f1-f8) may comprise: Pulse Transit Time (PTT), defined by the difference in latency within each pulse between the dicrotic notch and the absorption peak of the same pulse (f1), Amplitude/Latency of emission peak (f2), Amplitude/Latency of absorption peak (f3), Amplitude/Latency of dicrotic notch/incisura peak (f4), Heart Rate (HR) (f5), Time span between Maximum peak of first derivative and dicrotic notch/incisura peak of PPGd pulse (f6), the second derivative time span between maximum and minimum peak (f7) and PPGdIR ratio of PPGd epoch emission and absorption peak amplitude (f8).

In an example, the output features may then be used to compute the biomarkers for instance by method 1000 (later on disclosed).

FIG. 10 depicts a block diagram of method 1000 for calculating biomarkers, which may be subsequent to method 900 of FIG. 9. Firstly, the features may be extracted from the PPGd signal, in block 1001, and, in block 1002, the extracted features may be compared with previously stored features. In the event of statistically significant non-matching features, such features may be rejected.

The biomarkers may then be calculated, in block 1004, and calculated biomarkers may be compared, in block 1006, with values of previous measures e.g. of previous day, week, month, etc., in search or variability patterns i.e. capture at same time and similar activity level.

FIG. 11 depicts a method 1100 for calculating Blood Pressure (BP) which may accurately estimate the BP using an output feature set f1-f8 calculated according to any of the disclosed examples. Firstly, an ideal noise free environment may be secured, in block 1101.

To get a noise free environment and improve the measurements, the user may be in a resting position, e.g. sitting upright on a chair. The user may use a system for non-invasive measurements according to any of the disclosed examples, i.e. comprising at least a back case, an interface module and a mobile device, attached thereto. In an example, the device may be levelled with the user heart and the feet flat on the floor.

In order to avoid obstructing blood flow, the excess of clothing or anything that could possibly block unobstructed blood flow in the arm may be removed. As a precautionary measure, reading or talking while taking measurements should be avoided for ensuring proper measurements. Before taking the first measurement, the user may tighten the watchband, in block