Method and detector device for determining concentration of substance

Abstract

The invention relates to a method for determining the concentration of a substance in a sample comprising a liquid comprising particles, in particular glucose in blood, in which the refractive index of the liquid depends on the concentration of the substance dissolved therein and the particle density in the liquid is substantially constant. The method includes obtaining a PPG first signal during a first time period and estimating a set of complex coefficients from the first obtained signal. The phase difference is determined from these coefficients or a subset thereof and then used to derive a substance concentration or a change in substance concentration.

Description

TECHNICAL FIELD

[0001] This application claims priority to German application DE 10 2023 130 811.4, having a date of November 7, 2023, the disclosure of which is incorporated herein by reference in its entirety. The invention relates to a method for determining a concentration of a substance in a sample containing particles in a liquid, in particular glucose in blood, wherein the refractive index of the liquid depends on the concentration of the substance dissolved therein. The invention further relates to a detector device. BACKGROUND

[0002] Current standards for blood glucose measurement typically use an invasive technique, wherein a small amount of blood is extracted and subsequently analyzed electrochemically using a handheld device. This method is not suitable for continuous monitoring, since for each measurement a finger has to be pricked to obtain a fresh blood sample.

[0003] A more recently developed technique uses a button on the skin and employs a small needle-like sensor to miss the interstitial fluid in the part of the subcutaneous fatty tissue. However, the needle penetrates permanently into the skin. This method is at risk of infection. Furthermore, it can have to be removed in certain situations, for example during sports activities, swimming, etc.

[0004] In addition to these invasive methods, there are non-invasive methods based on optical IR measurements or Raman spectroscopy. In the first case, the selection of suitable emitters and detectors poses difficulties, while the method based on Raman spectroscopy is challenging due to the very poor signal-to-noise ratio. Recently, optical measurements using parts of the visible spectrum have been proposed, wherein the backscattered light is analyzed. However, depending on the skin type, the environment and other parameters, the signal-to-noise ratio can vary and significantly reduce the overall signal quality.

[0005] In this regard, there is a need for a method that can detect a substance in a liquid in a simpler manner and allows for continuous measurements. SUMMARY

[0006] This object and other objects are solved by the subject matter of the independent claims. Features of the proposed principle and further aspects are outlined in the dependent claims.

[0007] The inventors have recognized that the PPG waveform and its morphology can be used to determine whether the glucose concentration in blood is rising or falling. From the observations, the inventors can confirm that there is a morphological evolution during an oral glucose tolerance test (OGTT). More particularly, another aspect is observed that is able to describe the changing glucose concentration. This is given by the relative phase between the complex harmonics of the PPG waveform in frequency space.

[0008] The inventors now propose to estimate the relative phase between two harmonics or between the fundamental peak and one of the harmonics in the frequency space of the PPG signal. The phase shift tracks the morphological changes of the PPG wave, which in turn are associated with rising blood glucose concentrations measured with a reference blood glucose meter. It turns out that the relative phase of the signal, which is affected by noise, shows low variability. Furthermore, the optical measurement seems to lead the reference measurement by about 5 to 10 minutes, even though the reference measurement is taken using a finger prick method in case of blood glucose measurement.

[0009] The method proposed by the examiner is not limited to blood glucose measurement and can be applied to any flow liquid containing particles with volume or pressure modulation. This opens up possibilities for a variety of different applications, such as in the treatment of sewage, optical measurements in biological laboratories, determination of SPO2 or other substances in blood or liquids, etc. The improved robustness against noise and other disturbances also opens a window of opportunity for other diagnostics. The method can be combined with data and results of other optical parameters such as noise, PI and distance dependent signals to further improve the resolution and accuracy of such measurements.

[0010] In some aspects, a method for determining a concentration of a substance in a sample of a liquid containing particles, in particular glucose in blood, is proposed, wherein the refractive index of the liquid depends on the concentration of the substance dissolved therein and the density of the particles in the liquid is substantially constant. In the proposed method, the liquid is modulated in its volume or in its density. If the liquid is blood, the concentration of the substance in the blood is to be measured, the heartbeat provides the required modulation.

[0011] For determining the concentration of the substance, a first signal is obtained during a first time period. The acquisition rate of the first obtained signal is set to be at least two times larger than the periodicity or frequency of the modulation of the liquid to satisfy the Nyquist criterion. For example, the acquisition rate can be 10 times, 20 times or even 100 times larger. Typically, the acquisition rate is at least 40 times larger than the modulation frequency. In case of blood with a heartbeat of about 1 Hz to 2 Hz, the acquisition rate can be at least 10 Hz and more particularly between 20 Hz and 200 Hz.

[0012] Likewise, the first time period is adjusted such that a plurality of periodicities of the volume or pressure modulation of the liquid fits into the first time period. For example, the first time period can be between 3 times and 30 times larger than the modulation time period of the liquid. In case of a heartbeat frequency of 1 Hz, the acquisition time can be between 5 seconds and 20 seconds.

[0013] The method ensures that during the first time period and the acquisition of the first signal, the noise is captured with sufficient resolution and with sufficient number of signal peaks.

[0014] In a second step, at least some complex coefficients are extracted from the AC part in at least a part of the spectrum of the first obtained signal, wherein the complex coefficients cover at least one of the fundamental peak value and the peaks at positions in the frequency space that are integer multiples of the fundamental frequency and the defined phase between the two. This can correspond to the harmonic peaks. The complex coefficients are typically at least partially derived from a complex Fourier transform applied to the obtained signal in the time domain to transform it into the frequency domain. The spectrum in the frequency domain can be further processed or only parts of the spectrum are used to extract the complex coefficients therefrom.

[0015] With respect to the term "part", it is to be understood that the overall signal spectrum is itself band-limited due to real-world signal processing. It contains a noise part, the actual signal refers to the AC part and the DC part. For the proposed method, the AC part of the heartbeat signal, i.e. the fundamental peak value and the harmonics are relevant. Thus, complex coefficients can be associated with the fundamental peak value and one or more harmonic peak values, but it can not be necessary to have all respective complex coefficients. Thus, the "part" of the spectrum can comprise the entire spectrum up to a certain frequency limit, the AC part thereof or the DC part or the noise part, the latter two being the focus of interest in different applications.

[0016] It has been observed that certain subsets of complex coefficients are more suitable than others for determining a substance concentration or a change thereof. A variety of methods can be employed for such extraction, some of which are presented herein.

[0017] In a final step, a substance concentration is estimated and determined based on the phase difference between different peaks using a subset of complex coefficients extracted from at least a part of the spectrum and an optional reference value.

[0018] Alternatively, a change of a substance concentration can be determined based on the phase difference between different peaks using a subset of complex coefficients extracted from at least a part of the spectrum and an optional reference value. In this regard, the reference value can be zero or any other benchmark value. For example, the reference value can correspond to a ground or normal level of the substance concentration in the liquid.

[0019] The proposed method makes use of the fact that in certain cases a substance concentration can be determined by deviating from a shot noise limited system, as the substance or a change thereof introduces an additional term of systematic noise. This method can be beneficial in highly noisy systems or when the signal-to-noise ratio is significantly deteriorated. It has been observed in this regard that the phase difference extracted from the noise data shows less variability compared to other signals, such as the perfusion index that is also associated with the glucose concentration. Thus, the proposed method based on the use of the phase difference appears to be more robust than other measurement methods.

[0020] In some aspects, the reference value is based on the processed second signal. Thus, the proposed method comprises the step of obtaining a second signal during a second time period at an acquisition rate that is at least two times larger than the periodicity of the volume or pressure modulation of the liquid. The second time period can be after or before the first time period, with some time interval between the measurements. Further measurements can be performed to obtain a plurality of such signals. Suitably, the first and second time periods are identical to reduce possible residual effects, but subsequent method steps can compensate for different periodic measurements, e.g. by normalization.

[0021] A set of complex coefficients is extracted from at least a portion of the spectrum of the second obtained signal. The portion of the spectrum can be the same for the first and second obtained signals. In some aspects, the reference value is then based on the complex coefficients extracted from the spectral portion of the second obtained signal. This allows to obtain a change in the concentration of the substance between two or more measurements. In some cases where the concentration of the substance changes over time, the rate of change of the concentration of the substance and other parameters can be obtained according to the proposed method by evaluating a plurality of obtained signals. Also, in some aspects, the reference value is given by a predetermined set of complex coefficients or by values derived from these complex coefficients.

[0022] In some aspects, the step of extracting at least some complex coefficients from the spectrum comprises the step of identifying an AC portion of interest in the spectrum of the obtained signal, which corresponds to the fundamental frequency of the volume or pressure modulation of the liquid and some of its harmonics. Typically, the harmonics occur at multiples of the fundamental frequency. In the case of a heart rate, the fundamental frequency corresponds to the heart beat but is not fixed as it changes with the heart beat speeding up or slowing down. After identifying the AC portion of interest, a set of complex coefficients is computed from the fundamental frequency and at least one of the harmonics using the identified portion. In some aspects, the set of complex coefficients can be computed directly using the overall spectrum and all harmonics (band-limited).

[0023] In this respect, the values of the complex coefficients can vary depending on the spectrum and the number of peaks used for their computation. In some cases, the coefficients are computed individually for each identified AC portion, in some other aspects, the complex coefficients are computed in combination for the identified peaks of the fundamental frequency and the harmonics. Thus, a subset of complex coefficients is associated with each peak.

[0024] As mentioned above, the AC part determined by this method can then be separated from the spectrum. Not all harmonics are visible or have an effect on the spectrum due to noise levels. Therefore, in most cases, it can be sufficient to consider only the fundamental frequency and its harmonics up to the ninth, and especially up to the seventh, and especially up to the fifth, and especially up to the fourth. However, the maximum order is not limited. If the signal is clean enough, e.g. there is enough light on the finger, etc., or in non-VSM / non-biomedical applications, even higher orders can be present and extracted. Considering even higher order harmonic peaks can have particular interest for certain applications, i.e. for distinguishing glucose concentration changes from other substance changes, or in some cases, considering other substance concentrations, pressure, subject age, gender and condition.

[0025] It is noted in this respect that the original spectrum contains not only the AC part and the DC part, but also noise, which is present below the fundamental peak at the fundamental frequency and at the harmonic locations. Therefore, in some aspects, it can be appropriate to reduce or remove the noise at the locations below the peaks of the fundamental and harmonics before calculating the complex coefficients.

[0026] To this end, the step of extracting a set of complex coefficients as discussed above comprises the step of estimating the noise part within the spectrum around the fundamental frequency of the volume or pressure modulation of the liquid and its harmonics based on the values of adjacent points in the spectrum. More especially, the noise around this frequency with a determined width is identified, estimated and then removed (further explained below) to obtain a clean AC part of the spectrum. It should be noted that noise outside the peaks of the fundamental and its harmonics can also be removed.

[0027] The method uses the fundamental peak width or a given mask to define the width of the noise part to be estimated. The generation and application of such masks will be explained in more detail below. The estimation itself uses the noise of directly adjacent respective peaks and assumes that the noise level below the peaks is approximately constant over the width of the base of the peaks. As an alternative method, the noise part below the masked part of the spectrum can be estimated based on the points of the spectrum at the borders of the masked part. The estimated noise part is then subtracted from the overall spectrum, essentially leaving a clean AC part, which is purely the result of the signal modulation due to the beating of the heart.

[0028] Some aspects relate to the step of identifying the AC part of interest in the frequency spectrum of the obtained first or second signal. These can include a number of tasks and pre-processing steps, which can be used individually or in combination with other steps, to enable a distinction of the peak signal from the noise part. In some aspects, a smoothing function is applied to the frequency spectrum of the first and second obtained signal. Smoothing reduces noise; it should therefore not be too strong, but also help to reduce outliers. The smoothing function can be applied before obtaining the frequency spectrum to flatten out the potentially jagged structure of the harmonics.

[0029] In some aspects, at least the fundamental peak is enhanced, in particular by taking the absolute value of the spectrum to the power M, where M is an integer larger than 2. This is particularly useful when the data is noisy and the peak in the spectrum only slightly stands out above the adjacent noise. By applying this function, the peak is significantly enhanced with respect to the noise, which also simplifies subsequent processing steps.

[0030] In some aspects, a mask is determined, which has a width based on a predetermined amplitude level of at least the fundamental peak. For this, a reference level can be set to define the mask boundaries, for example at 10% of the peak level, or at 5% of the peak level. Other levels are also possible, for example 3%, 2% or even 1% and below. The reference level for defining the noise can depend on the overall signal level (i.e. a fraction thereof), and also on the enhancement level.

[0031] In some aspects, a dynamic peak width is determined, for example by f max -f min is determined, and the frequencies are given by the respective reference level at the rising and falling edge of the respective peak. In some aspects, a static offset can be added to the dynamic width, but the advantage is that this offset only needs to be determined once for all measurements.

[0032] In some aspects, the center of gravity of at least the fundamental peak is determined, optionally by using a width based on a predetermined amplitude level of at least the fundamental peak. The center of gravity can be determined by summing the product of the frequency and the FFT amplitude over the width of the previously determined fundamental. The result is then divided by the sum of the amplitudes. It should be noted that the center of gravity of the fundamental peak does not necessarily correspond to the maximum: this is only the case for symmetric peaks, which cannot be guaranteed in general. The center of gravity of the fundamental also provides information for identifying the correct frequency of the harmonics.

[0033] The mask with the determined width is applied on a predetermined number of possible peaks, and more in particular, on the respective center of gravity of the fundamental, and then repeated at the frequencies of the harmonics. The peaks of the harmonics are calculated from the fundamental within the spectrum, which has been determined as described above. The mask thus covers a predetermined number of peaks corresponding to the fundamental and its harmonics of the volume or pressure modulation of the liquid, thereby identifying the AC part.

[0034] As mentioned above, the AC part determined by this method can then be separated from the spectrum. Alternatively, the noise can be removed from the spectrum. In both aspects, a substantially pure spectrum is retained which only contains the fundamental peak and its harmonics. Not all harmonics are visible or contribute to the spectrum due to the noise level. Therefore, in most cases it can be sufficient to only consider the fundamental and its harmonics up to the seventh, and especially up to the fifth, and especially up to the fourth.

[0035] In some aspects, the first signal and the second signal are pre-processed. For example, the first signal and the second signal can be filtered to remove transients, especially with a frequency below 2 Hz, and especially below 1 Hz, and especially below 0.5 Hz. Heartbeat rates can be rather rare at 0.5 Hz or below. Breathing usually occurs in the range of 0.2 Hz (5 seconds per breathing cycle / 12 breaths per minute) and falls into the < 0.5 Hz range. These transients can be generated by breathing, for example in case of measuring blood. In some cases, the pre-processing is performed by applying a moving average filter with a window width of about one heartbeat, which captures slow transients. Then, the transients are removed and the DC part is re-added to generate a PPG signal which is rather stationary but still retains the AC modulation due to the heartbeat.

[0036] In some aspects, background subtraction has been performed on the measurement signals to remove dark counts and environmental effects. The first signal and the second signal can be converted into digital signals before further processing. The above-mentioned pre-processing can be performed in the analog domain or in the digital domain (e.g. low-pass filtering and background subtraction), while the Fourier transform for extracting the noise part is usually done in the digital domain, but can also be implemented, for example, in a system-on-chip (SoC) in the analog domain. In some aspects, the first obtained signal and the second obtained signal are processed using a fast Fourier transform.

[0037] In some further aspects, the phase difference is determined by first calculating complex numbers. The respective complex numbers correspond to a sum of complex Fourier coefficients associated with one of the fundamental peak and a harmonic. For example, a complex number can be given by the sum of complex Fourier coefficients associated with the fundamental peak, while another complex number can be given by the sum of complex Fourier coefficients associated with the first harmonic peak. The overall complex number of each measurement corresponds to the identified AC part therein.

[0038] At least two complex numbers are determined based on respective sums of complex Fourier coefficients. Then a first phase is based on a first complex number, and a second phase is based on a second complex number. In this respect, the first phase and the second phase can be determined using a geometric function, especially an arctangent with the imaginary part and the real part of the complex number as input.

[0039] In some further aspects, the complex number is based on the respective AC coefficient at the center of gravity frequency corresponding to the predetermined harmonic, as described above. However, it is also possible to use the sum of the complex Fourier coefficients within a mask covering the respective harmonic or fundamental peak value to determine the complex number. Using a mask and the signal under the mask can simplify the necessary calculations, as the mask has already been used to determine the AC part and its center of gravity.

[0040] Some aspects relate to a detector device. The detector device can be implemented in a single housing, such as a watch or a medical instrument, but can also comprise distributed components. In some aspects, a detector device is proposed for determining a concentration of a substance in a sample comprising a liquid containing particles, in particular glucose in blood, wherein the refractive index of the liquid depends on the concentration of the substance dissolved therein and the density of the particles in the liquid is substantially constant. The detector device comprises at least one light source and at least one detection component, wherein the at least one detection component is optically separated from the at least one light source. The at least one light source is configured to emit light through an exit window onto the sample containing the liquid with the substance.

[0041] The detection component is configured to detect a light component corresponding to the emitted light scattered through the tissue. The detector device further comprises a control circuit coupled to the at least one light source and the at least one detection component. The detector device is configured to perform the method steps proposed above.

[0042] In some further aspects, the detector device can comprise an evaluation circuit for extracting a set of complex coefficients from the first obtained signal, in particular in the frequency domain, and obtaining a phase difference between two peak values of the frequency spectrum therefrom. The evaluation circuit is further configured to determine the concentration of the substance from the phase difference or to determine a change in the concentration of the substance from the phase difference.

[0043] Optionally, the evaluation circuit can be configured to utilize a reference phase, which in turn can be derived from the phase difference according to the method proposed above.

[0044] The evaluation circuit can be implemented together with other elements in a single housing, but can also be arranged separately and communicate using wired or wireless interfaces. For example, the evaluation circuit can be a computer program executed on a mobile device, a computer or a medical device.

[0045] In some aspects, the control circuit is configured to control the at least one light source to emit a light signal in a first time period at a plurality of different consecutive times and to obtain a signal from the at least one detection component. The control circuit can be controlled and triggered by the evaluation circuit.

[0046] Some aspects relate to the implementation of a detector device. The detector device can comprise a plurality of photodetectors arranged in a ring shape or a quadrangle shape and optionally especially arranged at different distances around at least one light source especially centrally. In other aspects, the at least one light source comprises a plurality of optoelectronic devices arranged with different distances to at least one detection component; and optionally comprises a ring shape or a quadrangle shape, optionally arranged around at least one detection component especially centrally.

[0047] It can be appropriate if the at least one light source is configured to emit light of different wavelengths. This can compensate for different skin types and other parameters. In other aspects, the at least one detection component comprises a filter comprising a low transmissivity in a spectrum different from the spectrum of the light emitted by the at least one light source.

[0048] Although the optoelectronic devices presented herein are only explained with respect to their functionality with respect to the determination of a concentration of a substance, it should be noted that various implementations are possible. In this respect, the evaluation unit need not be implemented within the housing itself containing the emitter and the detector, but can be located separately therefrom. In some aspects, the evaluation unit is implemented in a separate device separate from the housing itself. The communication between the emitter and the detector and the evaluation unit as described above is facilitated for example by wireless communication. This would allow for example to implement a master-slave configuration, where the evaluation unit requests measurements to be taken periodically. Furthermore, the evaluation unit can be implemented mainly in software, for example as an application executed on a mobile device, while the rest of the optoelectronic device is implemented in a separate housing.

[0049] In some aspects, the housing (or more generally the optoelectronic device) is implemented as a ring, an earplug, a watch or any other wearable, which can in some aspects adapt to the user's daily environment and can be carried continuously. These rings, earplugs, watches or any other wearables are in communication connection with the evaluation unit or a mobile device or any other device implementing the evaluation unit. The rings, earplugs, watches or any other wearables can cover a larger portion of the user's skin, for example encircle a finger, clamp from both sides on an ear, etc. They can contain a plurality of emitters and detectors located at various positions, such that they can take measurements not only at one point, but also at multiple points at once or subsequently. As a result thereof, skin irregularities or other problems can be overcome and the overall measurement quality is improved.

[0050] Besides devices and hand-held devices, other applications are possible. For example, the optoelectronic devices can be implemented in medical devices or laboratory equipment, for example for testing and measurement purposes. These devices can also be stationary or mobile.

[0051] Some further aspects involve mobile displays in which detectors are directly implemented. In such applications, display LEDs (e.g., for red and green, and even blue) can be used as emitters according to the proposed principles. A finger is placed directly on the display surface, and then illuminated by the display to obtain a first signal and / or a second signal. Similarly, the proposed principles can be implemented in VR or AR glasses and devices. Another application involves safety concerns, such as during certain labor-intensive work and while driving a vehicle. Such optoelectronic devices can be implemented in cars, for example on a steering wheel, according to the proposed principles, to obtain a perfusion index and thus a noise spectrum during driving. This makes it possible, for example, to warn the driver of potential health threats while driving. Attached Figure Description

[0052] In contrast to the various embodiments and examples described in detail with reference to the accompanying drawings, other aspects and embodiments based on the proposed principles will become apparent, as illustrated in the drawings.

[0053] Figure 1 A detector device based on some aspects of the proposed principles is shown; Figure 2 It shows that it can be done Figure 1 Exemplary PPG measurements obtained by the detector device; Figure 3 It shows a reference curve illustrating the change of glucose concentration over time based on some aspects of the proposed principle, and the corresponding phase difference measurement obtained using light in the green spectrum; Figure 4 Two PPG measurements at points #1 and #33 are shown over a time interval of approximately 10 seconds, based on some aspects of the proposed principles. Figure 5 Detailed views of two PPG measurements are shown to illustrate some aspects of the traces and their characteristics; Figure 6 The spectra of two different data points obtained at low and high glucose concentrations are shown. Figure 7 A detailed view of the enhanced spectrum and fundamental frequency of data point #1 is shown to illustrate some aspects of the proposed principle; Figure 8 An exemplary mask generated from information obtained from the fundamental wave shown in the previous figure is illustrated, based on some aspects of the proposed principles; Figure 9A and Figure 9B The processed AC portion of the spectrum of data points #1 and #3, including the determined mask, is shown according to some aspects of the proposed principle; Figure 10Phase difference curves according to some aspects of the proposed principles are shown compared to reference measurements; Figure 11A and Figure 11B Unprocessed and processed Perfusion Indices are shown compared to Figure 3 embodiments of the proposed principles to illustrate some benefits of the proposed principles; Figure 12 Measurements of the Clarke error grid analysis for Figure 3 are shown; Figure 13A and Figure 13B Phase difference measurements for different wavelengths according to some aspects of the proposed principles are shown; Figure 14A Phase difference measurements using different harmonics in the infrared part of the spectrum according to some aspects of the proposed principles are shown; Figure 14B Corresponding measurements of the PI for Figure 14A are shown; Figure 15 Some data from multiple different parallel phase difference measurements taken at different wavelengths and different distances or angles of incidence are shown; Figure 16 Average traces of individual measurements of Figure 15 according to some aspects of the proposed principles are shown together with a reference trace; Figure 17A and Figure 17B Measurements according to some aspects of the proposed principles and their corresponding Clarke error grid analysis are shown, which show some characteristics that can be obtained by this method; Figure 18A and Figure 18B Yet another measurement according to some aspects of the proposed principles and its corresponding Clarke error grid analysis are shown, which show some additional characteristics that can be obtained by this method. DETAILED DESCRIPTION

[0054] The following embodiments and examples disclose various aspects of the proposed principles and combinations thereof. The embodiments and examples are not always to scale. Likewise, different elements can be shown as exaggerated or reduced in size to emphasize various aspects. It goes without saying that the various aspects of the embodiments and examples shown in the figures can be combined with each other without contradicting the principles of the invention. Some aspects show regular structures or forms. It should be noted that in practice slight differences and deviations from the ideal forms can occur, but this does not contradict the idea of the invention.

[0055] Furthermore, the various illustrations and aspects are not necessarily drawn to scale and the dimensions of the various elements can not be necessarily be substantially correct. Some aspects are highlighted by exaggerating them. However, terms such as "above", "over", "below", "under", "larger", "smaller" and the like are correctly indicated in the figures with respect to the relationship of the elements. Thus, such relationships between elements can be inferred from the figures.

[0056] Figure 1 A detector arrangement according to some aspects of the proposed principles is shown. In this embodiment, the detector arrangement facilitates glucose measurements in blood vessels of a human or animal body. However, it should be understood that the present invention and the proposed method are not limited to such measurements. Instead, for example, an artificial tube can be employed instead of human tissue to measure the concentration of a substance in a liquid flowing through the tube. The various elements of such a detector arrangement are similar to the arrangement shown in Figure 1 This enables the proposed method to be used for a variety of measurements in different applications. Thus, the method and detector proposed herein are able to determine the concentration of a substance in a particle-containing liquid.

[0057] The arrangement is used for a variety of measurements, including for example PPG measurements or SPO2 measurements. In this regard, as long as the detector arrangement is suitable to provide a PPG signal comprising a noise portion, the proposed method as explained herein can be implemented in existing hardware by software. In the present embodiment, the detector arrangement 1 comprises a housing 11 comprising at least one light source 12 configured to emit light pulses of at least one wavelength in the visible or IR spectrum.

[0058] The light source 12 can also comprise a plurality of light sources (not shown here) configured to emit light of different wavelengths or located at different positions. This enables to compensate for skin irregularities and also provides an adjustment of the signal-to-noise ratio, since the penetration of light to deeper layers is strongly dependent on the wavelength. For example, typical wavelengths suitable for PPG measurements as well as for measurements for determining glucose concentrations include light in the green and red visible spectrum as well as in the near infrared spectrum.

[0059] The housing 11 also comprises a detector arrangement 13 arranged separately from the light source 12. A light locking element 14 is arranged between the light source 12 and the photodetector. In this example, the photodetector 13 comprises a plurality of detector regions 13.1, 13.2, etc. The detector regions are located at different distances from the at least one light source. The various distances between the light source 12 and the detector regions 13.1, 13.2 can compensate for skin irregularities and also provide distance-dependent signals. These signals can also be used to determine the concentration of a substance, since the total amount of scattering depends on the distance.

[0060] In the present application, the systematic noise is increased by the scattering of the emitted light within the user's skin, a plurality of detector areas 13.1, 13.2 at different distances from the light source 12 enables to determine the optimal distance, where the AC part of the signal is optimized.

[0061] Finally, the detector device 1 also comprises an evaluation and control device 10. The evaluation device 10 can be placed within the housing 11 of the device or located separately. For example, the housing 11 can be a part of a smart watch or a wearable device that communicates with another device via a wireless interface. The other device contains the evaluation and control device 10. The evaluation and control device 10 is implemented as hardware, software or a combination of both.

[0062] The evaluation and control device 10 controls the measurement and can be regarded as an integrated data acquisition system that can provide control signals to a current driver for the at least one light source 12, to an analog-to-digital converter for sampling the current from the detector device 13 and areas 13.1, 13.2 and to a clock for triggering and timing the measurement.

[0063] During the measurement, the user places his finger with the tissue 30 on the surface 20, which is a part of the glass interface 21 or another transparent interface 21. The interface 21 is transparent for the light emitted by the light source, but can be non-transparent or comprise at least a reduced transparency for another wavelength. This reduces the dark current at the photodetector 13 and its areas 13.1, 13.2.

[0064] The light emitted by the light source 11 propagates through the tissue 30 along various light paths 32 in the tissue 30, where interactions with various blood carrying and non-blood carrying layers occur. Due to the resulting periodic blood flow, the heartbeat modulates the blood fraction within the measurement volume and thus encodes a modulation of the light flux re-emitted by the tissue and detectable.

[0065] The measurement modulation corresponding to the heartbeat contains various information that can be used for determination and is generally referred to as PPG measurement. For the purpose of this application, three components will be explained in more detail.

[0066] Figure 2 Such a PPG signal is shown. The PPG signal comprises a signal level or DC component (about 5.2 x 10 4 counts in the shown example), a main modulation AC of the signal due to the heartbeat (which is usually the object of direct interest in various applications) and (3) a noise component (most easily identified by the pronounced point-to-point variability in the shown example). The heartbeat is given by the maximum in the signal and lies in the range of about 60 / min (1 Hz). The perfusion index or PI is defined by the ratio of AC to DC, i.e. about 0.1 x 10 4The amplitude of the AC part of the count is divided by the DC part. In the given example, PI is approximately 1.9.

[0067] The PPG measurements and the measurements of the proposed method are typically taken over multiple heartbeats, for example within 5 to 20 seconds, and the acquisition rate is substantially larger than the heart rate. Since the proposed method requires in-depth knowledge of the noise imposed on the actual signal, it is suitable to take PPG measurements at a rate at least 20 times higher than the actual heartbeats. Since the heartbeats themselves can vary between approximately 60 beats / minute to approximately 180 beats / minute, it is suitable to use a data acquisition rate between 50 Hz and 200 Hz. In Figure 2 In the depicted example, an acquisition rate of 100 Hz is used, and each trace has 2048 data points.

[0068] In the following, a method is shown that first extracts the AC part from the overall spectrum and subsequently computes a set of complex coefficients from the spectrum. The set of coefficients is used to determine the phase difference between two harmonics or between the fundamental peak and one of the harmonics. It was found that for a given PPG measurement, the phase difference between two harmonics or between the fundamental peak and one of the harmonics, for example the second harmonic, is not constant but related to the glucose concentration. Thus, the absolute level of the substance concentration in the liquid can be determined from the phase difference or at least the change between two or more consecutive measurements. More in particular, information about the substance concentration is present in the DC component, the AC component and the noise component of the measurement signal. The DC component and the noise component are handled by different applications filed by the same applicant, but the focus is on the extraction of the AC part and its subsequent processing. While there are several different methods available to extract the noise from the signal in the time domain and in the frequency domain, the proposed method applies a method in the frequency domain.

[0069] Figure 3 A typical oral glucose tolerance test is shown, where the reference data, referred to as reference, is taken from the Abbott Freestyle Libre 3 system. During the test it constitutes the reference measurements, referred to as BG and BG offset. Food intake occurs at about 09:45, indicated by the vertical bar and the absence of data points. Shortly after, the optical measurements (in the green visible spectrum) and the reference measurements are performed at a frequency of about once per minute. The optical measurements GREEN are taken for 10 seconds, a time period long enough to obtain a proper PPG measurement and short enough that the change in glucose concentration during a single measurement is not too large. They are subsequently processed according to the proposed principles and the phase difference between the fundamental peak and the first harmonic is computed and visualized in the graph. The reference trace BG offset is the time-offset reference curve, details are given below.

[0070] After ingestion of a sugar-containing meal, the glucose concentration rises over time, reaching a peak at about 40 minutes after the meal and decaying over the next 40 minutes. Figure 3 The phase difference between the fundamental peak and the first harmonic for green light but without additional averaging is shown in Figure 6. The phase difference is significantly responsive to blood glucose concentration and will be explained in more detail below. This is further supported by the Clarke error grid analysis corresponding to the data of Figure 5, shown in Figure 6. The Clarke error grid analysis (EGA) was developed to quantify the clinical accuracy of a patient's estimate of their current blood glucose value compared to the blood glucose value obtained in their glucose meter. It is then used to quantify the clinical accuracy of the blood glucose estimate generated by the glucose meter compared to a reference value. The EGA is considered a standard for determining the accuracy of a glucose meter. Figure 3

[0071] For the purpose of extracting the noise, in the exemplary embodiment the PPG traces from indication points #1 and #33 will be used. Of course, the proposed method can be performed for every data point, i.e. every PPG measurement. For simplicity, only two points are used to explain the proposed method. The exemplary points correspond to a low and a high blood glucose concentration, respectively.

[0072] Figure 4 The actual PPG traces of data points #1 and #33 obtained by detecting light scattered by human tissue are shown. The measurement signals are pre-processed by using a low-pass filter to remove slow transients, e.g. caused by respiration. The transient removal is suitable to suppress end-point discontinuities, which would typically increase the 1 / f component in the frequency domain. Likewise, the dark current is at least partially removed. From the two traces it is visible that an increasing blood glucose concentration leads to an increase of the AC part and a decrease of the DC part. Thus, point #1 with a low glucose concentration comprises a higher DC component of about 2.6 x 10 5 and a lower AC part, which is also rather noisy. The trace of point #33 comprises a lower DC value of about 1.2 x 10 5 and a higher AC part. The noise can be similar to other noise, at least from a naked eye point of view.

[0073] Figure 5 A detailed overlay view of a portion of the PPG signals of the two traces corresponding to a single heartbeat is shown to examine the respective waveforms. The Y-axes comprise different scales to distinguish the different amplitudes of the traces of point #1 and point #33. The scale between point #1 and point #33 is about 2 between the minimum and maximum values of the respective traces.

[0074] Furthermore, during the heartbeat, there is a persistent bulge in the trace of point #33 close to the maximum value and a corresponding indentation close to the minimum value, Figure 5 ​The two circles in the middle emphasize this. The trace at point #1 (i.e. with lower glucose concentration) comprises a shape more similar to a trapezoid shape close to the maximum, while the trace at point #33 is more triangular. On the other hand, the notch at the minimum is clearly visible in the trace at point #33, i.e. high glucose concentration. The bulge clearly visible in the trace at point #33 is not visible or possibly shifted upwards (towards the peak) in the trace at point #1.

[0075] The following method enables not only to identify qualitative changes in the heartbeat between different glucose concentrations, but also to provide a quantitative measure that correlates the phase difference between the traces with the change in glucose concentration.

[0076] After pre-processing of the PPG signal, a Fourier transform is performed and normalized to the respective DC value (zero frequency) to provide a relative comparison of the modulation amplitudes. From Figure 6 It can be seen in that the fundamental frequency of about 1.3 Hz from the heartbeat is clearly visible in both traces. However, the amplitude is reduced for point #1 compared to the trace of point #33, and the harmonics are hardly visible. The amplitude of the fundamental frequency of about 1.2 Hz of point #33 is about 3 times larger than for point #1 corresponding to the lower glucose concentration.

[0077] The following figures describe a method to extract the AC part from the frequency spectrum and to determine complex coefficients from the frequency spectrum and more particularly from the complex Fourier transform of each PPG measurement, which are associated with the fundamental peak and the harmonics. In a system without correlation, the phase difference between the fundamental peak and the harmonics or between the harmonics should be highly similar in different measurements and thus not change over time. However, as already shown in the previous figure, the traces of the PPG measurements are influenced by the glucose concentration and so is the phase difference. More particularly, the phase difference between the fundamental peak and one of the harmonics or between the harmonics is used to determine the change in concentration.

[0078] To improve the data processing, some further pre-processing is performed to enhance the effect of the fundamental peak and the harmonics over the noise. The enhancement is achieved by taking the absolute value of the FFT to the power of M, where M can be an integer such as 2, 3, 4, etc. The enhancement is necessary when the actual signal from the heartbeat is rather small or the noise level is large, so that the harmonics only slightly stand out over the adjacent noise. Furthermore, a smoothing function is applied, in the exemplary embodiment a 3-point smoothing function, where a 3-point moving average is taken over the PPG measurement. The result of this additive pre-processing is visible in Figure 7 for the fundamental of point #1 in where M = 6 and a 3-point average is taken.

[0079] The amplitude of the peak is larger than the noise level. In a first aspect, the frequencies f min and f maxwhere the amplitude reaches a certain level, in this case 5% of the overall maximum. Other peak values can also be chosen and can depend on the enhancement factor. The level detection determines the dynamic peak width as f max -f min To get the actual peak width, a static offset can be added to the dynamic width, but the advantage is that this offset only needs to be determined once for all measurements. Furthermore, the fundamental peak is not symmetrical, so the center of gravity must be determined in the next step using the dynamic width, but with an offset given by the difference between the middle frequency (f max -f min ) / 2 and the center of gravity.

[0080] This can be done by summing the product of the frequency and the FFT amplitude over the dynamic width of the previously determined fundamental frequency. The result is then divided by the sum of the amplitudes. In Figure 7 and Figure 8 In the depicted example, the center pixel is at position 13.93 (arbitrary value), which roughly corresponds to a frequency of 1.27 Hz. Although the center of gravity roughly coincides with the maximum in this example, it does not have to be so, and this depends on the distribution of the amplitudes. The final mask of the fundamental peak boundaries is shown in Figure 8 , which is applied to the original FFT of point #1, where the reference level of the peak is about 5%.

[0081] The mask, whose width and center frequency have both been determined, is then applied to the spectrum. More specifically, the total number of masks to be applied to the spectrum can be adjusted, typically between 3 and 10. The reason for the limited number of masks used to mask the harmonics and the fundamental is that the harmonics decay very quickly and will soon be completely drowned in the noise. Therefore, it is mostly not necessary to interrogate more than 10 harmonics, as there will not be any further contribution.

[0082] For this process, the mask is placed with the respective width of the fundamental peak obtained in the previous step and with a center frequency corresponding to a multiple of the fundamental frequency. If necessary, an additional offset can be applied. This process is repeated for all higher harmonics until the maximum number of harmonics that should be resolved and analyzed is reached.

[0083] This method is performed individually for each point. That is, the fundamental peak of each PPG measurement corresponding to one of the points on the trace of Figure 2 is used to determine the mask. The determined. Then, as discussed above, each of the masks is applied to the respective spectrum.

[0084] In a next step, the respective mask is multiplied with the spectrum of each point, thereby separating for each point the fundamental and the harmonic peaks (and the noise part below the peaks) from the noise between the respective masks. Then, in a subsequent but sometimes optional step, also the noise below the masked peaks of the fundamental and the harmonics is estimated, thereby being removed.

[0085] To this end, it is assumed that the noise is frequency independent and constant within the masking range, i.e. for each mask constant between f min and f max It is also possible to assume a slight frequency dependency, as the noise will typically decay with increasing frequency. To estimate the noise below the peaks, the gap in the noise spectrum caused by the mask is filled with the average of the neighboring points, corresponding to the limits of the mask on the lower and higher frequencies.

[0086] Figure 9A The mask obtained for the fundamental of point #1 applying up to the 10th harmonic is shown, with the noise between and below the peaks removed. Thus, the spectrum is assumed to be the pure AC part. Figure 9B The mask obtained for the fundamental of point #33 applying up to the 10th harmonic is shown, and also any noise between and below the peaks is removed. Note that the mask width and its center of gravity are determined separately for each measurement point over time. This is necessary, as the heart rate can change during the overall time period.

[0087] The complex Fourier transform of the respective PPG measurement for each point provides a plurality of complex coefficients, wherein these coefficients can be associated with the respective fundamental peak and its harmonics. Thus, the complex coefficients within each respective mask of each point are summed and referred to as the complex number where I is the respective measurement point (e.g. point #1 and #33 as shown in Figure 9A and Figure 9B The number of fundamental peaks or harmonics. For example refers to the 2nd harmonic of the 10th data point. Figure 9A and Figure 9B The respective complex numbers of the peaks and harmonics are shown. It should be noted that each complex number consists of one or more complex coefficients as predicted, which are determined during the complex Fourier transform.

[0088] The phase is thereby defined as:

[0089] where imag is the imaginary part of the complex number, real is the corresponding real part. Atan is the four quadrant arctangent. The phase difference between the fundamental peak and the 2nd harmonic of data point I is thereby defined as:

[0090] This process is repeated for all data points I acquired throughout the measurement period, therefore, for example, for... Figure 2 All points in the oral glucose tolerance test.

[0091] Figure 10 The relative phase of red light between the fundamental peak and the first harmonic is shown during the oral glucose tolerance test (OGTT), where food intake occurred at approximately 9.45. For each data point shown, a PPG trace consisting of 2048 samples was obtained. No averaging or outlier removal was performed on the data.

[0092] The reference curve was 5 minutes ahead in time. A time lag was observed in the reference curve relative to the optical measurement in all measurements. One possible explanation lies in the nature of the OGTT itself. The reference curve measures glucose concentration in the interstitial fluid near the skin surface, while the optical measurement reaches deeper layers, including blood vessels. In this interstitial fluid, an increase in glucose levels may simply require a longer time. Figure 3 The same observations were obtained, delayed by 8 minutes, using completely different samples. In addition to this observation, a strong correlation was found between phase difference and glucose, allowing for the determination of changes in glucose measurements based on observations of the phase difference.

[0093] Return to reference Figure 3 And the phase difference between the fundamental peak and the first harmonic or green light, as shown therein. It has been shown that for different wavelengths, the noise and variability of PI and other measurements related to glucose concentration in blood are greater. It can be assumed that this behavior exists not only in glucose measurements but also in the measurements of other substances in liquids. Therefore, the proposed method of using the phase difference between individual peaks to determine the concentration of substances in liquids appears to be more robust than other measurements.

[0094] This is about Figure 11A and Figure 11B An explanation was given. Figure 11A and Figure 11B The same test sample was shown, similar to Figure 3 The test sample shown. The optical measurement results are the same, but... Figure 11A and Figure 11B The perfusion index, i.e., the ratio between AC and DC, was evaluated. Figure 11A PI is also shown, but with a 5-point moving average and outlier correction. Without smoothing and / or outlier correction applied to the perfusion index results, Figure 3 The variability of the relative phase shift appears to be significantly smaller than the variability of the perfusion index. Figure 10). But even after some smoothing, the traces only match the phase difference determined without any additional post-processing and trace optimization. Under noisy conditions, the relative phase between harmonics remains relatively intact, and the proposed method is more robust against small glitches and other disturbances during the measurement.

[0095] Figure 13A and Figure 13B The same measurement results are shown for Figure 3 but with different wavelengths. Both show the phase difference between the primary fundamental peak and the first harmonic, while the phase difference is also expressed in terms of glucose concentration. No additional averaging or any other post-processing has occurred. The three graphs obtained at three different wavelengths, i.e. red, green and IR, show that for all wavelengths with a strong correlation to blood glucose measurements, the phase difference is a very robust measurement. Thus, by providing a reference value for the optical measurement, e.g. the phase at a certain reference glucose level, it is easy to determine the change in glucose concentration by evaluating the phase difference between and applying an obvious linear relationship.

[0096] Due to the correlation of all wavelengths, the proposed method is applicable to various skin types and different situations where other optical measurements can have deficiencies or require additional corrections.

[0097] Another aspect relates to the evaluation of the phase difference between the primary fundamental peak and the first harmonic as well as the phase difference between the fundamental peak and another harmonic. Figure 14A and Figure 14B The measurement is shown using infrared light. Figure 14B The perfusion index is obtained during approximately 10:25 to about 11:35. Each PPG measurement lasts about 10 seconds, with one measurement per minute. The data of the perfusion index is processed with a smoothing function of a 5-point moving average. The reference trace REF is shifted in time by 8 minutes to compensate for the delay. The perfusion index trace generally follows the trace of the time-shifted reference trace REF, still with some feasibility, especially during the rise of the glucose level.

[0098] Figure 14AThe trace depicting the phase difference between the fundamental peak and the second harmonic is shown, referred to as PHASE 3-1. This trace depicts the rise in glucose level very well and closely follows the reference trace. The relative phase starts at 0 during the intake and increases to a level of about 15°. The drop in glucose level is also well depicted together with the reference trace, including the shoulder on the falling side in the reference trace and the phase difference trace PHASE 3-1. It can thus be concluded that the correlation between the phase difference and the glucose concentration exists not only between the fundamental peak and the first harmonic, but also between other phase differences. These statements hold equally for the phase differences between harmonics, such as the second and third harmonics.

[0099] However, before enhancement by raising the peak to the Mth power (as mentioned before), the signal level of the harmonics should be sufficiently higher than the noise level. This will typically limit the approach to the fundamental peak and at most the fourth or fifth harmonic.

[0100] The proposed method is suitable for different distances between the emitter and the detector, as well as for different angles of incidence of the light source onto the probe and the detector. For example, measurements were made at various distances between the light spot on the tissue by the light source and the light detection by the detector, i.e. at 2 mm, 4 mm and 6 mm. Likewise, different angles were used for the incident light, i.e. 10° and 25° for the incident light from the light source relative to the normal to the tissue. The detector was also tilted 10° and 25° relative to the normal vector to the tissue. All three wavelengths of green, red and IR light were measured.

[0101] This setup enables 12 PPG channels to be acquired in parallel with a single 2048 PPG trace acquisition. For each of the 12 acquired PPG channels, the phase shift between the fundamental peak and the first harmonic was calculated. The results are shown in Figure 15 as raw and unprocessed data, where all 12 traces are plotted. Some of the traces seem to show a better correlation than others. This is partly due to the measurement and partly due to the processing of the PPG measurements over time. In particular, there is a banding of the phase in some of the traces and some of the traces are transitioning from one band to another. The spacing between the bands is approximately 2π. By using the appropriate mathematical functions, such as atan2 in Matlab and subsequently the Unwrap function, it is possible to return the four quadrant arctangent with phase jumps > π by adding multiples of ±2π until the jumps are less than π. This approach is triggered by the behavior of the light with 2π properties, resulting in a better correlation between the various phase differences and the glucose concentration.

[0102] Averaging all 12 processed and unwrapped traces results in an average phase difference, which is shown in Figure 16The phase difference between the fundamental wave peak and the first harmonic with respect to the reference trace is shown in Fig. 16. Here, the correlation between the phase difference and the glucose measurement is significant. The respective concentration of glucose in the blood can be obtained with the phase difference, either its change or - with a proper reference value - even the absolute value.

[0103] A further aspect is given by the amount and the content of the consumed sugar. It can be reasonably assumed that different amounts of sugar also lead to differences in the rise, the plateau and the subsequent decline. Figure 17A The phase difference between the fundamental wave peak and the first harmonic with respect to the reference trace is shown in Fig. 16. Here, the correlation between the phase difference and the glucose measurement is significant. The respective concentration of glucose in the blood can be obtained with the phase difference, either its change or - with a proper reference value - even the absolute value. Figure 17B The corresponding Clarke error grid analysis with high confidence is shown.

[0104] Figure 18A and Figure 18B Now is another measurement, in which only half of the juice was consumed. In addition, a slightly different wavelength in the green spectrum was used. Figure 18A The measurement shown in Fig. 17 is still comparable to the measurement in Fig. 16, but a different wavelength was used, since the previous measurement showed a phase difference of about 15° between the low glucose measurement and the maximum glucose measurement, while the other parameters were identical. It can therefore be assumed that the phase difference is independent of the various wavelengths, as also indicated by the previous results. Figure 17A

[0105] Figure 18A The phase difference is about half of the previous measurement. The rise of the glucose level from the starting point to the maximum is roughly the same, indicating that all the glucose has been transferred to the bloodstream within about half an hour. The time taken for the decline is roughly the same as the rise, but the plateau is slightly shorter than in the previous measurement. Figure 18B The corresponding Clarke error grid analysis with high confidence is shown.

[0106] The two measurements in Figs. 17 and 18 show that not only can the glucose concentration be identified quantitatively and qualitatively, but also information about the glucose metabolism in the human or animal body can be provided. The optical measurement according to the proposed principle therefore opens up a new opportunity window for diagnostics.

[0107] List of reference signs

[0108] 1 detector device

[0109] 10 control and evaluation circuit

[0110] 11 housing

[0111] 12 light source, LED ​

[0112] 13 detector

[0113] 13.1, 13.2 detector area

[0114] 14 optical barrier

[0115] 21 glass interface

[0116] 20 surface

[0117] 32 optical path

[0118] 30 tissue

[0119] 32 optical path.

Claims

1. A method for determining a concentration of a substance in a sample comprising a liquid containing particles, in particular glucose in blood, wherein, The refractive index of the liquid depends on the concentration of the substance dissolved in the liquid and the density of particles in the liquid is substantially constant, wherein the volume or pressure of the liquid is modulated, comprising the steps of: - obtaining a first signal during a first time period at an acquisition rate at least two times larger than the periodicity of the volume or pressure modulation of the liquid; - extracting at least some complex coefficients from an AC part in a frequency spectrum of the first obtained signal, wherein the complex coefficients are associated with at least one of a fundamental peak and a harmonic peak in the AC part; - determining a relative phase between different peaks based on at least a subset of the extracted complex coefficients; - obtaining one of: o the substance concentration based on the extracted relative phase and optionally based on a reference value; and o a change in the substance concentration based on the extracted relative phase and optionally based on a reference value.

2. The method according to claim 1, further comprising: - obtaining a second signal during a second time period at an acquisition rate at least two times larger than the periodicity of the volume or pressure modulation of the liquid after obtaining the first signal; - extracting at least some complex coefficients from an AC part in a frequency spectrum of the second obtained signal, wherein the complex coefficients are associated with at least one of the fundamental peak and peaks at positions in frequency space that are integer multiples of the fundamental frequency, in particular harmonic peaks in the AC part; - determining a relative phase between the complex coefficients; wherein optionally the reference value is based on the extracted relative phase of the second obtained signal.

3. The method according to any of the preceding claims, wherein, The reference value is given by a predetermined extracted relative phase or by a value derived from the extracted relative phase.

4. The method according to any of the preceding claims, wherein, The step of extracting at least some complex coefficients from a frequency spectrum comprises the steps of: - identifying an AC part in the frequency spectrum of the first and second obtained signals, the AC part corresponding to the fundamental frequency and harmonics of the volume or pressure modulation of the liquid; - computing a set of complex coefficients from at least one of the fundamental frequency and the harmonics.

5. The method according to any of the preceding claims, wherein, The AC part corresponds to the fundamental frequency and harmonics up to the ninth order, in particular up to the seventh order and in particular up to the fifth order and in particular up to the fourth order.

6. The method of claim 4, wherein, The step of extracting at least some complex coefficients from a frequency spectrum comprises the steps of: - identifying a noise part within the spectrum of the fundamental frequency and harmonics of the volume or pressure modulation of the liquid based on values of adjacent points in the spectrum; or - identifying a noise part of a masked part of the spectrum based on points in the spectrum at a border of the masked part; and - removing the identified noise part at the fundamental frequency and harmonics from the spectrum.

7. The method of claim 4, wherein, The step of identifying an AC part in the frequency spectrum of the first and second obtained signals comprises at least one of the steps of: - applying a smoothing function to the frequency spectrum of the first and second obtained signals; - enhancing at least the fundamental frequency peak, in particular by taking the absolute value of the spectrum to the power M, where M is an integer larger than 2; - determining a mask having a width based on a predetermined amplitude level of at least the fundamental frequency peak; - optionally determining a center of at least the fundamental frequency peak by using the width based on a predetermined amplitude level of at least the fundamental frequency peak; - applying a mask, in particular the determined mask, on a predetermined number of peaks corresponding to the fundamental frequency and the harmonics of the volume or pressure modulation of the liquid, thereby identifying the AC portion.

8. The method of any of the preceding claims, wherein, The step of obtaining the first signal and / or the second signal comprises at least one of: - low pass filtering the first obtained signal and / or the second obtained signal to remove transients in particular with a frequency below 2 Hz and in particular below 1 Hz and in particular below 0.5 Hz; - removing dark counts from the first obtained signal and / or the second obtained signal; and - converting the first obtained signal and / or the second obtained signal into a digital signal. The step of obtaining the first signal and / or the second signal comprises the step of:

9. The method of any of the preceding claims, wherein, - performing a complex Fourier transform or a fast complex Fourier transform on the first obtained signal and / or the second obtained signal. The step of determining the phase difference comprises:

10. The method of any of the preceding claims, wherein, - determining a first complex number corresponding to a sum of complex Fourier coefficients, in particular associated with one of the fundamental wave peak and the harmonics; - determining a second complex number corresponding to a sum of complex Fourier coefficients associated with the other one of the fundamental wave peak and the harmonics; - determining a first phase based on the first complex number and a second phase based on the second complex number. The first phase and the second phase are determined using a geometric function, in particular using a four quadrant arctangent taking the imaginary and real part of a complex number as input.

11. The method of claim 10, wherein, The liquid is blood and the volume and / or pressure modulation of the blood is defined by a heartbeat.

12. The method of any of the preceding claims, wherein, 13. The method according to any of the preceding claims, further comprising: - calculating a heart rate from the identified AC portion in the frequency spectrum; and / or - obtaining a blood glucose dependency from a change in the heart rate. The refractive index of the liquid depends on a concentration of the substance dissolved in the liquid and a density of particles in the liquid is substantially constant, the detector device comprising:

14. Detector device for determining the concentration of a substance in a sample comprising a liquid containing particles, in particular glucose in blood, wherein, - at least one light source and at least one detection component, wherein the at least one detection component is optically separated from the at least one light source; - wherein the at least one light source is configured to emit light through an exit window onto a sample; and - the detection component is configured to detect a light component corresponding to scattered through emitted light; - a control circuit coupled to the at least one light source and the at least one detection component; wherein the detector device is configured to perform the method according to any of the preceding claims. The control circuit is configured to control the at least one light source to emit a light signal in a first time period at a plurality of different consecutive time instances and to obtain a signal from the at least one detection component.

15. The detector device of claim 14, wherein, The detector device comprises a plurality of photodetectors arranged in a ring shape or a quadrangle shape and optionally in particular at different distances around the at least one light source in particular centered.

16. The detector arrangement of claim 14 or 15, wherein, ​ 17. The detector device according to any one of claims 14 to 16, wherein, The at least one light source comprises a plurality of optoelectronic devices arranged at different distances from the at least one detection component; and optionally comprises a ring shape or a quadrangle shape, optionally arranged around the at least one detection component, in particular centrally.

18. The detector device according to any one of claims 14 to 17, wherein, The at least one light source is configured to emit light of different wavelengths.

19. The detector device according to any one of claims 14 to 18, wherein, The at least one detection component comprises a light filter comprising a low transmissivity in a spectrum different from the spectrum of light emitted by the at least one light source.

Images