WO2011026986A1

WO2011026986A1 - An optical device for sensing a plethysmographic signal using a matrix imager

Info

Publication number: WO2011026986A1
Application number: PCT/EP2010/063061
Authority: WO
Inventors: Salim Mimouni
Original assignee: Salim Mimouni
Priority date: 2009-09-07
Filing date: 2010-09-06
Publication date: 2011-03-10
Also published as: FR2949658B1; FR2949658A1

Abstract

The invention relates to a system (1) for detecting signals of the PPG type from a biological tissue, including: a) a supporting area (2') for a portion of the tissue, b) a matrix (2) of nxm elementary optical receivers, and means for forming an output signal of each elementary receiver, depending on radiation which it receives from a tissue bearing upon the supporting area and as a function of time, c) means for forming a PPG signal from the output signals of the receivers of the matrix.

Description

AN OPTICAL DEVICE FOR SENSING A PLETHYSMOGRAPHIC SIGNAL

USING A MATRIX IMAGER

DESCRIPTION

TECHNICAL FIELD AND PRIOR ART

The invention relates to a medical signal sensor, notably for a photoplethysmographic

(abbreviated in the following by: « PPG ») signal.

It also relates to an apparatus for measuring optical constants of a biological tissue examined by a sensor according to the invention.

It may be used for detecting cardiovascular diseases, for example via still (or video) cameras existing as measurement terminals. These mobile, nomad devices or computers capture the signal of nomad applications connected to servers and health web sites.

Coupled with engines for extracting health parameters, the invention allows delivery of a first analysis of the operation of the patient's cardiac cycle (pulsation frequency, detection of arrhythmias, follow-up of the arteries and of their stiffness, upstream detection of cardiovascular diseases or abnormalities) and possibly of other health parameters (oxygen saturation, alcohol level, glucose concentration) .

In many developed countries, cardiovascular diseases affect more than one person out of three, kill more people than cancer, and remain the first cause of mortality in developed countries with 45% of the recorded causes of death. The particularity of this disease is its difficulty of being detected: it is often necessary to await a first cardiovascular stroke for starting treatment.

In hospitals, there exist advanced means for complete cardiovascular examinations, including electrocardiogram (ECG) apparatuses (available in hospitals and at cardiologists) which require slow and repeated examinations, in a very specialized environment, in order to result in diagnosis. Moreover, the processing of ECG signals has taken a leap forward during this last decade and saw the introduction of advanced interpretation techniques based on fuzzy or classification logic based on neuronal networks.

But the accessibility to this type of cardiac examination is highly reduced relatively to the relevant population. In this major public health context, the problem is posed of providing a means for preventing cardiovascular diseases, which is easier to apply .

There is therefore an increasing need for low cost simple-to-use and accessible sensors.

When light is sent with a LED in an adequate spectrum (generally in the red and infrared) , blood absorbs a variable amount of light depending on the volume of the vein or artery and on its oxygen saturation. The heart beat generates an acoustic wave which propagates along the arteries and deforms them like a wave in a pipe. Having arrived at the walls, the acoustic wave returns. This wave form comprises a systolic and diastolic peak. This deformation of the arteries generates a change in the volume and therefore in the light absorption. By integrating a photodiode, the voltage changes of the latter therefore correspond to the cardiac cycle. As the measurement is optical and measures a change in volume, it is called a photoplethysmographic (PPG) measurement. The signal recorded by the photodiode has two components, called the pulsation component (AC) and the slowly variable component (DC) . The AC signal is synchronous with the heart beat while the DC signal reflects the total change in the absorption from the blood due to different physiological parameters such as breathing.

This standard type of device applies a discrete source/detector pair. But a measurement conducted with the type of device suffers from a major defect: that of the noise due to the micromovements between the finger and the probe.

For these reasons, the probe is accompanied by a clamp allowing the finger to be fixed against the source/detector pair, and, the movement artefacts to be thereby reduced in principle.

By using a clamp it is in principle possible to benefit from the following advantages:

A reduction in the movement of the finger relatively to the emitter,

A reduction in the movement of the finger relatively to the source, but which remains present, and changes the distribution of the light in the biological tissue in an unstable way.

Constant pressure is maintained on the finger or the tissue during the measurement.

But these advantages do not obviously appear upon use: the noise is still very present, and globally remains not very exploitable in the case of motion .

Indeed, in the literature, it is seen and reported that the noise is still very present, and that the signal globally remains not very exploitable in the case of motion.

An exemplary device used for this type of detection is the one of document US 2006/0287589. It includes a standard probe structure which is found in apparatuses with which the signal may be obtained. The clamp includes one or more source/detector pairs.

In this type of device, the signal is first perturbed by tremors of the finger under the clamp, and requires a plurality of « hardware » filters for efficient filtering and automatic suppression of coarse artefacts. Further, perturbations will be added, which result from possible sudden movements.

For reasons both of cost and of accuracy, the problem is therefore posed of finding a novel device, as an alternative to the existing systems with clamps .

DISCUSSION OF THE INVENTION

A sensor and method according to the invention apply a set of detectors or receivers juxtaposed and positioned as a matrix, on which a biological tissue or an organ, for example a finger of a user may be positioned, or else onto which the image of this tissue or this organ may be projected.

The invention thus relates to a method for detecting signals of the PPG type, and possibly other optical constants, from a biological tissue, these signals being formed from output signals from receivers of a matrix of nxm elementary optical receivers, depending on radiation received by each receiver.

During the capture of the signal, the set of detectors among those which actually perceive the signal or a significant fraction of the latter, may be selected. This selection may change over time and therefore during the displacements of the tissue or of the organ on or facing the matrix or as a projection on the matrix. In other words, a set of detectors which actually perceive the signal or a significant fraction of the latter are selected with this device almost in real time. The effect of a movement of the organ on the measured signal is therefore neutralized.

Thus, in the course of time, extraction of a signal is achieved with an optimum signal-to-noise ratio .

This reduction of the impact of movements on the signal allows production of a signal which is easy to exploit by using simple signal processing.

It is therefore possible not to use any clamp, such as the one often associated with this type of probe. In a device or a method according to the invention, the signal is present as long as the organ is laid upon, or positioned facing the matrix (regardless of the exerted pressure) and remains of very good quality even in the case of movements.

Preferably, the method according to the invention is applied with only ambient natural light in the absence of sources of additional external lights. Before any measurement on a tissue or an organ, it is possible to analyze the contents of the ambient light and thereby use the data relating to this ambient light .

The PPG signal may be formed by calculating the average of the intensities from the receivers and by evaluating the time-dependent change of this average .

The PPG signal may be formed from a histogram of the intensities of the output signals from the receivers of the matrix, the time-dependent change of which is identified. This time-dependent change may be defined by the time-dependent change of the position of the maximum and/or of the position of the barycenter or of the isobarycenter of the histogram and/or of the peak of a function for fitting the histogram.

It is possible to filter the time-dependent change of an histogram in order to suppress non-PPG contributions and/or extract the slowly variable envelope of the PPG signal.

It is further possible to calculate the oxygen saturation rate if a second source with a different spectrum but in the red-infrared range is available .

Moreover, by obtaining time-dependent data as matrices (or images) it is possible to obtain a mapping of diffuse reflectivity and to therefore calculate one or more optical characteristic constants of the relevant biological tissue, for at least one wavelength, for example the diffusion constant or coefficient (μ₃) and/or the absorption constant or coefficient (y_a) to the optical constants ys (diffusion coefficient) and ya (absorption coefficient) .

It is also possible to form a representative signal of the period including a systolic lobe and a diastolic lobe and/or of the regularity of the heart rate and/or of the resting heart frequency and/or of time intervals between a peak of a systolic lobe and a peak of a diastolic lobe.

These various processing operations may be achieved by an individual device associated with a user, and a portion of them may be achieved by a remote server .

The invention also relates to a system for detecting signals of the PPG type from a biological tissue, including:

a) a supporting area for a portion of the tissue,

b) a matrix of nxm elementary optical receivers or sensors, and means for forming an output signal from each receiver or sensor, depending on the radiation which it receives from a tissue bearing upon the supporting area and depending on time,

c) means for forming a PPG signal from output signals of the receivers of the matrix.

Preferably, such a device only uses natural light, in the absence of any additional light source, for example any light source which would be mechanically interdependent with the system.

Thus, a representative signal of the contents of the ambient light may be formed beforehand before laying any organ tissue on the supporting area. It is then possible to use ambient light.

A device according to the invention may further include means for measuring optical constants of a biological tissue (μ and y_g) , which can only be measured by having available spatial or mapping data of the diffuse reflectivity over time.

Such a system may further include at least one radiation source, which may be mechanically interdependent with the system.

The supporting area may be defined by:

- the whole of the surfaces of the receivers intended to receive an illumination,

- or the surface of a transparent layer positioned on the whole of the surfaces of the receivers intended to receive an illumination,

- or the surface of an optical device positioned on part or all the surfaces of the receivers intended to receive an illumination.

The architecture of a system according to the invention is compact, simpler and less costly than known devices.

On the contrary, a device of a known type, as presented in the introduction, has discrete detectors, i.e. they are positioned in an unitary way or isolated from each other and have an optical source interdependent with the system as an integral part of the latter. The addition of a second detector element then does not bring any functionality, since it reproduces a similar signal. Such a system may include means, for forming in a dependent way, a mask defining at the surface of the matrix, an area of receivers receiving radiation having passed through a portion of a tissue bearing upon the supporting area.

The means for forming a PPG signal may include means for calculating the average of the intensities from the receivers.

The means for forming a PPG signal may include means for forming a histogram of the intensities of the output signals from the receivers of the matrix and means for identifying the time-dependent change of a histogram. The time-dependent change of a histogram may then be defined by the time-dependent change of the position of the maximum and/or of the position of the barycenter or of the isobarycenter of the histogram and/or of the peak of a function fitting the histogram.

Such a system may include means for filtering the time-dependent change of a histogram in order to suppress non-PPG contributions.

It may further include means for calculating the oxygen saturation levels.

It may include means for forming a two-dimensional distribution of the diffuse reflectivity of the tissue on the matrix of elementary optical receivers of the system (which is also called a matrix imager) , and for calculating one or more optical constants or characteristics of the relevant biological tissue, for example the diffusion constant (μ₃) and/or the absorption constant (y_a) of the relevant biological tissue, for at least one given wavelength.

It may further include means for extracting the slowly variable envelope of the PPG signal.

According to a particular embodiment, a system according to the invention includes means for forming a representative signal of the period including a systolic lobe and a diastolic lobe and/or of the regularity of the heart rate and/or the resting heart frequency and/or time intervals between a peak of a systolic lobe and a peak of a diastolic lobe.

A system according to the invention may include a wireless telephone which includes at least the means a) and b) . The wireless telephone may further include additionally at least one portion of the means c) .

Alternatively, a system according to the invention may include a camera which includes at least the means a) and b) , and a computer.

In such a system, it is the computer which may further include at least one portion of the means c) , and which will therefore perform the calculations and processing operations in order to obtain a representative signal of a PPG signal.

A system according to the invention may further include a server which includes at least one portion of the means c) , and which, will therefore perform the calculations and processing operations for obtaining a representative signal of a PPG signal.

A system according to the invention may further include at least one infrared filter for filtering radiation received from the tissue bearing upon the supporting area. Shape recognition processing may advantageously be applied.

The invention may therefore be implemented as a health application in a mobile telephone or a PC peripheral by exploiting pre-existing sensors in a mobile telephone or a « webcam » of a micro-computer. A signal may then be transferred, from such a device, towards a web site hosted on a centralized server, for example of the type providing access to a health service through a web portal.

The invention is not a substitute for public health means, but is located upstream and allows presentation of the patients in a more targeted, more precocious way.

With the invention it is possible to reveal cardiac or coronary abnormalities, and/or arrhythmias, and/or the formation of blood clots, causing few or no symptoms before the cardiovascular stroke. It also allows the tracking of the time-dependent change of a certain number of health parameters: the regularity of the beat, and/or the resting heart frequency, and/or the rebound rate of the diastolic peak, and/or the transit time between both peaks which allows arterial stiffness to be calculated.

Applying the invention entails a reduction in the costs for individuals and the community, related :

- to reducing the management of emergencies and transports to locations where sophisticated detection and analysis means are presently found, - to applying palliative rather than curative treatments.

SHORT DESCRIPTION OF THE DRAWINGS

- Figs. 1A-1C illustrate three embodiments of a sensor according to the invention,

- Fig. 2A is a diagram explaining the composition of the signal of the imager,

- Figs. 2B and 2C schematically illustrate the formation of a mask,

- Figs. 3A and 3B respectively are an example of measurement of the spectral transmission of a finger and of color filters,

- Fig. 4 is an example of breaking down an image acquired with a device according to the invention,

- Figs. 5A-5C illustrate various signals obtained with a device and a method according to the invention,

- Fig. 6 is the spectrum of ambient light through the finger of a hand,

- Fig. 7 illustrates a portion of a signal obtained with a device and a method according to the invention and various characteristic quantities of this signal ,

- Figs. 8A-8B illustrate an embodiment of a device according to the invention, with application of a webcam and a computer,

- Figs. 9A-9C illustrate various signals obtained with the device such as the one of Figs. 8A and 8B, - Fig. 10 illustrates an embodiment of a device according to the invention, with application of a mobile telephone,

- Figs. 11A-11B illustrate images obtained with a device according to the invention;

- Fig. 12 illustrates an embodiment of a device according to the invention, with application of a server,

- Figs. 13A-13B illustrate another embodiment of a device according to the invention,

- Figs. 14A-14D and 15A-15C illustrate various signals obtained by a method according to the invention,

- Fig. 16 illustrates the course of a method according to the invention.

DETAILED DISCUSSION OF PARTICULAR EMBODIMENTS

It is recalled here that the photoplethysmographic (PPG) signal is due to the acoustic wave sent by the heart for pumping the blood through the arterial and vascular tree. This wave is expressed by a volume increase V of the arteries, of the arterioles, veins and other blood conduits. Various means exist for detecting this pressure wave. The invention applies an optical method, which may use ambient light or another source of light, and it uses positioning, for example direct contact, between a biological tissue or an organ of a living being and a matrix imager.

Now, a probe or sensor for acquiring physiological signals according to the invention are described, including an optical probe and suitable processing of the data. In the following both expressions « probe » and « sensor » are equally used.

As illustrated in Fig. 1A-1C, a probe 1 according to the invention no longer includes one or two receiver elements, but a matrix 2 of nxm (for example: n=640, m=480, more generally, n is for example comprised between 300 and 1, 000, and m is for example comprised between 300 and 1,000) receivers positioned as in an imager or sensor of the CCD or CMOS type for example. In most cases, an individual receiver such as a commercial receiver, delivers a signal on three channels, red R, green (designated by V hereafter) and blue B, and is equipped with three filters, one for each of these spectral ranges. An exemplary spectral profile for each of these filters is given below in connection with Fig. 3B.

These receivers form a detection field with a surface for example comprised between a few mm² and a few cm², for example between 1 mm² and 1 or 2 cm² or 30 mm² or 50 mm² or between 1 cm² and 5 cm² or 10 cm².

In the following, both expressions « receiver)) and « pixel» are equally used for designating an individual detector of the field of the matrix .

Preferably, a device or system according to the invention applies only ambient light without any additional source of radiation. But, optionally, one or more additional sources of radiation 4, 6 may be used, for example a source 4 of the LED type emitting in the red spectrum and/or a LED 6 emitting in the infrared spectrum. It is also possible to use only ambient light, without any additional source. The spectrum composition of ambient light may vary during the day, but it may however provide a usable source. Therefore, passive use of the system may thus be achieved.

Generally, exploitable measurements correspond to sources containing a spectral range in the red/near infrared (for example comprised between 600 nm and 1,100 nm) . For example, the signal may be obtained with a 950 nm LED source, as well as with white source such as the sun or an incandescent source emitting through the natural filter of the finger (Fig. 3A) .

The PPG signal reflects dynamics of the wave sent by the heart into the arterioles/veins and arteries, and the latter is then reflected on the walls and impacted by the distension and reflection properties of the arteries and of their inner walls.

The light source probes at various instants these time-dependent changes in terms of absorbed or transmitted intensity, and the various properties are visible and analyzable on the PPG signal.

The field of the matrix may directly be in contact with the biological tissue, for which extraction of the signal is intended, this is the case of Fig. 1A: in this figure, a finger 10 is directly in contact with the field of the sensor.

In the embodiment of Fig. IB, the surface of the sensor is separated from the biological tissue by a planar layer 12 which totally or partly lets through the light used (for example a protective plate, or a filter, for example a standard color filter, promoting the near-infrared or another spectral band) . The finger 10 will then come into contact with the surface of this layer and the sensors or detectors see as a projection, an image of the contact area of the finger on the layer 12.

In the embodiment of Fig. 1C, the surface of the sensor is separated from the biological tissue by at least one optical element or component 14, for example a lens. The tissue 10 will then come into contact with the surface of this optical element or component. There may further be a layer such as the layer 12. There again, the sensors or detectors see an image of the contact area of the finger, optionally transformed by an optical element.

In all these figures, reference 10 designates a biological tissue, here a finger, from which a PPG signal is intended to be extracted. This tissue bears upon the surface of the matrix, or of the layer 12 or of the means 14, this surface therefore forming a supporting area 2 ' . Each pixel provides a time-dependent signal s(i,j,t). The portion of the light absorbed by hemoglobin may therefore be recorded as a function of time by means of the different pixels of the imager.

In all the configurations described above, and in those which will be described below, the signals provided by each of the pixels are digitized, and may then be stored in the memory of memory storage means, in order to be then processed by signal processing means specially programmed for this purpose, by applying one or more of the processing operations described below.

In order to properly explain the advantages of these structures of sensors according to the invention, the simple configuration is considered here, which consists of a single source (in the sense that only a narrow range of wavelengths around a given wavelength λο is present) , the light of which passes through the biological tissue (end of the finger for example) which is laid on a matrix 1 of a device according to the invention with nxm pixels (this is the configuration of Fig. 1A) . The light is partly absorbed by the veins 14 (see Fig. 2A) , the change in volume of which is in the course of time (due to the heart beat) accompanied by a change in intensity (or in illumination) of the light received per pixel, and on the whole of the imager.

More specifically, Fig. 2A is a transverse sectional diagram of a detail of a finger 10 on a sensor 1 according to the invention, explaining the composition of the obtained signal. A spatial contrast due to the relief 12 of the digital imprint is at the origin of the stationary envelope in which a time-dependent signal develops mainly due to the absorption of light by the hemoglobin of the blood and by other components of the tissue 14.

When a biological tissue is put into contact with a sensor according to the invention, all the elementary receivers of the matrix 2 are not necessarily affected as such: some will be totally covered by a portion of the relevant tissue, while the other ones will not at all be covered; or else all will be covered by the tissue, but only a portion of receivers delivers a signal. It is therefore of interest that a system associated with a sensor according to the invention first allows real-time identification of the area of interest in the field of the matrix of receivers.

The area of interest 20 may first be defined as being the surface of the sensor with which the tissue is maintained in firm contact with the receivers of this surface. In the case of a configuration such as Fig. IB or 1C, this is the whole 15 of the receivers which are located facing or under the area of the layer 12 or of the optical component (s) 14, with which the tissue is in contact. In other words, an image of the tissue is projected on these receivers. This area is generally surrounded by a shaded area 17, due to the three-dimensional shape of the finger, or more generally by tissue which is bearing upon the supporting area 2 ' . The area of interest 20 of the sensor consists in these two areas 14, 17. Because of the movements of the finger, it changes in size and moves in the course of time. It may therefore be used for defining a mask in order to extract the useful datum or data. The area outside this mask represents parasitic light having quasi-zero probability of having been absorbed by the blood contained in the tissue.

Figs. 2B and 2C illustrate this aspect: this is a top view of the field 2 of the pixels of a device according to the invention. Different areas (area 15 with dense hatchings, area 17 with less dense hatchings and white area 19) are visible therein. The area 15 is the one which is strictly the image of the area of the image of the finger with the device. The area 17 corresponds to the shaded area indicated above. Both of these areas form a mask 20. Beyond, the area 19 is found which was only able to sense parasitic light.

A way for identifying the receivers (or pixels) located inside the mask, is to identify the receivers for which the signal varies over time. Indeed, the PPG signal, which is desirably detected, is by definition a signal which has time-dependent changes. The pixels, the output of which has remained substantially constant over time, are therefore suppressed from the masking area 20.

In the case when the field of the whole of the receivers is, relatively to the examined tissue, much smaller than that illustrated in both of these figures 2B and 2C, i.e. in the case when the tissue entirely covers the whole of the receivers with it, other techniques may be used. In particular, it is possible to define the mask 20 to be used as a binary mask determined from a threshold value equal to a fraction, for example the half or third of the value of the pixel receiving the maximum of transmitted light. All the pixels which deliver a greater intensity than that of the threshold value are comprised in the mask area. Other types of more developed masks may be used, for example having a smooth transition between the strongly weighted pixels and the less significant pixels; for example the value of the actual intensities may be used as a mask, or quite simply a set combining a uniform mask with absence of a mask.

But such a mask may change over time, with the position of the tissue or of the image of the latter.

With the invention it is optionally possible to provide a pressure index, which may be calculated depending on the ratio between the area of interest 20 and the total area (an area of interest to which the area 19 is added) . This pressure index is a tool which allows calibration of the respective participation between the venous and arterial signal, in the measured signal.

Moreover, the spatial distribution (i,j) of the diffuse reflectivity of the tissue on the imager allows, by means of diffusion models known in the literature, with knowledge of the size of the elements of the imager, calculation of the absorption constants (y_a) and of the diffusion constants (μ₃) of the relevant tissue.

The individual receivers will provide a PGG signal (relative to the change in volume of the veins/arteries and therefore to the changes in received light) of the video type: s (i, j, t) , wherein s is the signal of the pixel (i, j) of the matrix nxm at instant t. With this signal it is possible to again find the PPG signal with the best signal-to-noise ratio, and is richer than in the case of a standard clamp (where movement of the finger is not recognized) , since it contains additional data: time intervals of strong movement, and then pressure index of the area of interest .

As the signal is distributed over a significant surface (from a few mm² to a few cm²) , it has particularities which are such that a PPG signal of better quality (than with a standard probe with discrete receivers) may be extracted therefrom.

Reduction of the movement artifacts is first observed. On a single receiver element, the movements of the finger in the plane (translation) and out of the plane (elevation) cause sudden signal fluctuations which have to be suppressed by algorithms or corrected by elaborating adequate models. With a device according to the invention, the same movement always causes artifacts on a group of particular pixels (io, jo) (the « 0 » group) but the displacement of the finger promotes another group of pixels (ii, ji) (the « 1 » group) . In other words, the area of Fig. 2B will change over time. It is then possible to drastically reduce the amplitude of the artefacts either by selecting (with the mask) the signal on the area of interest, or by performing a spatial average (on the pixels i,j) which has the effect of compensating the noise of the « 0 » group by that of 1 group (which may be an alternative to the mask already defined) . The movement artefacts indeed are not uniform white noises, but noise related to the movement of the finger; now, for each pixel which passes from a contact condition to a non-contact condition with the tissue or its image (because of the displacement of the latter) , there exists another pixel which, conversely, passes from a condition of non-contact with this tissue or its image.

Further, local errors have a lesser effect. For example, if a parasitic light reaches a single receiver, the signal-to-noise ratio is degraded by this. This same degradation is divided by nxm (which is an order of magnitude of the improvement of the signal-to-noise ratio) . It is therefore possible to directly extract a signal from a matrix imager with lower external noise.

Thus, the mask function M promotes by its high value (the mask assigns to each pixel a value, the greater this value, the more the pixel counts in the calculation. A value of 0 indicates that the relevant pixel is removed from the calculation, a value of 100% indicates that the pixel is relevant and contains a large amount of signal. An intermediate value such as 50% indicates that the pixel is taken into account, but is twice less represented as the maximum pixel) M(io,jo,t) on the pixel {io,jo} at instant t, the signal s from the pixel group receiving the most light and having a better signal-to-noise ratio. This optimization is ensured at any instant t since the mask adapts to the position of the tissue in real time, and follows the vertical movements. When the pressure decreases, the tissue-imager contact becomes slight and the selection area is reduced, thereby suppressing the unnecessary noise from the pixels which do not receive any signal .

The advantage provided by the novel device which associates the laying of a biological tissue on a naked matrix sensor or protected by a transparent layer or with an optical function, was examined above: by using the masking function selectivity of the receivers may be achieved during the measurement. This selection may be achieved at every instant and it is thereby possible to give way to natural movements of the tissue (often a finger) on the sensor, without any stresses, and to thereby compensate induced micro-accelerations. This results in a simple-to-use device and with which a signal of excellent quality may be obtained.

With this advantageous technique it is possible to collect a signal transmitted from the ambient source towards the sensor. The resultant signal for each pixel is a time-dependent signal fo(i,j,t), which depends on the normalized spectral contents of the ambient light s (λ) , on the power of this source in terms of spectrum illumination (Watts/m²/nm) noted as Eo, on the spectral transmission of the biological tissue T (λ) taking into account the whole of the components of the skin (for example the stratum corneum, the epidermis, the dermis, and the absorbent components of the dermis, such as water and hemoglobin) and finally on the characteristics of the pixel which perceives this light. The latter take into account

its surface s_p,

its response as a photon receiver

< i >

R(k) = F( ) (^rati° between the photo-generated electric current and the light flux at wavelength λ, thereby expressed in A.W^-1) , the integration time of the pixel noted as τ

the equivalent capacitance of the stack of the pixel C_v

- and finally the total gain of the chain G.

Thus, in the case of the presence of polychromatic light (natural light for example, the signal level in binary LSB (low significant bit) units may be written as:

(1)

This recorded value of the pixel when the finger/tissue laid on the imager is compared with that recorded in open air (without positioning the tissue) noted as fo(i, j), which is written as:

The finger or biological tissue has a spectral transmission T (λ) which depends on its constituents and its thickness. However, the template (shape of the transmission curve) varies very little from one person to another. It then becomes possible to initiate calibration by using an absolute reference measurement, carried out by means of a calibrated spectrophotometer. In the case of color matrix sensors/imagers three reference values

(fo,i,R ;fo,i,v ,'ίο,ι,β) are thus obtained before application of the biological tissue, each corresponding to the successive impact of one of the sources 4, 6 through each Red (R) , Green (V) or Blue (B) filter, the responses of which are respectively noted: R_R,R_V and R_B. The following values are thereby obtained:

C ) = ^J_x ,_R(X)E(X)S_pd\

fo,i,v(i ) = g J R_v(X)E(X)S_pdX

J( .BU-:<) = ¾ fx RB(X)E(X)S_pdX ₍₃₎

These values are references for the following measurements, in the case of a color imager: fiA i) = g Λ RnW ) pd

The index 1 represents the active source (here the source 4) which was used for acquiring the measurement or the reference. For another source 6, the index 2 may be used.

The example of color filters does not restrict the use of these successive measurements of pixels of different sensitivities. For example, there are techniques for manufacturing sensors by printing a semiconducting ink, with which the filter effect may be replaced with adjacent pixels not having the same spectral sensitivity or response. It is then sufficient to replace in the formulae written above the response of the standard pixel combined with the Red filter with the specific response of the constitutive semiconducting material of the pixel R_Semiconductor (λ) . Fig. 3A is an exemplary measurement of the spectral transmission of a finger under ambient light. It is seen that the light in the red and infrared (beyond 600 nm or 650 nm) is transmitted with high intensity. Fig. 3B is an example of color filters, with which the matrix sensors may be equipped: there are three main bands, one in the red (R) , one in the green (V) and one in the blue (B) .

Fig. 4 represents a breakdown into Red/Green/Blue channels of an image obtained by the device according to the invention. The histogram H relating to each of the V and R channels is extracted (here on a basis of 256 elements) . The histogram of the red channels is designated by R, and that of the green channels by V.

How the PPG signal is obtained will now be described .

The raw signal is recorded during the acquisition period, for example of a few tens of seconds up to continuous acquisition. It is generally expressed as the data of a curve or set of curves over time, relative to each pixel, /_S,_R (i,j,t), wherein s represents the index of the source used, R is the specific response of the pixel receiving the light, (i, j) identifies the relevant pixel and t is the time. Thus, a raw recording typically appears as a video which may have several response channels, in the Red, the Green and the Blue, and several source channels, for example SI (first source, for example a red LED) , S2 (second source, for example a near infrared LED) and S3 (third source, for example ambient light) . The size of the video is that of the mxn matrix (size of the image) multiplied by the number of recorded images N equal to the acquisition time τ multiplied by the sampling frequency F_e.

There may be then compression of information. Indeed, the raw video signal is both cumbersome in terms of memory required for its storage, and especially unnecessary since the device does not operate in an imaging mode. Indeed, the plane of the sensor and the plane of the biological tissue are not conjugate planes in the sense of an optical system. An image is not therefore really seen on the sensor but simply intensity areas which correspond to light rays having passed through the finger, optionally the optical system (for example a lens, case of the configuration of Fig. 1C) or the protective layer (case of the configuration of Fig. IB) . Intensity areas are therefore obtained such as those illustrated in the right portion of Fig. 4 or in Figs. 11A and 11B.

Advantageously, an algorithm in the course of time selects areas with strong intensities, considered as signal carriers (i.e. this is masking) . For each instantaneous image of size mxn , a reduction algorithm only retains the distribution of the intensities, without taking into account their spatial distribution. This function is called a histogram, and it may be applied to each mode (defined by the channel/source pair) at instant t.

With this method it is possible to only retain information of the statistical type, not related to the position of the pixel but rather to the overall behavior of the population of light rays having probed the tissue.

Thus, for each measurement relative to an identified channel-source pair, the histogram is calculated on the population of the pixels on a basis of 32 elements, (or 64, 256,.... depending on the quality-storage space compromise which one intends to achieve) . A set of elements (which form a vector) at instant t, on the base k is thereby obtained, which is noted as h_trk(J).

This method amounts to count the pixels for which the value of the intensity of the instant t is between the values V± and V±+i . The values V± for i ranging from 1 to k form a regular interval extending from the minimum value which a pixel may assume (0) to the maximum value (256 if this is an 8-bit pixel) .

The values of each channel-source pair, which change over time are therefore grouped in h_1/k(j). This vector gives at instant t and on a base of resolution k the distribution of the pixels according to their intensities. Thus, at each intensity interval [j,j+l], the value of h indicates the number of pixels for which the intensity is comprised in this interval.

A histogram H which represents the overall intensity of the signal of each pixel may be used. But, when the source contains wavelengths for which blood has not the same absorption coefficient, it is also possible to only use the signals of the set of channels of same color, for example the red signals.

The function H, the red R and green B channels are represented for a given source at a given instant in Fig. 4. The corresponding « images » from which these histograms are formed are illustrated on the upper right of the figure (in fact, the 3 images are seen for the 3 sets of channels R, V, B) . These histograms may be normalized relatively to the reference values, the calculation method of which has been explained earlier.

During the acquisition, each of these histograms will change over time. This time dependent change, or dynamic change, represents the raw PPG signal and it may be characterized by different parameters of the histogram:

- the position of the maximum;

- and/or the position of the barycenter or of the isobarycenter of the curve. For this, the barycenter of the histogram is calculated: each interval is weighted by the number of pixels which it contains. A barycentric value (or weighted average) is then extracted which is noted as P_k(t) . This value is refined when the resolution of the histogram k increases; the coefficients of this barycenter result from the representativity of each channel/color in the contents of the information;

- and/or the average of the values of the red or green channel in order to obtain a PPG synchronous signal.

Alternatively, and notably when the shape of the histogram has a marked peak, the curve is fitted with a Gaussian function or a Lorentzian function, and the time dependent change of the maximum of this function is tracked. This curve is of the exp(-(x-p_k) /xo ) type wherein x₀ represents the width of the peak, p_k its center and a its amplitude. A value representative of the dynamics of the histogram over time P_k(t) is thereby obtained.

These various possible uses of the histogram or of the histograms, are particularly well adapted when the shape of the histogram is irregular or asymmetrical .

When the histogram is regular and symmetrical, it may be sufficient to consider the average of the intensity for the whole of the pixels: there again, this average will change over time and is then representative of the PPG signal.

In order to characterize the regularity or, on the contrary, the irregularity of the histogram, the average value and the standard deviation of the histogram are considered, as well as optionally the moments of order 2, 3 and other moments.

Generally, the consideration of the average may be limited in the following cases:

1. The matrix 2 of receivers has a substantially smaller size than that of the examined tissue. In this case, the image is quasi-uniform on the whole of the receivers. The histogram has a standard deviation of the intensities which allows it to be described as regular;

2. The configuration is that of Fig. 2C, and the tissue is located at, or close to, the object focal point defined by the optical means 14. In this case, a quasi-uniform illumination is obtained on the whole of the detectors. Generally, the exploitable measurements correspond to sources obtaining a spectral range in the red/near infrared (comprised for example between 600 nm and 1,100 nm) . For example, the signal may be obtained with a 950 nm LED source, as well as with a white source such as the sun or an incandescent source, emitting through the natural filter of the finger (Fig. 3A) .

It is therefore preferable to use a signal in this wavelength (near IR) range. But in certain cases it may occur that this signal is saturated. In this case, it is then attempted to exploit the signal which stems from another channel, the green channel.

An example of a method according to the invention, which groups steps as indicated above, will be given in connection with Fig. 16.

First, it is proceeded (step SI) with determination of the saturation of the selected channel or channels. If the red channel is saturated, a selection of the green channel (step S2) is made.

From the moment when the adequate channel is determined, the mask may be determined in the way which has been described above (step S3) . The signals which stem from pixels located inside the mask will allow formation of the PPG signal.

The average of the intensities of the whole of the signals of the retained channel (step S4) is then formed.

It is checked whether the histogram is regular (step S5) ; this regularity may be characterized in the way described above) . If there is actually some regularity, then it is possible to be content with the average which was calculated above, and the time-dependent change of which may make up the raw PPG signal (step S6) .

Otherwise (step S7), if the histogram has a minimum of irregularities, it is then proceeded in the way as indicated above : the raw signal is obtained by the time-dependent change of a parameter of the histogram, or its barycenter, or of a parameter of a curve which fits or models the histogram.

The PPG signal reflects dynamics of the wave which the heart sends into the arterioles/veins and arteries, and the latter is then reflected on the walls and impacted by the distension and reflection properties of the arteries. The source light probes at different instants these time-dependent changes in terms of absorbed or transmitted intensity, and the different properties are visible and analyzable on the PPG signal.

Therefore it is further possible to achieve filtering of the PPG signal.

Notably, once the dynamics of the histograms are obtained, certain non-plethysmographic participations of the signal may be removed by using several methods, whereof the first two shown below have been tested experimentally and give conclusive results:

• By standard linear filtering, the very low frequencies (< 0.4Hz) are removed, as well as the high frequencies above a frequency between 5 and 10Hz;

■ Or by an adequate time-scale breakdown.

By transforming the raw signal into wavelets the signal may be isolated from artifact and measurement noises on « levels » different from those where the signal is found;

^■ Or by a shape recognition algorithm based on a parent function consisting of two curves of the Gaussian or Lorentzian type, of different amplitudes, widths and positions to be fitted to the signal .

Figs. 5A-5C are the experimental proof of obtaining and extracting the optical plethysmographic signal .

The spectrobarycentric movement described above is observed in Fig. 5A, obtained from a channel source of the ambient light type (a red/infrared LED source gives a result comparable with different processing) . The signal is normalized relatively to the spectral reference of the color filters. Fig. 5B is obtained by extracting the raw signal of Fig. 5A, according to a breakdown into wavelets and recomposition . Finally, the graph of Fig. 5C shows the extent of the filtered signal in the frequency space, thereby giving the frequency ranges containing the information (the unit is the Hertz (Hz) and seconds for the PPG graphs) .

In a device and method according to the invention, the oxygen saturation parameter may be obtained simultaneously with the measurement of the PPG signal .

Blood consists of oxygenated and unoxygenated heamoglobin, noted as [HbO] and [Hb] respectively. The oxygen saturation noted as St(¾ measured by a source s at instant t is the oxygenation level of the blood which absorbs in majority the light from s: s = «o,= [Hb^~O^m] +^0][Hb]

Beer-Lambert's law stipulates, that in the case when the absorption phenomenon is predominant relatively to the diffusion, the optical density D of the absorbing elements i depends on their concentration C±, on their line absorption coefficient and on the length covered by the light ray undergoing absorption L. Thus, at a set wavelength λ, it is possible to write :

D(X) = ₁(X)c₁L + a₂(X)c₂L + ...

By therefore using two spectrally defined sources at λι and at ₂ which provide illumination in a range of light where both amounts [HbO] and [Hb] predominantly absorb with respective coefficients of cxHbo and for each source, one has:

,!¾ ~- D ;;A: ) :~ J:(.ft/_:r&_f;_;.(,V:)i; + (1 - ίίί5;ί_ί ¾ ; )ί£

As the optical densities Di and D_∑ are measurable by the incident flux on the pixel depending on its distance and orientation, both preceding equations allow the oxygen saturation in hemoglobin to be obtained provided that the wavelengths of the sources and the properties of the biological tissue in these ranges are known, information which is easily found in the literature. This resolution method may widely be used within the scope of the invention, but with a determining advantage. For each source-pixel pair (active area pixel) , an optical density D_s , i , _j may be defined wherein indicates the relevant pixel.

Instead of discretely measuring a single value of D, a whole population of density values is available, for which the histogram may be established and the dynamics may be recorded with the method described for PPG. A statistical position of the value D is thereby obtained on a base with k elements, noted as P_k,D.

In the case when the source used is not defined in a spectral range around a value λ, such as for example the ambient light of spectral power E (λ) , then a processing operation specific to this type of system architecture is invoked. This method (continuous average) consists of replacin a discrete spectral parameter _P( 0, with its .

Thus, with this same wide spectrum source as defined by E (λ) and the use of two channels with distinct spectral responses Rl (λ) and R2 (λ) , it is possible to obtain two values of optical densities. There remains to substitute the values of discrete absorptions with continuous values.

Therefore the following system is obtained:

(A? >ι

<∑>λ} > ··· ii_¾K>(Aj >·, -Hi - S.s <■: s¾¾ .A) >;i ;l-

This system forms a system equivalent to a standard system, and may therefore be solved. This requires the knowledge of the tabulated values of the absorbances for both types of hemoglobin over the whole of the spectrum of the source in order to establish the continuous averages thereof. The discrete or continuous optical densities may be calculated by using the reference measurements (before and after laying the tissue) since they are directly related to the transmittance of the biological tissue.

Fig. 6 is an illustration of a spectrum of ambient light through the finger of a hand (blue curve) which will be used for weighting the absorption values, on the Green and Red channels of standard pixels. The formula which was used for extracting the oxygen saturation is the following:

Obtaining the stabilized PPG signal over a time interval and repeated at a certain weekly or monthly frequency gives the possibility of a significant number of functionalities, due to the richness of the signal and to its quality.

The device may simultaneously deliver the wave form of the PPG signal and the oxygen saturation as functions of time. Both of these combined measurements deliver significant information on cardiovascular operation and health condition.

Fig. 7 illustrates the parameters which may be extracted from a typical PPG period. The health observables which are inferred therefrom are monitored and the history thereof is kept. A parameter related to the arterial stiffness S_a is the parameter ΔΤ, the time interval between the maximum of the two peaks and the one related to reflection in the arteries is b/a, b being the height of the diastolic peak and a that of the systolic peak.

More specifically, the recorded PPG signal over a given acquisition period may be used for notably extracting either one of the following parameters:

1. The typical period, which is representative of the average period over an acquisition time; it consists of a first systolic lobe S and of a second diastolic lobe D (sometimes occulted) (Fig. 8A) . For this the signal is broken down into two lobes, either by using cancellations of the second

,d²PPG ₀

derivative of the PPG ( ;— ) as markers of

dT²

inflection points, or by fitting with a set of two (either Gaussian or Lorentzian) curves

(ae^~(t~tl)2'^Tl + be^~(t~t2)2'^T2) , (ti, t₂) giving the positions of both peaks (systolic and diastolic peaks) ; a and b are the respective amplitudes of these peaks, and (ΤΊ, T₂) their respective widths;

2. The regularity of the occurrence of the periods and therefore of the heart rate. On an acquisition, it is either possible to achieve a frequency analysis by passing into Fourier space (by a Fourier transform) or to statistically study the distribution of the positions t± of the peaks in the course of time. The average At may thereby be inferred as well as the statistics on this parameter, i.e. its standard deviation σ . Thus it is possible to have a parameter available which increases with the irregularity of the heart beat. This overall parameter does not exempt from analyzing the distribution of the instants t±. Cardiac arrhythmia is often reflected on the signal by a defect or an irregularity on the occurrence of the lobes S and which returns periodically. Detection of arrhythmias may therefore result from a time analysis of the regularity, or from a transform of the elements t± in phase space (by a transform of the phase) . The statistics and the variability obtained on the whole of the values t± allow establishment of the daily or weekly cardiac variability, or over long periods. With this measurement, it is possible to access the resting heart frequency FC_rest which is difficult to obtain by methods other than by repeated measurements. This FC_rest is a strong health and sports ability indicator of the person. A low resting heart frequency, for a given age, is a good health indicator (for example, in beats per minute (BPM) , between 45 BPM, and 60 BPM for an amateur sportsman. If this frequency deviates upwards with age, cardiovascular health deteriorates. An overall index of the time-dependent change of this parameter may thereby be obtained over long periods, and observation of cardiovascular health may be maintained. 3. The time intervals ATi=ti+l-ti between the systolic and diastolic peaks which successively occur during the acquisition, are an element which reflects the condition of the vascular walls both of the coronary arteries (around the heart) and of the arterial tree between the heart and the tissue on which the measurement is conducted. The heart sends a bulk wave which is accompanied by an acoustic wave, and the time taken by the wave for covering the outgoing path and the return path depends on the propagation and rebound coefficient (reflection) in the arteries. Thus, the elasticity, the stiffness of the arteries and the surface condition of the inner walls (fat deposit or not) are reflected on the propagation velocity of the wave, and therefore on the average value of the interval ΔΤ (ΔΤ: average time separating both peaks) . Therefore it is possible to observe this internal time-dependent change in the arteries in the course of time, all the year round and from year to year, in order to prevent: a too large fat deposit, and/or increased stiffness due to addiction to smoking, and/or distensibility or any other alteration caused by another disease (diabetes for example) , or any other coronary or myocardial abnormality. A follow-up parameter is defined as indicated in the literature which is designated as arterial stiffness (S_a) and which is defined by the height of the patient divided by the interval AT :

Sa = h/AT

4. Simultaneous measurement of the oxygen saturation provides new information which completes that obtained from the PPG. A low saturation level, for example below 80%, is designated by the term of ischemia, and may confirm the presence of a blocked artery or an occlusion in the arterial tree. This type of vascular diseases and alterations may be found by the analysis described earlier of the PPG signal, such as an irregular perturbation of the form of the lobes, a sudden but regular disruption in time or further alteration of the arterial stiffness. But the measurement of blood oxygenation at the specific instants where this abnormality is revealed confirms these observations.

These specific analyses are those with which most current cardiovascular diseases may be prevented. But moreover, the PPG signal further contains rich informations which provide cardiac and cardio-respiratory analysis. The wave shape of the PPG reflects the arterial pulsation for a wavelength channel with strong penetration into the tissue (near infrared) while a smaller wavelength in visible red allows access to vein pulsation. This logic also applies to the measurement of the oxygen level in the veins, which is generally less than that of the arteries .

The slowly variable envelope of the PPG signal, formed by the height of the systolic peaks, shows a slow beat, with a period equal to a few seconds which correspond to the breathing rhythm. These pieces of cardio-respiratory information may allow the observation of rhythm abnormalities, which may be caused by premature ventricular contractions or tachycardia. Real-time saturation remains an observable which may reinforce these analyses.

Mathematically, it is possible to extract the slowly variable envelope of the PPG signal by various algorithms, notably a low-pass linear filter. Also certain parameters of the PPG signal may be added to this, such as:

- the regularity of the positioning of the first (systolic) peak;

- the first derivative and the second derivative of the PPG curve for identifying the positions of the peaks, the positions of the inflections and their statistical variabilities ;

- the area (integral) under the systolic peak calibrated by the saturation. The saturation gives the distribution of hemoglobin, either oxygenated or not, and thus indicates a correction factor on the amplitude of the PPG signal, if it is obtained with a single wavelength (non-calibrated PPG) ;

- the interval between the reference point

(beginning of the curve) and the first inflection,

- the reflection coefficient on the calibrated PPG (obtained with 2 channels or wavelengths, or normalized relatively to the saturation) which is the ratio between the amplitude of the diastolic peak and that of the systolic peak r = b/a.

The following configurations of a device applying a method according to the invention are given as an example. As understood from the explanations given above, the introduction of optical means 14, whether they form a complex system (in the case of a still camera objective) or are a simple lens, does not perturb the operation of the measurement device. Indeed, the composition and the sharpness of the image obtained on the matrix of pixels have little importance, only the statistical characteristics and the global dynamics are retained.

An embodiment is first presented where only ambient light is used with 3 output channels which are the Red/Green/Blue pixels of a standard commercial apparatus. This example is illustrated in Figs. 8A and 8B. Moreover it is optionally possible to apply one or more external sources attached to the finger or integral with the still camera, for example the auto-focus source.

The configuration applied is that of Fig. 1C, but the apparatus used is a computer camera 25 (or « web cam », integrated or connected through an USB cable to a computer) . It is at the surface of the objective of the camera which therefore forms the supporting area, where the user positions his/her finger. The viewing screen of the computer is seen in Fig. 8A. The latter may apply processing according to the invention of data of the signals from the matrix 2 of pixels.

Fig. 8B illustrates an image on the screen 28, this image groups the results of processing according to the invention : the PPG signal 32, three histograms R, V, B and an area 30 which in fact is the image as seen by the sensor and from which the histograms are calculated. Generally, in a device according to the invention, it is possible to achieve a display of time curves, and notably of the PPG signal (a raw but also processed signal) and/or of the saturation signal, and/or of one or more histograms and/or of digital values such as for example the heart beat .

The results displayed in Fig. 8A show that the ambient light/webcam acceptance pair is not always sufficient for the second cardiac lobe (diastolic lobe) to clearly appear relatively to the background noise. Moreover, the infrared filter which equips this type of camera may further attenuate it. Adequate signal processing may therefore be applied. For example, a shape recognition or filtering algorithm is applied with which it is possible to extract more efficiently the dynamics of the histograms, from the noises and parasitic signals. This algorithm may beneficially be used on other apparatuses.

Figs. 9A-9C illustrate the results obtained with this first embodiment. The signal of Fig. 9A is the one which reflects the dynamics of the population of pixels, with spectral normalization. The superposition of a noise at a higher frequency is observed there. This noise is suppressed during the shape recognition and the time scale breakdown of the signal (Fig. 9B) . A PPG signal of very good quality is therefore obtained. Fig. 9C illustrates the spectral contents of the processed signal, and for example allows determination of the average heart frequency during the acquisition.

Another embodiment, illustrated in Fig. 10, takes up the configuration of the invention again by using an optical system and ambient sources. The finger is laid or pressed down on the objective 40 of a mobile telephone 42, located in the rear portion 42' of the latter. This objective is therefore the supporting area 2 ' . The assembly is then oriented towards the most intense source of light, such as light from the sun or from a lamp. By means of an application loaded into the telephone, and a specific button provided for this purpose in the menu of the telephone and therefore available to the user on the front face of the latter, a video recording is thus started, which may last for a few tens of seconds. A video is recorded, and then processed separately, for example by specially programmed data processing means, arranged in the telephone. Several types of telephones were tested, including the IPhone of Apple or HTC which is provided with the « Android » system of Google as well as telephones from the Samsung range.

The configuration is then that of Fig. 1C, with an optical component 14 (the lens) located between the area where the finger may be laid (supporting area) and the detector.

Figs. 11A and 11B are typical examples of thereby obtained images: they have the shape of a halo of quasi-circular symmetry (due to that of the optical system) , the center of which is more intense and less turbulent than the edges of the image. Quasi no notable dynamics or variation are observable to the naked eye. The video is then reduced to histogram curves for each instant t, as already explained above.

As understood from both previous embodiments, it is possible to produce a software application applying a method according to the invention on a portable computer of the PC type or on a portable telephone. This software carries out processing of the video, allows display of the results obtained by the method according to the invention, storage of the measurements in memory storage means and delivery of a report on the time-dependent change of one or more cardiovascular parameters.

More generally, an electronic device, applying a method according to the invention, includes in addition to an imaging device for connecting data of the type described in connection with Figs. 1A-1C, data processing means, for example a processor programmed for applying data processing according to the invention, and display means for viewing of more representative curves of the PPG signal and/or one or more results calculated from these curves. Such a device is preferably of the portable or mobile or wireless type.

Given the considerabe reduction in the amount of data during video compression with the histogram algorithm, allowing the transition from about hundred Mb to about an amount of data comprised between 20 kb and 280 kb, it is also possible to achieve a centralized application on a single server and which uses PCs and mobile phones as measurement terminals. This is illustrated in Fig. 12. The data collected by each terminal 42, 52 (here: a wireless telephone and a computer) are reduced on this actual terminal by means of a method such as the one described above. The reduced data may be sent to a processing server 60, so as to be processed there in a robust way. At the same time, it is possible to benefit from overall statistical data on the whole of the users. Customized results from data processing operations carried out by the server may then be sent to each user and/or to an apparatus of a physician 70, for example via the Internet network 66. These data may be displayed on the terminal of the user and/or of the physician, in connection with the identification of the thereby monitored patient.

The platform or the server 60 mentioned above may include memory storage means 63 for the data relating to each user. Alternatively, these data may be stored in the memory of memory storage means of another server. Schematically, the server 60 also includes various components, such as a microprocessor 65 connected through a bus to a set of RAM memories 67 for storing data, and of ROM memories 69 in which program instructions may be stored. This system further includes a viewing device (not shown in Fig. 12), or screen, and peripheral means such as a keyboard and a mouse. It may further include means for interfacing with the Internet network 36.

Each user is equipped with the device according to the invention 42, 52. In one of the memory areas of the server 60, the data or program instructions are stored for applying a processing method according to the invention and as described above, and notably for, depending on the data received from the user 42, 52, establishing a histogram and inferring therefrom a signal of the PPG type and/or of oxygen saturation.

These data or instructions may be transferred into a memory area of the server from a diskette or from any other medium which may be read by a microcomputer or a computer (for example: hard disk, read-only memory (ROM) , dynamic random access memory (DRAM) or any other type of RAM memory, a compact optical disk, a magnetic, electric or optical storage item) .

If the processing takes place in the individual device 42 of a user, it is in a memory area of the apparatus of this user where the data or program instructions are stored for applying a processing method according to the invention and as described above, and notably for, depending on the data measured by this user, establishing a histogram and inferring therefrom a signal of the PPG type and/or of oxygen saturation. These processed data may be displayed on the terminal of the user and/or transmitted to the physician of this user.

Figs. 13A and 13B represent another device according to the invention.

The base configuration is that of Fig. IB. The matrix sensor 1 is extracted from a commercial webcam : this is a color CCD imager on which the finger is directly laid. A LED source 2 with an infrared wavelength (950 nm) was soldered on the substrate 7 used and points above the imager. This source may be used in the absence of a sufficient ambient source. Indeed, it is possible to obtain a good quality signal by using ambient light. The absence of an infrared filter between the tissue 10 and the sensor (which is currently found in still cameras) allows more intense passage of the light, even weak to the naked eye and therefore allows the PPG signal to be obtained.

This experimental device is connected to a computer through a USB port.

Another simple exemplary embodiment of a device according to the invention is the following:

- the matrix of pixels is connected to electronic data processing means,

- the illumination is that of ambient light ;

- only the signal of the red channel of each detector is selected;

- the average of the whole of these channels in the red is performed.

The matrix of pixels is associated or connected with for example a microcomputer or with the electronic means of a wireless telephone, which carry out all the calculations and processing operations.

Figs. 14A-14D are various signals obtained by a method according to the invention.

Once the video is reduced or compressed by the algorithm, the signal illustrated on the curve of Fig. 14A is typically obtained. Lower frequency sampling (less points per period) may be noted, due to the video compression applied by the system during the recording of a not very dynamic sequence. By controlling the recording system, it would be possible to improve the signal. The curve of Fig. 14B shows as an example the influence of the processing, which is reduced in this case to filtering and localization of the position of the peaks by the wavelet method. Both figures 14C and 14D are respectively an analysis of the frequency content (and therefore give the average heart frequency in beats per minute: the peak of the frequency content gives a position in Hz which corresponds to the heart frequency) and an analysis of the time content (regularity of the heart beats) .

Figs. 15A-15C are also various signals obtained with a method according to the invention.

Fig. 15A shows the aspect of the signal extracted from a recorded and then reduced video. A similar result is obtained with a real time application. The curve includes a pseudo-periodic signal (PPG) contained in a slowly variable envelope. This slowly variable envelope is related to breathing, and with its information it is possible to further extend cardio-respiratory analysis. Processing specific to this channel-source is applied to this signal in order to obtain the refined PPG curve (Fig. 15B) . The curve of Fig. 15C shows the spectral contents of the PPG signal and the average heart frequency may be inferred therefrom.

Claims

1. A system (1) for detecting signals of the PPG type from a biological tissue, including:

a) a supporting area (2') for a portion of the tissue,

b) a matrix (2) of nxm elementary optical receivers, and means for forming an output signal of each elementary receiver, depending on radiation which it receives from a tissue bearing upon the supporting area and as a function of time,

c) means for forming a PPG signal from the output signals of the receivers of the matrix, including means for forming a histogram of the intensities of the output signals of the receivers of the matrix and means for identifying the time-dependent change of a histogram.

2. The system according to claim 1, further including at least one radiation source (4, 6) .

3. The system according to claim 2, said radiation source (4, 6) not being interdependent with the system.

4. The system according to any of claims 1 to 3, the supporting area (2') being defined by:

- the whole of the surfaces of the receivers intended to receive illumination, - or the surface of a transparent layer positioned on the whole of the surfaces of the receivers intended to receive illumination,

- or the surface of an optical device positioned on the whole of the surfaces of the receivers intended to receive illumination.

5. The system according to any of claims 1 to 4, including means for forming, as a function of time, a mask (20) defining at the surface of the matrix, an area of receivers receiving radiation having passed through a portion of a tissue bearing upon the supporting area.

6. The system according to any of claims 1 to 5, the means for forming a PPG signal including means for calculating the average of the intensities from the receivers.

7. The system according to any of claims 1 to 6, the time-dependent change of a histogram being defined by the time-dependent change of the position of the maximum and/or of the position of the barycenter or of the isobarycenter of the histogram and/or of the peak of a function fitting the histogram.

8. The system according to any of claims 1 to 7, including means for filtering the time-dependent change of a histogram in order to remove non-PPG contributions therefrom.

9. The system according to any of claims 1 to 8, including means for determining whether the histogram is regular or not or in a saturation situation .

10. The system according to any of claims 1 to 9, including means for calculating the oxygen saturation level.

11. The system according to any of claims 1 to 10, including means for extracting the slowly variable envelope of the PPG signal.

12. The system according to any of claims 1 to 11, including means for forming a representative signal of the period including a systolic lobe (S) and a diastolic lobe (D) and/or of the regularity of the heart rate and/or of the resting heart frequency and/or of the time intervals between a peak of a systolic lobe and a peak of a diastolic peak.

13. The system according to any of claims 1 to 12, including a wireless telephone which includes at least the means a) and b) .

14. The system according to claim 13, the wireless telephone further including at least one portion of the means c) .

15. The system according to any of claims 1 to 12, including a video camera which includes at least the means a) and b) , and a computer.

16. The system according to claim 15, the computer further including at least one portion of the means c) .

17. The system according to any of claims 13 to 16, further including a server (60) which includes at least one portion of the means c) .

18. The system according to any of claims 1 to 17, further including at least one infrared filter for filtering radiation received from the tissue bearing upon the supporting area.

19. The system according to claim 15, further including means for applying a shape recognition processing operation.

20. The system according to any of claims 1 to 19, further including means for determining an output signal from each elementary receiver in a channel of a particular color (R, G, B) .