The present invention relates to a procedure as defined in the preamble of claim 1 and to a measuring apparatus as defined in the preamble of claim 29 for non-invasive determination of fractional oxygen satura¬ tion in blood. Moreover, the invention relates to a sen¬ sor as defined in the preamble of claim 20, designed for use in conjunction with tne measuring apparatus of the invention to collect measurement data about the patient. Specifically, the present invention relates to the monitoring of the oxygenation level of the body in patient monitoring systems. Measuring the oxygen satura¬ tion of arterial biood IP peripheral circulation is ge- nerally sufficient to determine the oxygenation situati¬ on and sufficiency of oxygen supply in the entire body. The oxygenation level of tne human body can be estimated via oxygen saturation measurement of arterial blood eit¬ her in a non-invasive manner using pulse oximeters or transcutaneous oximeters/blood gas analysers or in an invasive manner either by taking a sample of arterial blood and analysing In Vitro blood gases (In Vitro Blood Gas/pH Analyzers) or performing an optic measurement on the blood sample using so-called CO-oximeters or haemo- ximeters (In Vitro Multiwavelength Oximeters) .

Partial pressure measurements of gas in arteri¬ al blood samples and optic methods based on the ab¬ sorption of lignt by blood samples are part of long¬ standing tradition, but clinical use of pulse oximeters only became common in late 1980' s and the measuring principle itself is relatively new. There are numerous patents and patent applications relating to pulse oxime¬ ters. The most important of these as well as the most comprehensive general descriptions of prior art are found in patent specifications US 4,653,498, US 4,819,752, US 4,407,290 and US 4,832,484.

The prior-art technology described in the abo¬ ve-mentioned patent specifications, which is the basis of currently used equipment, is imperfect and inadequate for continuous and non-invasive monitoring of changes in the actual oxygenation level or degree of fractional oxygen saturation in a patient's blood. Although In Vit¬ ro oximeters are in principle capable of measuring frac¬ tional oxygen saturation from a normal blood sample, the measurement is neither non-invasive nor continuous. On the other hand, pulse oximeters measure continuously and non-mvasively, but they are not able to measure the ac¬ tual degree of fractional oxygen saturation of blood and are therefore inadequate for situations where only a part of the total amount of haemoglobin in a patient is functional. Pulse oximeters measure fractional oxygen saturation assuming that the patient's blood composition is the same as that of a healthy, non-smoking person. A high dyshaemoglobin level, i.e. a high relative amount of haemoglobin not participating in oxygen transport, always involves a danger to the patient because current pulse oximeters produce an incorrect estimate of the oxygenation level of blood.

The cause of incorrect measurement lies in the measuring principle: Since pulse oximeters use only two different wavelengths of light for the estimation of oxygen saturation, only two different kinds of blood ha¬ emoglobin, viz. oxyhaemoglobm (Hb02) and deoxyhaemoglo- bm (Hb) , can be accurately measured by this method. All other dyeing blood components (usually dyshaemoglobms or dyes used in clinical tests) have a disturbing effect on the measurement and can only be taken into account as average amounts. This type of average correction is ge¬ nerally made on the composition of healthy blood. Howe¬ ver, the composition of normal blood may change in an unforeseen manner and without a readily identifiable cause. The blood composition of a patient with a criti¬ cal illness may differ from the blood composition of a

healthy person as a result of medication, the nature of the illness or a medical treatment or measurement. A new and significant treatment of this type is the so-called nitrogen oxide (NO) treatment, which may cause a consi- derable rise in the patient's methaemoglobin (MetHb) le¬ vel. Another common case of incorrect measurement is carbon monoxide poisoning, which involves a high carbo- xyhaemoglobm (HbCO) level in the patient. Continuous non-invasive monitoring of the actual degree of oxygen saturation is particularly important during NO treatment because the dyshaemoglobm levels may rise relatively rapidly, which means that an analysis based on a blood sample is not sufficient. The measurement of fractional oxygen saturation is also of great importance in rescue operations and in follow-up monitoring after carboxyha- emoglobm poisoning.

It is obvious that accurate, continuous and non-invasive measurement of fractional oxygen saturation requires a sensor with several wavelengths used to pro- duce an analysis of blood composition. In the following, prior art will be discussed by considering a technique that uses the principle of non-mvasive measurement using an oximeter with several wavelengths.

A previously known oximeter based on non- invasive measurement uses eight different wavelengths to determine the average degree of oxygenation of the blood via a measurement on the ear (see Girsh et Girsh, Ann. Allergy, 42, pages 14-18, 1979) . The measurement does not use the pulse oximeter principle, whereby the measu- rement is only applied to arterial blood by distinguis¬ hing from the light transmission a component pulsating in synchronism with the heartbeat and normalising this component against the total light transmission. Instead, average oxygen saturation is measured directly from the total light transmission at the wavelengths used. The total transmission depends on the oxygen saturation and composition of both arterial and vein blood, but also on

the absorption and scattering caused by other tissues. Typically, blood accounts for only 1-2 % of the amount of tissue, so the signal may be very ambiguous. Such a method has many drawbacks: First, the person's comple- xion, the structure of the tissue in itself and especi¬ ally the scattering and absorption properties of the tissue as well as its other properties change and even dominate the total transmission. In fact, the method re¬ quires several wavelengths for the compensation of these properties, and it cannot produce reliable analyses of the composition of arterial blood. In addition, analy¬ sing the blood composition in terms of percentages is difficult because the relative amounts of arterial blood and vein blood and their different degrees of oxygen sa- turation affect the absorption. The amount of dysha- emoglobins is the same m both arteries and veins, but oxygen saturation varies with tissue metabolism and tem¬ perature or with the regulating mechanisms of the body.

In patents EP 335357 and US 5,421,329 it is suggested that by adding a third wavelength to a conven¬ tional pulse oximeter with two wavelengths it is possi¬ ble to improve the accuracy of functional saturation me¬ asurement with a pulse oximeter or to eliminate or redu¬ ce the artefacts caused e.g. by movement. In the former patent, the third wavelength is used to eliminate the irregular artefact signal from on top of the pulsation caused by the heartbeat. The wavelength is not used for the determination or identification of the dyshaemoglo- bm level. In the latter patent, the third wavelength is used to adjust the measurement of a low degree of oxygen saturation, but it is not used in conjunction with the measurement of the normal saturation range or for the measurement of dyshaemoglobin levels or the degree of fractional oxygen saturation. The latter patent also re- lates to the reflection principle and especially to the measurement of oxygen saturation in a baby during child¬ birth. In this situation, a third wavelength is natural-

ly needed to achieve a more reliable measurement. Simi¬ larly, patent application WO 94/03102 proposes the use of a third wavelength to eliminate artefacts caused by motion. Patent specification US 471434 (Minolta Camera) also uses a third wavelength for more accurate measure¬ ment of functional saturation. Like the others, this specification is not concerned with the measurement of fractional saturation or in general dyshaemoglobin le¬ vels. Patent specification EP 0 524 083 also proposes the use of a third wavelength for simultaneous measure¬ ment of carboxyhaemoglobin level and oxygen saturation. In the measurement, three different laser diodes with wavelengths of 660 nm, 750 nm and 960 nm are used. These three wavelengths are used to measure the modulation ra¬ tios, and the concentrations of three unknown kinds of haemoglobin, Hb02, Hb and HbCO, are calculated by sol¬ ving a linear system of equations. However, the procedu¬ re presented in patent specification EP 0 524 083 has two significant drawbacks. First, the procedure is not applicable for the measurement of MetHb; in other words, fractional saturation can only be determined for the three kinds of Hb mentioned above. Secondly, solving the aforesaid linear system of equations is not sufficient for the determination of the concentrations of the afo¬ resaid three kinds of Hb, as will become evident later on from the description of a preferred embodiment of the calculating procedure of the present invention. The ba¬ sic drawback is that the system of equations is not a linear one because the coefficients used in it are in themselves functions of the concentrations. For this reason, the use of this method is restricted to a very narrow range of oxygen saturation, and the procedure is not workable in the operating range generally required for pulse oximeters. In addition to the above drawbacks, the procedure involves the use of laser diodes and a fibre optic connection to the measurement point, which

makes the measuring system rather too expensive for practical measurements and difficult for the user. Mo¬ reover, laser diodes have a narrow choice of wave¬ lengths. Due to the use of fibre, the light is attenua- ted especially at the connection points and the signal- to-noise ratio is worse than in the conventional soluti¬ on employing light-emitting diodes. The official regula¬ tions relating to coherent radiation and the danger caused by the radiation e.g. to the eye also constitute a limitation of the application of the procedure in practical situations.

Further, patent specification Aoyagi et al, EP 0 679 890 Al, representing prior art, presents an appa¬ ratus designed for the measurement of light aosorbing blood components. According to the specification, the procedure and apparatus can be used to determine the de¬ gree of functional oxygen saturation of blood, the con¬ centrations of different kinds of haemoglobin as well as other dye components of blood, such as bilirubm and m- vein dyes. The proposed procedure and apparatus are based on a rather unusual optical model of light transmission through tissue and formation of a pulsating signal. Since the procedure is obviously one of the pri¬ or-art solutions related to the present invention, it is necessary to point out the erroneous assumptions lying behind the procedure and apparatus. The drawbacks listed below serve as examples, and the drawbacks are not described in full extent. For a more detailed explanati¬ on of the drawbacks, reference is made to the thesis Reindert Graaff, "Tissue Optics Applied to Reflectance Pulse Oximetry", Groningen University, 1.12.1993, which is an excellent description of tissue optics and its mo¬ dern representation. To those familiar with pulse oxi¬ metry or non-invasive measurement of blood properties, the drawbacks listed below are self-evident and can be recognised via empirical measurements. Accordingly, pa¬ tent application EP 0 679 890 is based on the following

erroneous propositions. First, diffusion approximation and its parametrised flux models (in the application re¬ ferred to, the so-called Arthur Schuster theory) can be applied to describe the total transmission through tis- sue, but they cannot be used in conjunction with pulsa¬ ting tissue components and the operation of pulse oxime¬ ters at short wavelengths (600-700 nm) , nor can they ge¬ nerally be used at a low saturation or for highly ab¬ sorptive blood dye components. Using the diffusion model (applies to situations where the scattering target area is considerably larger than the absorption target area) together with the Lambert-Beer pulse oximeter model (applies to situations where the absorption target area is considerably larger than the scattering target area) simultaneously generally does not lead to realistic re¬ sults. Second, it is stated in the patent application that the scattering term is known and independent of the wavelength. In fact, the scattering term is one of the adjustable parameters in the model and is also dependent on the wavelength and the tissue type. The pulsating portion of the scattering is also dependent on the siz'e, shape and number of blood cells, i.e. on the haemato- crit. Third, the tissue term (in the patent application, the pulsating component that is not blood) plays no sig- nificant role in the signal formation at all and there¬ fore it cannot be used in the way it appears in the for¬ mulas as a factor representing theoretical extinction coefficients of blood and empirically measured modulati¬ on ratios. Further, it is stated in the patent applica- tion that the tissue term is mostly water, which in fact does not absorb at all in relation to the dominating pulsating terms in the wavelength range used. In fact, the dominating pulsating tissue-type effect is produced by vein blood, whose absorption depends on all those things that patent application EP 0 679 890 presents as quantities to be measured, such as oxygen saturation, different haemoglobin varieties and their amounts and

dyes; thus, there would be no linear correlation between different tissue terms that could be defined in advance - although the application asserts to the contrary - which means that the degree of non-linearity of the problem increases considerably. Fifth, the number of unknown quantities in the procedure and apparatus of the patent application in question clearly exceeds the num¬ ber of equations available. Moreover, the patent appli¬ cation contains numerous other inaccuracies, so the app- lication or the procedure and apparatus presented in it do not, at least in respect of their basic assumptions, meet the quality criteria that are expected to be obser¬ ved in clinical patient monitoring measurements.

In the above, existing prior art has been dealt with from the point of view of systems using more than two different light sources each having a different spectral emission, yet so that the spectral emission in the same light source is always the same. Especially the use of more than two light sources is still associated with problems relating to maintaining the accuracy of the apparatus in use even in situations where the spec¬ tral emission of the light sources changes due to tech¬ nical aspects of fabrication of the light source and maintaining accuracy requires a correction to compensate this change. A method for such correction or rather a cheap method for maintaining sensor accuracy is presen¬ ted in US patents 4,621,643 (11/1986), 4,700,708 (10/1987) and 4,770,179 (9/1988) . All these patents pro¬ pose solutions in which information about the wave- lengths of the light sources is transmitted to the mea¬ suring apparatus by encoding the correction required by changes of wavelength into an impedance element or in practice into the resistance value of a resistor. From the resistance value or by some other similar coding method, the measuring apparatus receives information in¬ dicating the changes required in the calibration of the apparatus. This can be done with a single resistance va-

lue or other simple ^coding' when the unambiguity of the measurement signal is guaranteed via other techniques. In two-wavelength pulse oximeters, unambiguity is based on the apparatus forming substantially only one signal, i.e. a modulation ratio between the two wavelengths, a so-called R-value, which, via an unambiguous calibration curve, can be directly associated with the functional oxygen saturation or Sp02 value. No such unambiguous correlation exists when there are several light sources and more than two haemoglobin varieties or other blood dye components are to be measured. The present invention proposes a different calibration method, which is unam¬ biguous and is not - unlike prior-art technology - based on transmitting the emission properties of the light sources to a measuring apparatus but is instead based on light source-specific use of the absorption properties of each haemoglobin component or dye, i.e. the so-called extinction coefficients of blood. This allows e.g. situ¬ ations where the wavelengths of the light sources or the light source type itself can be changed or the sensor can be aligned for different kinds of Hb or dye measure¬ ments while preserving the compatibility and accuracy of the sensor with all apparatus. Thus, in the framework of the present invention, the "wavelength" of the light source is understood in the first place as meaning the current number or other identification of the light source, which is associated with the haemoglobin compo¬ nents or dyes measured with the sensor in question. The new method is applicable both for sensors and apparatus with two light sources and for those with more than two light sources.

As a summary of prior art, it can be stated that so far there is no method or apparatus capable of reliable measurement of fractional oxygen saturation of arterial blood and quantitative determination of the dyshaemoglobin level.

The object of the present invention is to eli¬ minate the problems and inaccuracies described above. A specific object of the present invention is to produce an effective and accurate measuring procedure for the determination of the relative concentrations or composi¬ tions of haemoglobin derivatives or dye components con¬ tained in a patient's blood. A further object of the present invention is to produce a measuring apparatus and a sensor which can be utilised to effectively apply the calculation method of the present invention for de¬ termining the level of oxygen saturation in a patient' s blood.

As for the features characteristic of the pre¬ sent invention, reference is made to the claims. In the procedure of the invention for determi¬ ning the relative concentration or composition of diffe¬ rent kinds of haemoglobin contained in blood, such as oxyhaemoglobm, deoxyhaemoglobm and dyshaemoglobins, and/or dye components, such as various m-vein dyes or the like, in a non-mvasive manner using the light ab¬ sorption caused by different haemoglobin varieties and/or dye components, light signals are transmitted at at least two predetermined wavelengths to a tissue com¬ prised in the patient's blood circulation, a light sig- nal transmitted through the target under measurement and/or reflected from it is received and the proportion of the intensity of the pulsating light signal received at each wavelength is determined in relation to the to¬ tal intensity of the light transmitted through the tis- sue or reflected from the tissue. The pulsation of the light signal is determined by the heartbeat frequency, which has a direct effect on the amount of blood flowing in the tissue and therefore also on the amount of ha¬ emoglobin derivatives and/or dye components. According to the invention, the effective ex¬ tinction coefficients of blood haemoglobin derivatives and/or dye components in the tissue are determined for

each light signal and/or light signal pair via a mathe¬ matical transformation from the extinction coefficients of the blood dye components according to the Lambert- Beer theory and the proportion of specific blood ha¬ emoglobin derivatives and/or dye components in relation to the total amount of haemoglobin contained in the blood is determined by means of the intensity of the signals received at different wavelengths. Thus, the procedure of the invention is based on a so-called modu¬ lation signal for each wavelength and on a comparison of these signals between two different wavelengths. The re¬ sult of the latter comparison is expressed in terms of a modulation ratio. This quantity describes the average blood dye difference between these two wavelengths. When this relative dye difference is measured using several wavelength pairs, the concentrations of different ha¬ emoglobin derivatives are obtained by solving a non¬ linear system of equations (1) formed from the wave¬ length pairs and extinction coefficients.

where % mod i is the modulation percentage for light transmission as measured at wavelength i, i.e. the proportion of light transmission varying at heartbeat frequency as a percentage of the total light transmissi¬ on;

C is a constant; the ij-element of the ε-matrix is an empirical extinction coefficient of blood haemoglobin variety and/or dye component HbX-, mathematically derived from the known extinction coefficient for wavelength I; and the unknown kinds of blood haemoglobin and/or dye components in percentages are placed in the vertical vector (HbXj _. , HbX HbX, The non-linearity of the system of equations is due to divergences between theory

and practice. The actual extinction coefficients ε ^s are also dependent on the scattering of light caused by the tissue and on the combined effect of absorption and scattering. The corrections needed in the extinction coefficients are the larger the larger is the proportion of the attenuation caused by absorption and scattering. In the Lambert-Beer theory, the scattering and the ef¬ fect of the tissue are not taken into account.

The system of equations presented above can be solved in several advantageous ways according to the present invention. In an embodiment, the system of equa¬ tions is solved for all blood dye components, the sum of whose proportions is 100%, and the number of independent light signals is selected so that it corresponds at le- ast to the total number of haemoglobin derivatives and/or dye components set as unknowns. Moreover, it is to be noted that blood may also contain haemoglobin de¬ rivatives or dye components whose concentration and/or composition is known. In this case, the known concentra- tions must be taken into account when calculating the total number of all haemoglobin varieties and/or dye components .

In an embodiment of the present invention, the haemoglobin varieties set as unknowns are oxyhaemoglobm and deoxyhaemoglobin, and at least one dye component is a blood dyshaemoglobin variety, such as HbCO, MetHb or HbNO.

In an embodiment of the present invention, the non-linear system of equations is solved by using modu- lation ratios, wherein the transformation between the known extinction coefficients according to the Lambert- Beer theory and the effective extinction coefficient of the blood-containing tissue is a function transformation between the measured modulation ratio of two independent light signals and the modulation ratio formed from cor¬ responding known extinction coefficients.

In an embodiment of the present invention, the non-linear system of equations is solved by dividing the non-linear system of equations into linear portions on the basis of blood composition and around a given compo- sition and solving the linear system of equations thus obtained using experimentally determined extinction coefficients derived for the blood composition in ques¬ tion. The extinction coefficients are preferably deter¬ mined in advance for different blood oxygenation levels. The non-linear system of equations thus reverts into a linear system of equations and it can be easily solved by known mathematical methods.

In a preferred embodiment, the non-linear sys¬ tem of equations is solved via an iterative process, part of which process consists in identifying the compo¬ sition and/or existence of the dyshaemoglobin and/or dye that leads to the best iterative result. Iteration is a known mathematical method and is therefore not explained here in detail. Further, in the determination of the concentra¬ tions of haemoglobin varieties and/or dye components, it is preferable to weight different modulation ratios on the basis of the patient's blood composition, in which case the non-linear system of equations (1) is solved iteratively by using weighted modulation ratios.

In a preferred embodiment of the present inven¬ tion, a light signal is transmitted via the same or nearly the same optic route to the tissue using at least two predetermined wavelengths. As the light signal pas- ses through the tissue at substantially the same point, errors due to divergences in the tissue are avoided.

The selection of wavelength ranges is explained below in greater detail by referring to Fig. 1, but in a preferred embodiment the wavelengths to be used are se- lected from four different wavelength ranges, of which at least one wavelength range is below 660 nm, prefera¬ bly so that the wavelength ranges are: 620 - 650 nm, 655

- 665 nm, 680 - 750 nm and 790 -1000 nm. The selection of wavelength ranges is further affected by the composi¬ tion of human blood so that for a normal blood composi¬ tion where the total amount of dyshaemoglobin is below a suitable level, e.g. below 3 %, light signals produced at medium wavelengths of 660 nm, 690 nm and 900 nm and/or modulation ratios formed from them are weighted in relation to the light signal in the wavelength range of 620-650 nm and its modulation ratios. For the measu- rement of a high methaemoglobm level, the wavelength range of 620-650 nm and its modulation ratios are used, which are weighted in substantially the same proportion as the other light signals used and their modulation ra¬ tios. In addition, for a normal composition of human blood, the measurement can be performed at four medium wavelengths 900 ± 10 nm, 690 ± 5 nm, 658 ± 5 nm and 632 ± 5 or 900 ± 10 nm, 690 ± 5 nm, 658 ± 5 nm and 645 ± 5 nm.

In a preferred embodiment of the present mven- tion, for each independent light signal an effective ex¬ tinction coefficient is calculated in the light signal emission range concerned from wavelength correlations of known extinction curves for blood haemoglobin varieties and/or dye components. Further, an effective extinction coefficient in the tissue is determined via the same mathematical transformation for all small spectral chan¬ ges in a given light signal.

In a preferred embodiment of the present inven¬ tion, changes in the spectral emission of a light source operating around a given medium wavelength, i.e. varia¬ tions in the wavelength of the light source are taken into account by utilising the changes in the effective extinction coefficient consistent with the Lambert-Beer theory, by determining the modulation ratio between two light signals for the effective modulation ratio measu¬ red in the tissue and the effective modulation ratio ac¬ cording to the Lambert-Beer theory and determining the

concentrations of haemoglobin varieties with the aid of the corrected effective extinction coefficients and the effective modulation ratio.

In a preferred embodiment of the present inven- tion, the measurement is calibrated separately for each sensor by storing the determined effective extinction coefficients consistent with the Lambert-Beer theory and/or corrected to correspond to the tissue separately for each light source, preferably for each light ele- ment, in a storage device provided in the sensor so that they can be used in the determination of haemoglobin va¬ rieties and/or dye components. Furthermore, the function transformations for each light signal pair and/or the medium wavelengths are stored in a storage device in the sensor, to be used in the determination of haemoglobin varieties and/or dye components.

In a preferred embodiment of the present inven¬ tion, to determine a given dyshaemoglobin variety, the proportion of other dyshaemoglobin varieties is set as a constant and the variety to be measured is set as an unknown. In addition, the proportion of other varieties is set to a value corresponding to an invasively measu¬ red proportion.

The invention also relates to a sensor for the collection of measurement data through tissue comprised in a patient's blood circulation in non-invasive measu¬ rement, said sensor comprising means for connecting the sensor to a measuring apparatus, a light source which transmits a light signal at at least two predetermined medium wavelengths and a receiver disposed to receive a light signal transmitted through and/or reflected from the target under measurement. According to the inventi¬ on, the sensor comprises a storage device for the stora¬ ge of predetermined sensor-specific data, where the sen- sor-specific data comprises an effective extinction coefficient for each haemoglobin variety and/or dye com¬ ponent desired, said effective extinction coefficient

being characteristic of each light source used. The sto¬ rage device may be a programmable read-only memory whose contents can be electrically deleted or altered, or a similar memory circuit. It is further preferable that in the storage device are stored the function transformations and/or medium wavelengths for each light signal pair and/or an identifier representing these and/or other corresponding information that indicates the connection between ex- tinction according to the Lambert-Beer theory and ex¬ tinction measured in tissue, to be used for the determi¬ nation of haemoglobin varieties and/or dye components.

In a preferred embodiment of the sensor of the present invention, the light source comprises a set of light elements in which the wavelengths to be used, of which at least one is below 660 nm, have been selected from four wavelength ranges as follows: 620 - 650 nm, 655 - 665 nm, 680 - 750 nm and 790 -1000 nm. Further, the light elements are thermally anchored on the sensor frame to keep the temperature of the light elements be¬ low a specific limit. Moreover, to keep the temperature of the sensor part to be connected to the patient below a specific limit, the sensor comprises a first optic fibre for passing the emitted light to the target under measurement and a second optic fibre for passing the received light to the receiver. In this way, the ele¬ ments, which tend to get warmed up, can be kept at a distance from the patient.

In a preferred embodiment of the present mven- tion, the sensor comprises a first set of light elements arranged to emit light to a first target and a second set of light elements arranged to emit light to a second target. Further, the first and second sets of light ele¬ ments have one common wavelength range, which is used for the compensation of variations due to the measuring point. In a preferred case, the measurement signals obtained in the common wavelength range are compared

with each other and, based on this comparison, divergen¬ ces caused by the tissue between measuring points are adjusted.

The sensor may also preferably comprise a set of light filters disposed in conjunction with the recei¬ ver to divide the received light into different wave¬ length ranges. In this manner, light can be transmitted in a wide range of wavelengths, preferably in the range of 600 - 1000 nm, and the received signal can be filte- red into desired wavelength bands.

The sensor may further comprise fixing means for attaching the sensor to the patient, preferably to the patient's ear or finger. It is also possible to at¬ tach the sensor to other limbs or organs of the patient. The present invention also relates to a measu¬ ring apparatus for the determination of the relative concentrations or compositions of different kinds of ha¬ emoglobin contained in blood, such as oxyhaemoglobin, deoxyhaemoglobin and dyshaemoglobms, and/or dye compo- nents, in a non-invasive manner using the light ab¬ sorption caused by different kinds of haemoglobin and/or dye components. The measuring apparatus of the invention comprises a sensor as described above and a signal pro¬ cessing device for the processing of the signals recei- ved. The signal processing device may be a computer, a microprocessor or an application-specific integrated circuit (ASIC) or the like. According to the invention, the apparatus comprises a calculating device and a rea¬ der device connected to the calculating device and sen- sor for reading the data stored in the sensor and transmitting the data to the calculating device. The reader and calculating device may be any electrically controlled component or application known in itself, and it can preferably be incorporated in the signal proces- sing equipment or in the same assemblage with said equipment.

In a preferred embodiment of the present inven¬ tion, the apparatus is arranged to measure at different wavelengths in accordance with a predetermined time di¬ vision principle in such manner that the measurement of the shares of certain dye components or haemoglobin va¬ rieties is weighted with respect to time so that in a given period of time the apparatus is only measuring in a given part of the entire wavelength range in use. The time division can be effected e.g. using a suitable channel arrangement and channelling device by selecting a given channel at a time for measurement.

In an embodiment the measuring apparatus of the present invention, the data stored in the sensor inclu¬ des specific wavelength values for special blood compo- sitions of patients, essential spectral emission infor¬ mation about the light sources used in the sensor, pre¬ ferably light elements, with respect to different dye components and/or haemoglobin varieties, and/or informa¬ tion about the extinction coefficients according to the Lambert-Beer theory and about the mathematical transfor¬ mation between these and the extinction coefficients for the tissue. Further, the data stored in the sensor in¬ cludes information describing the sensor type. Moreover, the measuring apparatus may comprise an identifier devi- ce for the identification of sensor type. The identifier device may also be integrated with the reader device.

As compared with prior art, the present inven¬ tion has the advantage that using the procedure and ap¬ paratus of the invention it is possible to eliminate the problems described above relating to prior-art equip¬ ment, and above all the invention makes it possible to eliminate the drawbacks and inaccuracies of prior-art methods and apparatus.

Furthermore, the invention presents a new type of calculating method that takes into account the inac¬ curacies resulting from the scattering and absorption

caused by tissue. In addition, the procedure allows com¬ pensation of small variations in LED wavelengths.

In the following, the invention is described by the aid of a few example embodiments by referring to the attached drawing, in which

Fig. 1 presents the extinction coefficients of different haemoglobin varieties as functions of wave¬ length;

Fig. 2a-2c represents a preferred wavelength selection according to the present invention;

Fig. 3 is a diagram representing a sensor ac¬ cording to the present invention;

Fig. 4 is a diagram representing a preferred arrangement for mounting a light element according to the present invention;

Fig. 5 is a diagram representing a sensor ac¬ cording to the present invention;

Fig. 6 is a diagram representing a measuring apparatus according to the present invention; Fig. 7 illustrates the correlation of theoretic and empiric oxygen saturation as functions of the ratio (R/IR) of modulation percentages determined at wave¬ lengths 660 nm (R) and 940 nm (IR) .

Fig. 1 shows the extinction coefficients of different haemoglobin varieties as functions of wave¬ length. Referring to Fig. 1, certain reasons lying be¬ hind the choice of wavelength according to the invention will now be explained. Preferred wavelengths are selec¬ ted mainly using two criteria: One of the wavelengths, e.g. λ ₂ , comprised in the modulation ratio (%modλ ₁ /%modλ ₂ ) is selected from a range near the lsobes- tic point of the dominating Hb variety or oxyhaemoglobm (Hb02) and the Hb variety whose amount is primarily being measured with the ratio in question, or in general from a range where a change in the relative amount of haemoglobin has only a minor effect on the signal. The other wavelength in the modulation ratio, in this case

λ _lf is so selected that the Hb varieties in question ha¬ ve a large difference between their extinction coeffi¬ cients at this wavelength. If the isobestic point cannot be used, it is preferable to select the wavelengths so that the difference between the extinction coefficients of the two primary Hb varieties is of opposite sign at the two wavelengths used. The selection of the primary modulation ratio used in the calculation with four wave¬ lengths from all the ratios possible or all six possible pairs of two wavelengths is also made on two grounds: The modulation ratio must have a high or at least a suf¬ ficient sensitivity to a change in the concentration of the Hb variety (Hbxx below) primarily to be measured with this modulation ratio; i.e. (%modA.)/(%mod2,) d(Hbxx) ' ² is large enough.

On the other hand, the sensitivity of the modu¬ lation ratio to temporary wavelength variation between the LED components must be low in relation to sensitivi- ty, i.e.

—(%modA.)/(%mod^ ₂ ) δλ is small enough.

Even for as few as four haemoglobin varieties or dye components, simultaneous satisfaction of all cri- teria is impossible. Selecting the wavelength near the minimum or maximum of the absorption curve reduces the calculation errors caused by variation in the wave¬ lengths of the LEDs. For this reason, the minima and ma¬ xima of the absorption curves and in general their flat portions are particularly good choices. If such a flat portion lies near the isobestic point, the wavelength in question is a good reference value, against which the dye differences are formed. In Fig. 1 there are two such ranges: in the range 790-1000 nm, the absorption curves for all Hb varieties are sufficiently insensitive to changes in wavelength, whereas in the range 680-750 nm

only the absorption curve of deoxyhaemoglobin is sensi¬ tive to wavelength changes. It is generally advisable to select the wavelengths so that two of the four or more possible wavelengths are the same as in currently used pulse oximeters. When this is the case, all the empiri¬ cal information available in the case of pulse oximeters is also available when fractional oxygen saturation is measured by means of modulation ratios. The third wave¬ length range selection is therefore 655-665 nm. Selec- ting the wavelengths in these three ranges improves the accuracy of fractional oxygen saturation measurement on normal human blood. The wavelength ranges mentioned are also suitable for the determination of HbCO.

As it is not possible to make selection of four wavelength ranges without conflicts, MetHb is treated as a special case. For MetHb measurement, a preferred wave¬ length range is 625-650 nm. The selection of four diffe¬ rent wavelength ranges according to the present inventi¬ on is: 625-650 nm, 655-665 nm, 680-750 nm and 790-1000 nm. The calculation procedure is then so adjusted that in all situations a maximal accuracy of determination of fractional oxygen saturation is achieved for these wave¬ length ranges.

In Fig. 2a-2c, a preferred optimal wavelength selection is considered using the Lambert-Beer model. Pulse oximeter type modulation signals are presented for three different special situations: 1. Hypoxemia or high Hb level with normal DysHb levels (Fig. 2a), 2. Carboxy- haemoglobimia or high HbCO while other Hb concentrations are normal (Fig. 2b) and 3. Methaemoglobimia or high MetHb level with normal blood composition in other res¬ pects (Fig. 2c) . In the case presented in Fig. 2, a pre¬ ferred optimal wavelength selection according to the in¬ vention is 900 ±10 nm, 690 ±5 nm, 658 ±5 nm and 630 ±5 nm. Since the isobestic point between oxyhaemoglobin and carboxyhaemoglobin lies near 645 nm, this is a prefe¬ rable wavelength for the calculation of the modulation

ratio as well. At this wavelength, MetHb can be deter¬ mined with sufficient accuracy. Another optimal wave¬ length selection according to the invention is thus 900 ±10 nm, 690 ±5 nm, 658 ±5 nm and 645 ±5 nm. Table 1 be¬ low further illustrates the contradictory nature of the wavelength selection when maximal accuracy of determina¬ tion is required simultaneously for all different situa¬ tions. Table 1 shows those wavelengths which should be used in the first place, as well as those which should not be used, for each one of the different descriptions of the patient's condition regarding oxygenation. Table 1

According to the present invention, the contra- dictions in the selection of wavelength can be reduced and a better selection accuracy can be achieved by weighting the correct modulation ratios in different ways in each blood oxygenation situation. For instance, in the case of hypoxemia, wavelengths 900 nm, 690 nm and 658 nm and modulation ratios calculated from these are always used and the fractional oxygen saturation is weighted more than the quantities for wavelengths 630 nm or 645 nm. In this case, wavelength 630/645 nm is prima¬ rily used to detect the presence of MetHb, but the mea- surement of the amount of DysHb is effected using other wavelengths. A preferable wavelength pair is 690 nm and 900 nm because both HbCO and MetHb affect this modulati¬ on ratio in the same way. When the level of MetHb is

found to be rising, the weighting coefficients are alte¬ red so that the weighting of wavelength 658 nm is redu¬ ced and that of wavelength 630/645 nm is increased. In the case of carboxyhaemoglobin, all wavelengths can be weighted more equally. However, since the absorption coefficients for carboxyhaemoglobin are exceptionally small in the whole range, the HbCO concentration is mainly evident through the fact that carboxyhaemoglobin replaces oxyhaemoglobm. The best wavelength to allow this to be detected is 900 nm, so the modulation ratios for this wavelength must be weighted above the average. A preferred calculation method according to the inventi¬ on is weighted calculation as described below, which is optimised for the wavelengths used and for the patient's illness condition. The present invention also uses a sensor calibration procedure to be described later on, which allows a good analysing accuracy to be maintained at all wavelengths.

Fig. 3 presents a preferred sensor according to the present invention, which is used in the procedure of the present invention for collecting measurement data in a non-mvasive measurement through tissue comprised in the patient's blood circulation. The sensor comprises means 1 for connecting the sensor to a measuring appara- tus. In this embodiment, the sensor is connected to the measuring apparatus via a cable, known in itself, de¬ signed for the transmission of signals. Furthermore, the sensor comprises a light source 2 forming an essential part of it, which emits a light signal at at least two, m this embodiment four predetermined medium wave¬ lengths. The light source comprises a number of light elements 2 ¹ , , 2 ⁴ , each one of which emits light at a selected wavelength different from the others. The sen¬ sor also comprises a receiver 3, which typically is a light-emitting diode or a so-called PIN diode and which is arranged to receive a light signal transmitted through and/or reflected by the target under measure-

ment. The sensor further comprises a storage device 4 for the storage of predetermined sensor-specific data, where the sensor-specific data comprises an extinction coefficient separately determined for each light source and each blood Hb variety and/or dye component that can be measured. The sensor presented in Fig. 3 further com¬ prises a sensor terminal 14, in which the storage device is mounted and to which the light elements 2 , ..., 2 and the receiver 3 are connected. Via the sensor terminal 14, it is possible e.g. to control the sensor by means of the measuring apparatus and to read the data con¬ tained in the storage device 4.

The light elements 2 ¹ , ..., 2 ⁴ in Fig. 3 are con¬ nected in a bipolar circuit, which uses only three con- ductors instead of five and is therefore simpler and mo¬ re advantageous in respect of cable structure. Light elements 2 ¹ , ..., 2 ² and, on the other hand, light elements 2 ³ , ..., 2 ⁴ are connected in reverse directions relative to each other and they are driven by operating current of opposite sign. Such a bipolar control circuit in itself is obvious to the person skilled in the art and is the¬ refore not described here in detail. Generally, the light elements 2 ¹ , ..., 2 ⁴ are small LED chips about 0.3 * 0.3 mm" ^" in size, which are adjusted on a hybrid or cir- cuit board or a corresponding mounting. Surface mounted devices can also be used, but to produce a light source unit as small as possible, chips are a better choice. Further, one light element 2 ⁿ may consist of two identi¬ cal light sources (not shown) connected in series. Such a connection is preferable if the LED has an insuffi¬ cient luminosity when used alone or if the attenuation caused by the tissue at this wavelength is so large that the light transmitted through it is of insufficient in¬ tensity. Fig. 4 presents a particularly advantageous so¬ lution that allows the surface temperature of the sensor to be reduced in relation to the conventional structure.

In this embodiment of the invention, a light source hy¬ brid 2 is thermally anchored on the sensor frame 5 and its thermal connection to the skin surface is minimised. Such a structure allows a low sensor surface temperature to be maintained even if several light elements are us¬ ed. Moreover, in the structure presented in Fig. 4, the¬ re is between the sensor surface and the light source hybrid a cavity 15 that causes diffuse reflection, thus diffusing and smoothing the light emitted. Therefore, the light emission reaching the skin surface is smoothly spread over a large area, which is an advantage e.g. in regard of the elimination of motional artefacts from the light detector signal. Comprised in the sphere of the present invention are also sensors that may use diffe- rent wavelength ranges and be intended for some other measurement purpose. Such a sensor may be e.g. one in¬ tended for the measurement of blood bilirubin or a given

Fig. 4 also illustrates a preferred embodiment of the invention, in which the receiver 3 is disposed on the same side relative to the tissue as the light ele¬ ment 2 , so that the receiver receives a signal reflec¬ ted from the tissue. The sensor depicted in Fig. 4 is particularly advantageous in cases where the attenuation caused by the tissue is so strong that the light transmission at all wavelengths is insufficient for the measurement of fractional oxygen saturation in the transmission geometry. In such cases, a reflection sen¬ sor as presented in Fig. 4 can be used instead of a se- ries connection of a number of LEDs . Since light transmission decreases especially in the case of short wavelengths, but on the other hand scattering increases at these wavelengths, the advantages of the reflection geometry become manifest via both mechanisms. Thus, the signal produced by the reflection sensor increases in relation to other sensor solutions.

Fig. 5 presents a typical finger sensor struc¬ ture, which comprises a first optic fibre 6 for passing a light signal emitted by a light source (not shown) to the target under measurement and a second optic fibre 7 for passing a light signal transmitted through the tis¬ sue to the receiver (not shown) . With this sensor struc¬ ture, too, excessive temperatures against the patient's skin are avoided. In the finger sensor, the fastening device 9 used to attach the sensor components near the target to be measured is a clothes-peg type clamp. Let it be further stated that the solution illustrated by Fig. 5 can be implemented e.g. as an ear sensor or the sensor components can be fastened by means of tape or some other type of separate fixing means. When a fibre-optic circuit as shown in Fig. 5 is used, the LEDs as well as the detector may be placed inside a patient monitor or a corresponding measuring apparatus, in which case the light is passed via the first optic fibre 6 to the sensor and via the second op- tic fibre 7 back to the measuring apparatus. In this so¬ lution presented in Fig. 5, sensor temperature rise is avoided altogether.

In a preferred embodiment (not shown) of the present invention, two separate sensors are used at dif- ferent measuring points. One of the sensors is prefera¬ bly a conventional pulse oximeter sensor. For the other sensor, one common wavelength is selected and used to correct the scaling of the modulation ratios between the sensors, calculated in a crosswise manner. Therefore, in two separate sensors, at least five LEDs must be used instead of four. However, this structure is preferable even in a case where one of the sensors is only used for short periods to measure the dyshaemoglobin level while the other sensor is used continuously e.g. for the mea- surement of functional oxygen saturation.

Referring again to Fig. 3, the light detector 3 or receiver can be divided into four separate wavelength

channels by using light filters. Fig. 3 shows light fil¬ ters 8 , , 8 arranged between the tissue and the recei¬ ver 3. Each filter has its own wavelength pass band in accordance with the wavelengths used in each measure- ment. In a preferred embodiment, the light filters can be electrically controlled so that the passband wave¬ length can be changed by means of the measuring appara¬ tus. When light filters are used, it is preferable to use a common wide-band light source. In principle, such a light source with sufficient bandwidth may consist of one or more LEDs with a medium wavelength of about 680 nm and a bandwidth of about 60-80 nm. In this case, the LED for the near-infrared range may have a wavelength of e.g. 910 nm. Another solution is to use a wide-band ha- logen or other conventional light source and pass the light via fibre to the oxygen saturation sensor.

Fig. 6 presents a preferred embodiment of the measuring apparatus of the present invention. The measu¬ ring apparatus in Fig. 6 comprises a sensor as described above by referring to Fig. 3, together with a sensor terminal 14. The measuring apparatus preferably compri¬ ses at least a signal processing device 10, which may be a microprocessor or a corresponding programmable compo¬ nent known m itself. Further, in this embodiment the light elements 2 , , 2 are controlled by two bipolar drive circuits (not shown) controlled by a microproces¬ sor. The radiation emitted by the light elements 2 ¹ , ..., 2 is passed through the tissue or reflected from it to the light detector 3 of the sensor, from where the sig- nal is passed via the sensor terminal 14 to the current- voltage converter (not shown) of the preamplifier of the apparatus. After this, the signal is amplified in a man¬ ner controlled by the microprocessor 10. Let it be furt¬ her stated that in Fig. 6 the measuring apparatus and sensor are presented in a greatly reduced form because most of the technology used m the measuring apparatus consists of electronics known to the person skilled in

the art. The user interface and display functions as well as other general properties of the measuring appa¬ ratus are defined via the microprocessor 10. According to the invention, the measuring apparatus presented in Fig. 6 further comprises a calculating device 11, which is programmed to carry out the measuring procedure of the invention with the aid of the sensor. Other essenti¬ al components in the measuring apparatus are a reader device 12 and an identifier device 13, by means of which the information stored in the memory element 4 of the sensor is read and the sensor type identified. In this context it should also be noted that with modern techno¬ logy even a hardware implementation is possible in which all the above-mentioned components, i.e. the signal pro- cessmg device 10, calculating device 11, reader device 12 and identifier device 13, are programmed mto a sing¬ le application-specific integrated circuit (ASIC) .

A preferred embodiment of the present invention is one in which the measuring time in the embodiment presented in Fig. 6 is flexibly divided between the channels. In this case, e.g. most of the time would be spent on measuring the fractional oxygen saturation whi¬ le a clearly shorter time is reserved for dyshaemoglobin measurement or the dyshaemoglobin level is only deter- mined when necessary. The time division used in the mea¬ surement may be a fixed division or it can be flexibly changed as required in each situation.

The theory of pulse oximetry is generally pre¬ sented as being based on the Lambert-Beer law. According to the theory, light transmission through the tissue at each wavelength is exponentially dependent on the absor- bance of the tissue. This theory is generally accepted and established in pulse oximetry. Omitting details, the theory is generalised for four different wavelengths in the matrix format described above, as follows:

where %mod i is the modulation percentage for light transmission as measured at wavelength i, i.e. the proportion of light transmission varying at heartbeat frequency as a percentage of the total light transmissi¬ on, the ij-element of the ε-matrix is the extinction coefficient of the haemoglobin component j of arterial blood for wavelength i and the haemoglobin components in percentages are placed in the vertical vector (HbX _x , HbX ₂ , ..., HbX.,) in this order j. Thus, the horizontal li¬ nes of the ε-matrix are the extinction coefficients of a given wavelength for different Hb varieties. The cons¬ tant C determines the units on the left side of the equation. In the above system of equations it has been assumed that the modulation percentage is small (< 10 %) . If this is not the case, instead of %mod i the exact theoretic form is used:

% mod i <r- ln(l ± % mod i / 100%) * 100%

The modulation percentages are the basic sig- nals of the pulse oximeter, so in the Lambert-Beer theo¬ ry it should be possible to calculate the concentrations of different Hb varieties directly by using a matrix re¬ verse to the ε-matrix. However, in practice this system of equations is not linear. As stated before, the diver- gences between theory and practice are due to the fact that the actual extinction coefficients ε~ are also de¬ pendent on the scattering of light caused by the tissue and blood and on the combined effect of absorption and scattering. The corrections are the larger the larger is the proportion of the attenuation caused by absorption and scattering. The Lambert-Beer law assumes that the scattering of light and the non-homogeneity of tissue are not taken into account. In practice, therefore, in non-mvasive measurement the elements ε _ι;) of the ε-matrix

differ from the extinction coefficients ε' _ι;) of real blood. This correlation between the actual and theoreti¬ cal extinction matrices can be represented by a trans¬ formation T that depends on the total absorption and scattering at the wavelengths used in measurement. This relationship can be presented in the following form:

The * sign at the top edge of the matrix means that the elements of the matrix are ε' _1D . Since the total absorption is the sum of the extinction coefficients weighted by the haemoglobin pro¬ portions, the transformation T is dependent on the ha¬ emoglobin concentrations themselves, or (T)=(T)(HbX„ HbX ₂ , HbX ₃ ,..., HbX _j ) In multi-component analysis of blood compositi¬ on, the divergence between theory and practical measure¬ ments has significant consequences, which are disregar¬ ded e.g. in patent application EP 0 524 083 Al . First, the system of equations (1) is not linear, i.e. it can- not be solved by using a reverse matrix. Second, any Hb variety and change in its concentration also affects the actual extinction coefficients of other Hb varieties. For this reason, the system of equations (1) must be solved for all Hb varieties present in blood, and the fairly strongly absorbing MetHb variety cannot be exclu¬ ded from the analysis.

In the following, referring to Fig. 7, which depicts the correlation of theoretic and empiric oxygen saturation as functions of the ratio (R/IR) of the modu- lation percentages determined at wavelengths 660 nm (R) and 940 nm (IR) , it will be demonstrated how the system of equations (1) is solved with the aid of the modulati¬ on percentage ratios for several Hb varieties, i.e. Hb02, Hb, HbCO, MetHb and Hbx, where Hbx is a haemoglo-

bin component that appears in the patient's blood in a special situation, e.g. nitrosylhaemoglobin HbNO or sul- phohaemoglobin HbS. In principle, Hbx may represent any blood dye component, such as an artificial dye. In Fig. 7, the significance of the matrix transformation (2) is that the transformation serves to transfer the theoretic Lambert-Beer curve onto the empi¬ ric curve (in the figure, Wukitsch et al) . This can be effected via a function conversion that changes the nu- meric values on the R/IR axis in the desired (non¬ linear) manner. With four different wavelengths, a cor¬ responding correlation between the theoretic and empiric modulation ratios and the oxygen saturation level is obtained for each one of the six different wavelength pairs. The transformation T contains the information needed for the transfer of all these pairs of curves, and conversely, the transfer of the curve pairs determi¬ nes the transformation T. Below is a description of how this kind of transformation functions in multi-component analysis of Hb varieties with a plurality of different wavelengths and a plurality of different haemoglobin components.

The description is limited to the use of four different wavelengths. However, under certain assump- tions, the procedure can be used to analyse more than four Hb varieties with sufficient accuracy.

In the analysing procedure of the invention, the total extinction coefficient of dyshaemoglobins at wavelength i is written as:

' ^■ > = ^ DysHb * ^ ^Mam > ⁺ ^ DysHb * ^"* ^C0 > ⁺ ^ DysHb * *.> < ^/M > ^{( 3} > where

DysHb =MetHb +HbCO+HbX (4) and HbX is a third dyshaemoglobin variety. In its general representation HbX can also be interpreted as an artificial dye concentration in blood.

With four different wavelengths, the unknown system of equations can be written in the form:

^Ϊ HbOl 1 - Hb0 ₂ - DysHb :5 ) DysHb where the ε-matrix is in accordance with the Lambert-Beer theory and contains the extinction coeffi¬ cients documented for blood in literature. The modulati¬ on percentages predicted by the theory are thus in the vertical %mod I vector. C again represents the transfor¬ mation of the units. The concentrations of different Hb varieties are presented as proportional shares.

The experimentally measured modulation percen¬ tages are:

It can be observed at this point that the transformation T is neither linear nor e.g. a matrix multiplication, but it does have an unambiguous inverse transformation. We shall now calculate a representation of this transformation.

Equation (5) is divided into modulation ratios as follows: % mod k _ ε _kl * Hb02 + ε _a * (1 - HbO2 - DysHb) + ε _u * DysHb

% mod 1 ε _n * HbO2 + -?„ * (! - Hb02 - DysHb) + ε _a * DysHb

where k,1 = 1, 2, 3, 4 and k # 1. In the case of four wavelengths, this division can be made in six different ways, i.e. there will be six expressions (8). Let us write [ (%mod k) / (%mod 1)]=Z and solve each equation (8) for Hb02:

_ (1 - DysHb) * (ε _a - Z* ε _Q ) +DysHb* (ε _ki - Z* ε _n )

HbO2

Z * (e _kl -£,,)- (ε _a - ε _a ) (9)

The theoretic modulation ratio Z can be ex¬ pressed using an experimentally measured modulation ra¬ tio Z' so that

Z=f _kI (Z') (10) which determines the transformation inv (T) : Z'→Z. The experimentally correct fractional oxygen sa¬ turation is obtained from equation (9) by substituting formula (10) for Z. Consequently, when the wavelengths are 660 nm and 940 nm, function f (660,940) transfers the experimentally measured curve in Fig. 7 onto the Lambert-Beer curve by changing the numeric values on the vertical axis in the manner determined by function f (660, 940). Function f depends on the total absorption and scattering appearing at wavelengths k and 1, but no longer on different Hb varieties. Thus, function f can be found e.g. via hypoxemia tests with normal HbCO and MetHb concentrations. This makes the calibration of fractional saturation measurement considerably easier. Function f is known via the calibration data for conven¬ tional pulse oximeter measurement at wavelengths R = 660 nm and IR = 900 nm (or 940 nm) . If the representations of transformation inv(T) are designated e.g. at wave¬ lengths 630 nm, 660 nm and 900 nm as follows: R = f660,900(R') Q = f660,690(Q') ( %modl ϊ f %modl V

P f630,690(P') % mod2 % mod2 o s f690,900(S') % mod 3 (τ- ¹ ) % mod3

T f630,660(T) % mod 4) V % mod 4) u f630,900(U')

then all functions f can be determined by chan¬ ging the concentration of one Hb variety and therefore the total absorption for the wavelength in question. Representations R' -R, Q' -Q, P'-P, S'-S, T'-T and U' -U being known, equations (3) and (4) are solved and the fractional saturation level is solved for 3-6 modulation ratios from the system of equations (9) iteratively so that the fractional oxygen saturation level Hb02, total concentration of dyshaemoglobin DysHb and the compositi- on of dyshaemoglobin varieties can be determined in a compatible manner. The calculation may include verifica¬ tion of the result by checking it against the condition built into the equations that the sum of all Hb varie¬ ties is 100%. The iteration process can be accelerated using clinical information about the patient's conditi¬ on. Typically, only one DysHb variety in the patient's blood has an increased value. Moreover, the amount of dyshaemoglobin changes slowly with time, so equations (3) and (4) need not necessarily be used in real-time calculation of the fractional oxygen saturation level. These equations thus have a greater importance in the identification of dyshaemoglobin varieties.

Above, the way in which multi-component analy¬ sis of haemoglobin composition of blood can be lmplemen- ted has been described. Next, we shall discuss the ques¬ tion of how the concentration of dyshaemoglobin can be identified. After that there follows a discussion of how the sensor calibration data are saved.

In clinical measurement of fractional oxygen saturation, the dyshaemoglobin variety to be monitored is generally known beforehand. It is generally also known that the concentrations of other dyshaemoglobin varieties are normal and remain unchanged in the planned treatment. In such a situation, the user may input the dyshaemoglobin variety of primary interest into the ap¬ paratus. In this case, the non-mterestmg dyshaemoglo¬ bin concentrations are set as normal parameters in ex-

pressions (3) and (4) while the dyshaemoglobin variety of primary interest is set as an unknown parameter. In principle, the user may also make haemoglobin values me¬ asured from a blood sample available to the apparatus. The apparatus performs the calculation using oxyha- emoglobin (Hb02), deoxyhaemoglobin (Hb) and the dysha¬ emoglobin variety of primary interest as unknown vari¬ ables while the amount of other dyshaemoglobin varieties is taken to be constant. For reasons of user friendliness, it is gene¬ rally required that it should be possible to "measure" the dyshaemoglobin variety in the same way as the amount of oxyhaemoglobin. In this case, the dyshaemoglobin com¬ position is not known in advance, and the effects of treatment administered to the patient are not known, either. In principle, the normal composition of human blood, which contains four haemoglobin varieties, can be analysed with four different wavelengths. In the blood of a person who is critically ill, other HbX varieties or other dyes may exist in consequence of treatment or medication. The existence of these may be identified by stipulating that the Hb02 value calculated via iteration from five different modulation ratios should fall within certain predetermined limits of variation. An additional stipulation that may be applied is the assumption inclu¬ ded in the equations that the sum of haemoglobin varie¬ ties should be 100%. Identification of a divergent HbX variety or dye is effected when the iteration result is not consistent. If the result of iterative calculation indicates an increased MetHb value and the patient is having NO-treatment, the HbX variety is assumed to be nitrosylhaemoglobm HbNO. The extinction coefficient of this is added to expressions (3) and (4) . The HbCO in the expressions is set to the normal level or about 1%. If the next iterative calculation yields a substantially better consistence, HbNO is identified. If in normal iterative calculation no increased MetHb or HbCO values

are observed but the accuracy of iteration is low, then a check is made to verify whether any other pre¬ programmed Hb varieties or dyes exist. If the iterative calculation yields a substantially better result, then the haemoglobin variety or dye in question is identi¬ fied.

Another advantage provided by the calculation method described above is that changes in the sensor LED wavelengths can be easily and simply taken into account in the calculation of fractional saturation. The sensor calibration process comprises the following steps:

1. The wavelengths and spectral emission of the sensor LEDs are measured.

2. For the absorption curves of Hb varieties documented in the Lambert-Beer theory or in literature, an effective extinction coefficient is calculated for the wavelength of each LED, taking into account the me¬ dium wavelength and spectral line width of the LED. The receiver sensitivity curve can also be included in this process.

3. The extinction coefficients ε _x - are stored separately e.g. in matrix format for all light elements.

4. The matrix of the extinction coefficients is stored in a storage device, such as an EEPROM, provided in the sensor.

5. Transformations of the theoretic extinction matrix and of an empirical matrix are stored in the same EEPROM as functions R, Q, P, S, T, and U. These functions are used in the procedure m the manner described above.

6. The sensor type, i.e. optionally MetHb sen¬ sor, CO sensor, etc., is stored in the EEPROM.

7. Data relating to the manufacture and guaran¬ tee of the sensor and other corresponding data are sto- red in the EEPROM.

Therefore, the sensor of the present invention provides a great advantage because the measuring appara-

tus can read the contents of the EEPROM and use the ca¬ libration data in the calculation of fractional satura¬ tion .

The invention is not limited to the examples of its embodiments described above, but instead many varia¬ tions are possible within the framework of the inventive idea defined by the claims.