DE69736556T2 - IMPLANTABLE SENSOR AND ARRANGEMENT FOR IN-VIVO MEASUREMENT AND CONTROL OF CONCENTRATIONS IN LIQUIDS AND GASES - Google Patents

Description

BACKGROUND THE INVENTION

The The present invention relates generally to medical devices and methods for in vivo determination the concentration of a chemical constituent in a body fluid by means of a direct spectroscopic analysis of the liquid and to remove or eliminate actual or potential Artifacts from measured spectral data.

• (1) Die derzeitigen handelsüblichen Instrumente sind invasiv und benötigen im Allgemeinen einen Tropfen Flüssigkeit zur Bestimmung der chemischen Bestandteile.
• (2) Die derzeitigen handelsüblichen Instrumente benötigen ein einzigartiges chemisches Reagens für die Bestimmung der einzelnen Bestandteile. Die Reagenzien sind im Allgemeinen Einwegprodukte, die häufig nach jeder Messung ersetzt werden müssen. Die Unterhaltung der Instrumente ist daher teuer und unzweckmäßig, insbesondere wenn es sich um ein implantierbares Produkt handelt.
• (3) Implantierbare chemische Sensoren werden aufgrund der mangelnden Stabilität der Reagenzien bei Körpertemperatur im Zeitverlauf abgedrängt und weisen eine nur kurze Lebensdauer aufgrund von Biokompatibilitätsproblemen auf.
• (4) Chemische Sensoren verbrauchen im Allgemeinen den Bestandteil, den sie messen sollen, und beeinflussen so die Genauigkeit. Bei einem implantierbaren Sensor führt die Gewebereaktion mit dem Sensor unweigerlich zu einer Verringerung des Vorrats der Bestandteile, wodurch auch die Genauigkeit reduziert wird.

In medicine, measuring the composition of blood and other body fluids is often a valuable diagnostic tool used as the basis for drug titration and therapy setting. Roughly speaking, there are two classes of devices for measuring chemical constituents: chemical and optical devices. Chemical techniques are the most common and in particular all techniques used in the clinical laboratory are based on chemical techniques. Chemical techniques inevitably rely on the reactant-responsive ingredient and either optical or electrical measurement of the reagent or reaction by-products Unfortunately, current chemical techniques have numerous disadvantages, including the following:

• (1) The current commercial instruments are invasive and generally require one drop of liquid to determine the chemical constituents.
• (2) The current commercial instruments require a unique chemical reagent for the determination of the individual constituents. The reagents are generally disposable products that often need to be replaced after each measurement. The maintenance of the instruments is therefore expensive and inconvenient, especially if it is an implantable product.
• (3) Implantable chemical sensors are stalled over time due to the lack of stability of reagents at body temperature, and have a short life due to biocompatibility problems.
• (4) Chemical sensors generally consume the component they are to measure, thereby affecting accuracy. In an implantable sensor, the tissue reaction with the sensor inevitably leads to a reduction in the supply of components, which also reduces the accuracy.

optical Techniques have been used by researchers as a means of overcoming all problems in the Related to chemical techniques because they are not reagents or need some kind of chemical reaction and no direct Interaction of the sensor with the chemical component required is. Optical techniques can theoretically for a continuous non-invasive simultaneous determination of several chemical ingredients. Despite the promising approach optical techniques and efforts numerous groups around the world to perfect these techniques, optical techniques are rarely used in clinical medicine Commitment. The pulse oximeter and an instrument for measuring the percentage Body fat are notable exceptions.

One Sensor specially sought by researchers is a blood sugar sensor for home use for persons with insulin-dependent diabetes. Ideally would be Such a sensor would be non-invasive or permanently implantable, the continuous or allow quasi-continuous measurement of blood sugar levels and would be over the physiological area of blood chemistry and other conditions exactly and he would only require minimal intervention in the patient. It are reportedly over 60 groups around the world who are currently trying such a sensor to develop, but so far no group is known that she is close to a success. The reason for these efforts is the rising costs the diabetes complications for the patient and the healthcare system. In studies could be shown will significantly reduce diabetes complications can, when the blood sugar is closely monitored becomes.

diabetes is an important chronic disease in the United States alone have health costs ranging from $ 90 to $ 110 billion yearly caused. Approximately 750,000 people in the United States suffer from insulin-dependent diabetes Mellitus (IDDM), also known as type I diabetes, juvenile diabetes or Insulin deficiency diabetes is referred to. Another 16 million people suffer from non-insulin dependent Diabetes mellitus (NIDDM), also known as type II diabetes or adult diabetes referred to as. Globally, the World Health Organization estimates the number of diabetics on over 80 million, of which many cases still are not diagnosed.

The acute complications of diabetes include metabolic abnormalities such as hypoglycaemia and ketoacidosis. Long-term complications affect the blood vessels and tissues, leading to retinopathy, stroke, myocardial infarction, heart failure, arterial occlusive disease, nephropathy, renal failure, and peripheral neuropathy. For example, diabetes has affected over 39,000 people in the United States last year Blindness. Approximately 25% of all IDDM patients and 10% of those with NIDDM suffer from proliferative retinopathy within 15 years after diagnosis of diabetes. Approximately 34% of patients with IDDM and 19% of cases of NIDDM have diabetic kidney disease within 15 years of diagnosis. Although these diseases all occur in the general population, they occur earlier and more frequently in diabetics and progress faster.

In Several studies have proven that the onset or progression of complications of diabetes in diabetics with IDDM significantly can be reduced if their blood sugar levels are on one almost normal level can be maintained. In the study Diabetes Control and Complications Trial (DCCT) was a nine-year randomized prospective study of more than 1400 subjects. The DCCT compared the complication rate after standard treatment from IDDM (mean blood sugar 195 mg / dl) with a close-meshed Control with frequent Finger blood tests and frequent Insulin injections (mean blood sugar 155 mg / dl). It showed that intensive care is at risk for the onset of retinopathy decreased by 76% and decreased the progression of retinopathy by 54%. A smaller study, the so-called Oslo study, observed 45 insulin-dependent Diabetics randomized into three groups: several Insulin injections, continuous subcutaneous insulin infusion (CSII) and conventional twice daily Insulin injections. With CSII and multiple injections could almost normoglycemia achieved, but not with conventional therapy. After two Years, the groups with CSII and multiple injections showed little Deterioration as measured by the number of microaneurysms and bleeding while the Group with conventional Therapy showed a significant deterioration.

With regard to these studies, several observations can be made:
Although close monitoring reduces the secondary complication rate of diabetes compared to loosely controlled diabetes, the risk of complications is still several times higher than in those without diabetes.

A Close control involves significant inconvenience and inconvenience Pain, so adherence to the treatment regimen for many, if not for Most diabetics can not be realized. For determination For example, the appropriate insulin injection needs a diabetic measure your blood glucose at least five to six times a day, where he has to prick blood into his finger, and then to measure the blood glucose value with a blood glucose meter. For many Patients taking blood is a painful and unpleasant one Process that is in fact considerably more painful and unpleasant as the actual insulin injection.

Both Studies showed a significant increase (two to three times) severe and sometimes asymptomatic hypoglycaemia in the patients tight control compared to those under conventional Therapy. A severe hypoglycemia leads to unconsciousness and sometimes even coma. In the closely monitored patients was the incidence of severe hypoglycemia 0.6 to 0.8 cases during the year they are common among patients Therapy in 0.2 to 0.4 cases a year. For this reason, a continuous or quasicontinuierlicher Sensor that can give an alarm, desirable.

The Blood glucose levels monitored closely Patients show a mean of 155 mg / dl and fluctuate in the Daily course between a minimum value of 120 mg / dl and a maximum value of 190 mg / dl. These values are still pretty much higher than the glycemic values of non-diabetics, which are on average 110 mg / dl and during the day run between 90 and 120 mg / dl. Although the DCCT study points out that further lowering of blood sugar is desirable, that speaks Risk of hypoglycaemia against another cut.

It It is believed that a system that raises blood sugar really normal values (90-120 mg / dl) can keep the diabetes-related Can potentially eliminate complications. Such a system would be with a Insulin infusion pump and a glucose sensor. Neither the pump would still be the sensor require an interaction on the part of the patient, and both would certainly be painless. The development of the pumps is already well advanced. The Missing limb is a sensor that continuously monitors blood sugar can measure, is non-invasive or implantable and has a large physiological Range provides accurate readings.

Researchers' pathways for measuring blood sugar include a variety of different techniques, both optical and chemical. Several researchers have attempted implantable chemical sensors, but these sensors seem to inevitably fail because of the above reasons. U.S. Patent No. 5,353,792 (Lubbers et al.) Discloses a modification of the chemical sensor which employs "optically excitable and easily readable substances" one of which could presumably interact with glucose. The disadvantage this way is the same as with the other chemical techniques, ie the substances probably have to interact directly with the component to be measured and thus may eventually become depleted over time, or alternatively the substances may be blocked from interacting with the component by encapsulation tissue. Another variation of the approach of Lubbers et al. are affinity sensors based on fiber optics and described in Schultz, JS et al., "Affinity Sensor: A New Technique for Developing Implantable Sensors for Glucose and Other Metabolites" [Affinity Sensor: A New Technique for Developing Implantable Sensors for Glucose and Other Metabolites] , Diabetes Care, Vol. 5, No. 3, p. 243, May-June 1982; and Mansouri, S. et al., Bio / Technology, Vol. 2, p. 385, 1984. These sensors also suffer in the long term a decrease in sensitivity.

The Optical techniques also include a number of different ones Methods. In most experiments, noninvasive transcutaneous Paths with infrared spectroscopy used. Infrared spectroscopy is based on the absorption of infrared light. The absorption amount depends from the concentration and the path of light through the to measuring liquid from. Every chemical ingredient has its own unique Absorption spectra, depending on the weight of each atom and strength of each Molecular binding in a molecule. If one is enough greater Signal-to-Noise Ratio is present, it should be theoretically possible the presence and concentration of each chemical constituent in a liquid, such as. Blood, precise too determine. Transcutaneous infrared spectroscopy techniques show variability in the optical coupling, a weak signal-to-noise ratio and an excessive number Artifacts. So far could not be shown in any disclosed technique Be that reliable is working. A technique disclosed in U.S. Patent Application No. 08 / 500,388, filed on 6 July 1995 with the title "IMPLANTABLE SENSOR AND SYSTEM FOR MEASUREMENT AND CONTROL OF BLOOD CONSTITUTENT LEVELS "[IMPLANTABLE SENSOR AND SYSTEM FOR MEASURING AND CONTROL OF CONCENTRATIONS OF BLOOD COMPONENTS], the problem with the Coupling and the signal-to-noise ratio by implantation of the sensor partially solve it instead of its transcutaneous application. A A more detailed analysis of these two approaches will be provided below discussed.

Besides were also tried several other optical ways. US Patent No. 4,704,029 (Van Heuvelen) discloses an implantable device, from the measurement of the refractive index of blood with an optical Device which is in direct contact with the blood, on the glucose concentration is closed. This device has numerous disadvantages one of which is the need for direct contact with the Device with blood is. U.S. Patent No. 5,209,231 (Cote et al.) reveals a slightly different way in which the rotation of a plane of polarized light by blood or other body fluid is measured. This technique and other similar techniques work with a very small turn (milligrams) and therefore produce a very small signal, making the method very sensitive to Scattering, movement and fluctuations in the path length reacts.

• (1) Körpergewebe und insbesondere Blut bestehen hauptsächlich aus Wasser. Wasser weist eine sehr starke intrinsische Absorption im Infrarotbereich (IR-Bereich) auf, insbesondere im mittleren IR-Bereich. Die starke Absorption durch Wasser führt zu einer starken Abnahme des Signal-zu-Rauschabstands, wenn die Weglängen verhältnismäßig gering sind. Im mittleren und nahen IR-Bereich müssen die Weglängen bei ca. 25 oder 1000 μm liegen, um einen akzeptablen Signal-zu-Rauschabstand zu erhalten.
• (2) Die Konzentrationen der Bestandteile sind im Vergleich zu Wasser gering. Die Spektren der verschiedenen Bestandteile von Blut und anderen Körperflüssigkeiten neigen dazu, sich selbst und das Spektrum von Wasser zu überlappen, und fast die gesamten Konzentrationen der Bestandteile schwanken über den physiologischen Bereich. Bei der Messung der Blutzuckerkonzentration muss die Messung beispielsweise gegen einen Hintergrund durchgeführt werden, bei dem die Konzentration der anderen Bestandteile sich ständig ändert und die Spektren der anderen Bestandteile das von Glukose überlappen.
• (3) Im nahen IR-Bereich verschieben sich beide Positionen der Peaks und die Intensität der Wasserspektren mit der Temperatur und dem pH-Wert, so dass die Messung komplizierter wird.
• (4) Blut und andere Körpergewebe neigen dazu, Licht erheblich zu streuen, so dass der vom Sensor nachgewiesene Signal-zu-Rauschabstand weiter reduziert und die Variabilität verstärkt wird.

As mentioned above, it should theoretically be possible to determine the concentration of constituents by infrared spectroscopy, provided there is sufficient signal-to-noise ratio, but in practice this determination has proved to be a major challenge. In particular, the blood glucose measurement proved to be very difficult, even in vitro. There are mutliple reasons for this:

• (1) Body tissues and especially blood are mainly water. Water has a very high intrinsic absorption in the infrared range (IR range), in particular in the middle IR range. The strong absorption by water leads to a sharp decrease of the signal-to-noise ratio when the path lengths are relatively low. In the middle and near IR range, the path lengths must be about 25 or 1000 μm in order to obtain an acceptable signal-to-noise ratio.
• (2) The concentrations of the components are low compared to water. The spectra of the various constituents of blood and other body fluids tend to overlap themselves and the spectrum of water, and almost all of the concentrations of the components vary over the physiological range. For example, when measuring the blood glucose concentration, the measurement must be performed against a background where the concentration of the other ingredients is constantly changing and the spectra of the other ingredients are overlapping with those of glucose.
• (3) In the near IR region, both positions of the peaks and the intensity of the water spectra shift with the temperature and the pH value, so that the measurement becomes more complicated.
• (4) Blood and other body tissues tend to scatter light significantly, further reducing the signal-to-noise ratio detected by the sensor and increasing variability.

Despite these differences, several researchers have been able to determine the concentration of glucose and other components in vitro with acceptable or nearly acceptable accuracy, using multivariate techniques for the evaluation of spotting tren used. It should be emphasized that these measurements were made in vitro using a prior art spectrophotometer, well-optimized setup, and generally with factors that are not reproducible in vivo (ie, temperature stability, constant path length). In addition, these measurements typically required a large number of spectral points (> 100) in order to make good and accurate predictions of the concentration of the component. Therefore, to be able to perform these measurements in vivo, extraordinary measures must be taken to ensure that the signal-to-noise ratio is as good or better than the measurements in vitro and that there are no additional artifacts.

in view of the difficulties in measuring in vitro should not come as a surprise that the transcutaneous routes did not work. In a transcutaneous Sensor "sees" the light a lot larger number of variables than is the case in vitro, indicating the interaction of light with skin, fat, bones and the like. The separating the spectra becomes very difficult. In addition, the signal-to-noise ratio, which is available in this way is due to the significantly poorer optical coupling, the Presence of additionally scattering Tissue (e.g., skin) and the absorption of tissue that does not contain glucose contains (e.g., fat), significantly lower than in vitro measurements. in view of the necessary accuracy in the spectra for an accurate determination of the Concentration of the ingredient can cause excessive artifacts Disguise measurements.

It was trying to address some of the problems associated with the transcutaneous To overcome infrared sensor, by using an implantable glucose sensor like his in U.S. Patent Application No. 08 / 500,388, filed July 6, 1995, entitled "IMPLANTABLE SENSOR AND SYSTEM FOR MEASUREMENT AND CONTROL OF BLOOD CONSTITUTENT LEVELS "[IMPLANT BRANDER SENSOR AND SYSTEM FOR MEASURING AND CONTROL OF CONCENTRATIONS OF BLOOD COMPONENTS]. This sensor improved For example, the signal-to-noise ratio by implantation of Sensors, which eliminates some of the sources of optical loss were. The sensor in this patent application consists of each other spaced pairs of infrared light sources and infrared sensitive ones Detectors. Each source and detector pair are so different spaced, that light from the source flows through a blood vessel and from the detector is recorded. Each source gives a different discreet one narrow band of light. The considerable spatial distance between each Pair of source / detector causes a significant spatial Distance of the spectral lines emitted by the sources. Each detector thus "sees" a different one spatial Region of the blood vessel. by virtue of this fact and the dynamic nature of the blood vessels assumed that the spectral information output by the detectors can not be used to obtain sufficiently accurate blood glucose levels.

Also corrected the measuring technique in the patent application described above not to spectral artifacts due to scattering and absorption by the blood vessel wall and other tissue located in the light path between the exit of each source and the input of the corresponding detector. Furthermore the diameter of a blood vessel is neither constant over time, there is a blood vessel with each Heartbeat bulges and again collapses, still uniform in diameter. Further is blood spatially not homogeneous, so that in the scatter from point to point considerable variability may be present. The variability in the path length the one listed above Patent application described measuring techniques can thus on the order of magnitude from 1 part to 100 (in absorbance units), which is big enough is to take advantage of the spectral information obtained from it to destroy.

Lots The problems discussed above also arise when other components of body fluids be measured. Overall, there are significant obstacles to accurate Measurements of body fluid components in vivo. Despite a long-standing need, the sensors could and spectral analysis techniques do not yet do this job to solve. The present invention fulfills this need and provides both sensors and spectral analysis techniques the exact measurements of body fluid components deliver in vivo.

Out EP-A-404562 is an implantable sensor for in vivo spectroscopy known for evaluating the signals coming from the sensor and for determining the concentration of a constituent of a body fluid uses a processor, where a light source emits light and a detector the emitted light, that with the body fluid interacts, proves.

The earlier filed WO-A-9701986 discloses a similar sensor in which light with a variety of wavelengths over one essentially collinear light path is transmitted through the body fluid.

Notwithstanding the above-mentioned prior art, it is the object of the present invention to develop a sensor which is a composite tes spectrum, ie body fluid and foreign tissue, and the means for correcting for the spectra of a foreign component.

These The object is achieved by the characterizing features of claim 1. Further advantageous embodiments of the present invention the dependent claims be removed.

SHORT DESCRIPTION THE DRAWINGS

The The following detailed description of preferred embodiments The invention is better in connection with the attached drawings understandable. To illustrate the invention, the drawings show embodiments which currently preferred. It is understood, however, that the invention is not limited to the exact arrangements and instruments as shown. In the drawings show:

1 a schematic view of an in vivo sensor according to a preferred embodiment of the present invention;

2 a schematic view of a source in the sensor of 1 can be used;

3A a sectional view of a first physical implementation of a device for attachment of implantable parts of the sensor 1 when the device is attached to a vein;

3B a lateral plan view of the device 3A ;

3C a sectional view of a second physical implementation of a device for attachment of implantable parts of the sensor 1 when the device is attached to a vein;

4A a sectional view of a third physical implementation of a device for attachment of implantable parts of the sensor 1 when the device is attached to a vessel membrane;

4B a side view of the device from 4A ;

4C a sectional view of a fourth physical implementation of a device for attachment of implantable parts of the sensor 1 when the device is attached to a vessel membrane;

5 a schematic view of the source with a single output in the sensor off 1 can be used;

6 a schematic view of another source with a single output in the sensor off 1 can be used;

7 a schematic view of an alternative source and a detector arrangement for use in the sensor 1 ;

8th and 9 schematic views of in vivo sensors according to further embodiments of the present invention, which are particularly useful for measurements in interstitial fluids or in a gel; and

10 a flow chart of the steps for processing spectra obtained from the in vivo sensor according to the invention.

DETAILED DESCRIPTION OF THE INVENTION

The The present invention is based on the discovery that the concentration of a chemical constituent of a body fluid in vivo can be when electromagnetic radiation or light energy (hereafter "light") with a multitude different infrared wavelengths and in particular near-infrared wavelengths the body fluid is guided so that the light of the many different wavelength a substantially collinear path to each other in the passage through the liquid occupies. It was also discovered that the concentration of glucose a chemical component of the body fluid is that can be accurately detected in this way and that the level of glucose in a blood vessel (e.g., a vein) in one vascular membrane or in an interstitial fluid or a gel (hereafter "interstitial Liquid ") are measured can. It has also been discovered that for accurate determination of concentrations of chemical constituents of body fluids, e.g. glucose, in fast changing anatomical structures, such as veins or ..., for sending and detecting all of the different wavelengths of the light in a substantially single period necessary are.

1 FIG. 10 is a schematic block diagram of an implantable sensor system. FIG 10 for analyzing the chemical constituents of body fluids in a mammal, according to a first embodiment of the present invention. This embodiment detects the concentration of a chemical constituent in blood flowing through a vein, and is particularly useful for detecting the concentration of blood glucose levels. The physi cal construction of an embodiment of the sensor system 10 is in 3A - 3C shown and discussed below. The components shown in the figures are not to scale.

In 1 contains the sensor system 10 (i) a light source 12 for sending a variety of different infrared wavelengths into the vein 14 , (ii) a detector 16 to detect light coming from the light source 12 was sent out and with the blood in the vein 14 and outputting a signal containing spectral information about blood components, and (iii) a processor 18 for evaluating the spectral information contained therein and for determining the concentration of at least one chemical constituent in the blood from the spectral information. An ad 20 and an alarm 22 are with the processor 18 connected to data and alarm conditions from the processor due to the detected concentration 18 to obtain. The output of the processor 18 Optionally, it can also be used for feedback control, such as automatic control of a drug dispenser 23 , A kind of drug donor 23 may be an externally attached or internally implanted insulin infusion pump.

The light source 12 and the detector 16 are positioned extravascularly and disposed opposite and facing each other so that the output of the light source 12 approximately through the middle of the vein 14 runs. The output of the light source 12 and the input of the detector 16 are near or in contact with the outer wall of the vein 14 as more fully related to 2 is described. In a suitable embodiment of the invention, wherein the diameter of the. vein 14 between 0.3 and 1.0 mm, is the distance between the output of the light source 12 and the vein wall and the distance between the input of the detector 16 and the vein wall is less than 100 μm (0.1 mm).

Depending on the needs of the patient and the sophisticated design and miniaturization of the system components will be some or all components of the sensor system 10 implanted. In 1 are the light source 12 and the detector 16 implanted in the person and the processor 18 is outside the person. The processor 18 can be worn for example on the wrist or belt of a person. The implanted components communicate via a suitable telemetry loop with the external components, such as a pair of RF receivers 24 and 26 which are associated with the implanted or non-implanted components. Alternatively, the processor 18 implanted and directly to the output of the detector 16 be connected, and the ad 20 and the alarm 22 can be outside the person. In this configuration, the processor 18 via rf receiver connected to the output of the processor 18 or with the inputs of the display 20 and the alarm 22 connected with the ad 20 and the alarm 22 communicate. If the output of the processor 18 As feedback control is used, the drug donor may become 23 either outside the person or in person.

An important feature of the invention is the light source 12 which emits light having a plurality of different infrared wavelengths, the light of the wavelengths having a substantially collinear light path to each other. This collinearity is important because of the sensor 10 In order to measure relatively dilute substances in a body fluid, such as glucose in the blood, relative differences in absorbency on the order of one part at 10 ° from one wavelength to the next must be demonstrated. If the light paths are not substantially collinear, variations in scattering, absorption, or path length differences may mask the desired spectral information.

1 shows a preferred light source 12 for generating an output that is particularly useful for measuring blood glucose levels.

The light source 12 contains a light-emitting device 28 in the form of a laser diode array having a plurality of individual laser diodes, for example, thirteen individual laser diodes 30 ₁ - _{13 13} , Every laser diode 30 generates a discrete wavelength λ ₁ -λ ₁₃ . Sometimes it may be necessary to check the output of one of the diodes 30 subjected to spectral filtration to obtain an output of higher spectral purity. The state of the light-emitting device 28 and the state of the individual laser diodes 30 ₁ - _{13 13} is from a regulator 29 controlled. The output of each laser diode 30 is with an input of a respective single mode fiber optic 32 ₁ - 32 ₁₃ connected, which has a core diameter of about 9 microns. The output of the respective fiber optics 32 ₁ - 32 ₁₃ is via a Bragg diffraction grating coupler 36 or a wavelength division multiplexer / demultiplexer having an input of a single-mode fiber optic 34 similar construction and with a similar core diameter as the fiber optic 32 connected. At the exit of the coupler 36 For example, the light from all individual wavelengths is essentially collinear. This coupling technique ensures that light at each wavelength has the same light path as the light of any other wavelength.

The output of the fiber optic 34 , the one with 38 is the output of the light source 12 , Part of the output of a fiber optic 34 becomes a calibration detector via a beam splitter (not shown) 40 diverted to from the processor 18 for the calibration of the output of the laser diodes 30 ₁ - _{13 13} to be used. The output of the calibration detector 40 is then fed to a preamplifier (not shown), an analog-to-digital converter (A / D converter) 41 and finally an RF transmitter / receiver 24 guided. Optionally, the output of the calibration detector 40 for adapting the light emitting device 28 spent energy (in 1 the laser diode arrangement) also with the regulator 29 be connected to receive a constant energy output from pulse to pulse. A suitable calibration detector is an InGaAs detector with a two-stage preamplifier. A simple temperature measuring device can sometimes be used instead of the calibration detector 40 to measure the output of each diode 30 ₁ - _{13 13} be used.

To detect the blood glucose concentration in the blood with an accuracy of ± 18 mg / dl over the physiological range of temperature, pH and blood chemistry, it is sufficient to use about thirteen different wavelengths, with two wavelengths (or more) used to normalize path length variations be used, a wavelength for the output offset and the remaining ten wavelengths are used for the glucose measurement. With accurate measurement of body fluid temperature, it may be possible to reduce the number of wavelengths. Ideally, the two wavelengths used to normalize the path length are not affected by "foreign" tissue surrounding the fluid to be measured (in this case, blood), that is, the difference in absorbance between the two wavelengths is "foreign". Tissue not affected. A wavelength set (in nanometers) suitable for the detection of blood glucose concentration at a blood path length of 0.5 mm or less is as follows:
1300, 2259, 1964, 2038, 1970, 1839, 2146, 1376, 2084, 1391, 2186, 1434, 2154 nm.

The first wavelength (1300) is for used the output offset while the next two wavelength (2259 and 1964) for the normalization of the path length be used.

If longer path lengths used, the optimal wavelengths shift slightly in the shorter Wavelength range. Alternatively you can suitable wavelengths for the detection of other chemical constituents be that in the infrared range or in the visible range lie. So there may be some wavelengths that are not in the infrared range lie. The exact number and value of the wavelengths depends on the component of interest and the desired accuracy of concentration from. In general, it is preferable to use spectral ranges to avoid high water absorption. That's why the near-infrared area generally opposite the mid-infrared range is preferred.

The each chosen wavelength and spectral ranges are based on a reiterative tentative basis in connection with the determination of an in vitro algorithm (as discussed below), the absorption peaks of the constituents of interest, the water absorption, hardware considerations and other considerations, like others, for example disturbing Tissue, considered become. In the case of the glucose sensor, the range was between 900 nm and 2500 nm, because wavelengths over 2500 nm excessive water absorption showed and limitations were subjected to the device, and wavelengths below 900 nm showed too weak absorption. In the face of it were over 450 Individuals were sampled from each of which a FT-IR spectrum was recorded between 900 nm and 2500 nm.

• 1. Erhalt der besten Kreuzvalidierung, Kalibrierung und prädiktiver Standardfehler
• 2. Verwendung von möglichst wenigen Wellenlängen
• 3. Verwendung von möglichst kurzen Wellenlängen

There are several well-known techniques for correlating the spectra with the measured values of the constituents, including the Norris regression algorithm, the partial least squares approach, the regression of the major constituents, and neural networks. On the blood samples of the 450 people, the different techniques were tried on a trial basis in order to achieve the following goals:

• 1. Obtain the best cross validation, calibration and standard predictive error
• 2. Use of as few wavelengths as possible
• 3. Use of the shortest possible wavelengths

Everyone these algorithms leads finally to the choice of wavelengths and the corresponding algorithm, the measured spectra with the concentration of the components correlated. In the section "DETERMINATION THE IN VITRO ALGORITHM "below will be optimizing this algorithm with the Norris regression discussed. This is the preferred embodiment of the invention.

For measuring the blood sugar concentration in blood, in the preferred embodiment, each laser diode should be 30 so much energy that at least 1 mW to the photodetector 44 is delivered. In general, this means that the output from the source should be at least 10 mW at approximately 2200 nm.

The detector 16 Contains a collecting fiber optic 42 on the other side of the vein 14 , the Output of the light source 12 is arranged diametrically opposite to that of the light source 12 emitted light after flowing through the vein 14 catch. The catching fiber optic 42 has a larger core diameter than the fiber optic 34 the light source output, optimally from about 0.2 to about 1.0 mm, depending on the size of the vein used for the measurement and depending on whether the vein from the sensor 10 is compressed or not. The catching fiber optic 42 must be chosen so that it has good catching properties. Therefore, ideally, it should have a numerical aperture (NA) of 0.39 or greater for the measurement of high scattering liquids, such as blood. In terms of capture efficiency and hence signal-to-noise ratio is a large diameter core for the capture fiber optic 42 preferable. However, too large a fiber may be due to path length differences (through the vein) of the collected light in different regions above the aperture of the capture fiber optic 42 or the photodetector 44 (as discussed below) lead to nonlinearities. The output of the collection fiber optics 42 becomes a photodetector 44 For example, in a non-cooled InGaAs detector with a two-stage linear preamplifier. The photodetector 44 and the preamplifier should have (1) a high degree of linearity, (2) a large signal-to-noise ratio and (3) a sufficiently broad spectral bandwidth to capture the entire spectral band of the source 12 exhibit. For the accurate measurement of the absorption of glucose, the light energy levels required by the photodetector 44 were measured to be 10 ⁵ larger than the signal-to-noise ratio, leaving the photodetector 44 must be linear over 5 decades. When using infrared wavelengths, the requirement after 5 decades leaves little room between the noise threshold on the one hand and the degree of linearity on the other hand. In front of the photodetector 44 No spectral filter is required unless the photodetector can detect wavelengths longer than 3000 nanometers where blackbody radiation becomes significant. In the preferred embodiment of the invention, an electronic bandpass filter (not shown) is used to limit 1 / f noise and high frequency noise. The electronic bandpass filter usually limits a frequency of about 1 / 2T to about 2 / T, where T is the pulse width of a laser diode pulse. The photodetector 44 is preferably a device with an element that is sensitive over the entire range of received infrared wavelengths.

The output of the photodetector 44 gets to an A / D converter 46 sent and the digitized output signal from the RF transmitter / receiver 24 transmitted from the external RF transceiver 26 received and sent to the external processor 18 cleverly. The RF transmitter / receiver 24 and 26 preferably have a range of about one to about three meters. Alternatively, the transmitter / receiver 24 and 26 be replaced by infrared transmitter / receiver.

In some cases, there is no catching fiber optic 42 necessary. Instead, the photodetector can 44 the light after flowing through the vein 14 record directly. An advantage of using a capture fiber optic to receive the light is that it is compared to the photodetector 44 is relatively light and small, so the likelihood is less that the vein is no longer consistent. In a collecting fiber optic, less capsule tissue can accumulate than in a photodetector 44 because a fiber optic is mechanically easier to fix with respect to a vein than a photodetector, so that abrasions can be more easily avoided.

The output of the calibration detector 40 is after its amplification and digitization for transmission to the processor 18 with the RF transmitter / receiver 24 connected. The processor 18 thus receives a signal from the calibration detector 40 and the catch detector 16 , The processor 18 uses the measured strength of the two signals to make certain corrections as discussed below.

When a foreign object is implanted in a living body, tissue tends to accumulate on the article, even if the article is made of a biocompatible material. This tissue is known as "capsular tissue." Capsule tissue usually forms over time to a certain extent, after which its growth stops and reaches a steady state, usually within 4-6 weeks. 1 shows capsule tissue 48 at the exit 38 the fiber optic 34 and at the entrance of the collection fiber optic 42 , In the preferred embodiment of the invention, the fiber optics 32 and 34 be surrounded by a hermetic and biocompatible coat, and at the exposed ends of an infrared window, such as sapphire have. The sheath and window may be coated with a hydrogel or like material to minimize growth of capsular tissue. The jacket can be made of titanium or stainless steel or another hermetic and biocompatible material. With such measures, although rejection and growth of capsular tissue can be minimized, this tissue is likely to form on exposed surfaces and must be considered in the evaluation of spectral information.

When measuring fast moving in itself Anatomical structures, such as pulsating veins, have also been shown to require the spectra to be taken in a "substantially single time period" for accurate measurement (ie, the spectra must be substantially coincident in time), which requires that the plurality of diodes be different Wavelengths are pulsed in a "substantially single period". In particular, the spectra should be recorded within a time that is so short that fluctuations between signal and path length over the measurement time will be much lower than the expected signal of the constituent of the fluid. Experimental data show that the "substantially single time period" in veins and the like for the measurement of blood glucose should be less than about 100 μsec. As defined herein, a "substantially single period" means not a time but a period of time in short, it is essentially the same in terms of the changes that the fluid component undergoes during the measurement period. For measurements in anatomical structures without rapid changes, such as interstitial fluid or gel, longer periods are acceptable.

In operation, each diode 30 the laser diode array sequentially pulsed over a certain period of time, with a time-out between the different diode pulses. In one embodiment of the invention suitable for measuring blood glucose concentration in blood vessels, a change in path length of about 10% over one second may be expected in certain veins as the vein pulsates due to blood flow. A blood glucose concentration of 100 mg / dl, with a blood path length of 0.5 mm, contributes an absorbance of the order of 0.7 × 10 ^-4 at 2273 nm, one of the strongest bands of glucose in the near infrared region. If two consecutive wavelength measurements with an accuracy of one part in 10 ^{5 are} desired, the measurements must be made within 0.1 milliseconds of each other to obtain a temporally coincident measurement. This means that the entire spectra must be recorded within 0.1 milliseconds. In an embodiment of the invention that satisfies this criterion, each diode becomes 30 pulsed sequentially for approximately 3 .mu.sec., with a timeout of 3 .mu.sec between the various diode pulses. Thus, a thirteen wavelength embodiment requires 75 μsec (0.075 milliseconds) for a complete pulse cycle. The pulsation of the diodes is controlled by the controller 29 controlled.

To further minimize time artefacts, diode pairs should be used 30 that are very close in wavelength or that are used for path length calibration, are pulsed one after the other. For the sample wavelengths listed above, the diodes become 30 wavelengths of 1964nm and 2259nm are pulsed side by side, as with the diodes 30 with the near wavelength pairs 1964 nm and 1970 nm and 2154 nm and 2146 nm, respectively.

To improve the signal-to-noise ratio, a number of repeated spectra N can be recorded over a relatively short period of time. The period should be so short that there is no significant change in the concentration of the ingredient. For a blood glucose sensor, this period can be several minutes. The processor 18 would then either average the N spectra or calculate the concentration of the component and then average the N predicted values of the component (s). Such a scheme yields a square root N improvement in signal-to-noise ratio. With changes in path length, pH, or temperature in the N samples, it is generally preferable to average the predicted concentration of ingredients rather than averaging the spectra and then predicting the concentrations.

In 1 (and 3C . 4C and 5 - 7 as discussed below), the fiber optics function 34 and 42 as a light delivery system for the active elements of the source 12 and the detector 16 , and they provide flexibility in the arrangement of the sensor components in the housing. In 2 has the fiber optic 54 a similar light-emitting function. If other means are used to adequately mix the light from the plurality of different source wavelengths, one or all of the fiber optics may be used 34 . 42 and 54 can be eliminated and light can be received or delivered directly from the respectively active sensor components. The output of the coupler 36 for example, directly into the vein 14 are discharged and the photodetector 44 can be arranged so that it emits emitted light through the vein 14 flows, receives directly. Nonetheless, the preferred embodiment of the invention uses fiber optics to deliver the light to the active sensor components.

2 shows an alternative embodiment of a light source 12 ' , In this embodiment, the light output of the laser diode array becomes collinear by mixing the beams in a long fiber optic. In particular, the outputs of the individual laser diodes 30 ₁ - _{13 13} into individual multimode fiber optics 50 ₁ - 50 ₁₃ guided. A suitable multi-mode fiber 50 has a core diameter of about 100 microns and forms a fiber bundle 52 with a diameter of approx. 0.5 mm. The output of the multimode fiber bundle 52 gets into the input of a mononuclear multimode fiber optic 54 guided, which has about the same diameter as the fiber bundle 52 , The mononuclear multimode fiber 54 has a significant length, for example a length of more than 1 meter, so that substantial wavelength mixing can occur as the light beam travels to the exit. (Although "wavelength mixing" is used, the individual diodes are pulsed sequentially rather than simultaneously.) To minimize the long fiber optic 54 required space can be the fiber 54 to be rolled up into tight loops.

In 1 must be the single mode fiber optic 34 have no significant length, because the wavelength mixing in the coupler 36 takes place, so that the light of the multiplicity of wavelengths already at the entrance of the fiber optics 34 is mixed. The for the embodiment of 2 The detector used is not shown, but it should be similar in diameter to the fiber optic 54 ,

3A shows a sectional view of a first embodiment of an annular device 60 for attaching implantable parts of the sensor system 10 and maintaining a fixed relationship between the output of the light source 12 and the input of the detector 16 , 3B shows a lateral plan view of the sensor 3A , Both the source 12 as well as the detector 16 are in 3B not to be seen. The source 12 is in 3B shown as a phantom. In 3A and 3B runs the vein 14 through the middle of the device 60 , The device 60 is preferably made of titanium or another suitable biocompatible material. The light source 12 and the detector 16 are in a pocket of the device 60 attached and their respective exits and entrances are through windows 62 and 64 made of biocompatible and infrared transparent material such as sapphire. The window 62 and 64 and the annular portion of the device 60 may be coated with a hydrogel or other polymer to minimize the growth of capsular tissue. To further inhibit the growth of capsular tissue, the device should 60 be designed so that sharp edges are minimized. The device 60 can be constructed in two C-shaped halves, which are at the joints 66 and 68 after placement around the vein 14 be connected to each other.

As discussed above, the light emitted by the source can be directly into the vein 14 blasted and detected directly from the respective active sensor components, or fiber optics can be used as a light delivery system for the active elements of the source 12 and the detector 16 be used. 3A shows the direct transmission. 3C shows a second embodiment of a fastening device 82 working with the light delivery system, where the output and input ends of the fiber optics 34 and 42 each on the device 82 are attached. In the light delivery system are the active components of the device 82 including the light source 28 and the photodetector 44 , from the device 82 arranged away.

The sensor 10 may also be used to measure the concentration of a liquid component in a vascular membrane, such as in a peritoneal membrane. 4A shows a sectional view of a third embodiment of a direct transmission device 84 for use with a vascular membrane 86 , 4B shows a lateral plan view of the sensor 4A , In 4A and 4B is the device 84 generally U-shaped as in 4A oriented (when rotated 90 degrees to the left or right C-shaped), leaving a slot 88 Are defined. The vascular membrane 86 is in the slot 88 arranged. Implantable parts of the sensor system 10 are in pockets of the device 84 on both sides of the slot 88 arranged. The device 84 so holds a firm relationship between the output of the light source 12 and the input of the detector 16 upright. The device 84 can the same coatings and windows 62 and 64 and consist of the same materials as the device 60 out 3A , Unlike the device 60 , for which a two-part housing part is required, the housing part of the device 84 optionally consist of a single integral U-shaped material. 4C shows a fourth embodiment of a device 90 for use with a vascular membrane 86 using the light emitting system, the output end of the fiber optics 34 respectively. 42 at the device 90 is attached.

In 1 and 2 can the light source 12 every light emitting device 28 which provides a light output with the required energy, wavelengths and spectral sharpness. Glucose is a dilute component of blood. Due to this fact, there are only a limited number of light-emitting devices for the measurement of the blood glucose concentration that fulfill all these criteria and have acceptable efficiency.

Light-emitting diodes (LEDs) are a suitable light source, which instead of one or more of the laser diodes 30 can be used. Some LEDs require a spectral filter, such as a narrow band filter, while others may be used without spectral filters. A disadvantage of LEDs is the fact that they are less efficient than laser diodes, especially after providing the required spectral filtering. In addition, LEDs are devices with high divergence, which, with efficiencies well over 20%, are difficult to achieve with fiber optics connect. Some advantages of LEDs over laser diodes are that they are inexpensive and the full spectrum from 800 nm to 2500 nm is commercially available.

The perhaps most important decision in the construction of a spectroscopic chemical infrared sensor concerns the selection of the spectral region (s) and wavelengths, especially when a white light source impractical or substantially non-optimal. The choice of wavelength and spectral region (s) ultimately determines the hardware and the Software (in vitro algorithm). For an in-vivo setup, the absolute minimum number of wavelength generally three points (unless other means are used can be to Background, scattering, extracting absorption by other ingredients, such as. the use of the pulsatile component, as in the pulse oximeter). At least one point will be for the path length calibration need and two more points are needed to measure the height of the absorption peak. at complex media, e.g. in blood, where there are a number of overlapping Components, the number of required wavelengths decreases considerably to. As a general rule, one can assume that one or two spectral points per chemical component, its concentration independently fluctuates and a spectral structure in and around the for the determination of the constituent of interest used have absorption peak added. For the Determination of blood sugar concentration in the near infrared range is needed additional Spectral points to compensate for overlapping globulin, cholesterol, Albumin, urea, hemoglobin, uric acid and of course water. About that In addition, there are undoubtedly other chemical species that are balanced Need to become. The means for selecting the wavelengths are discussed below.

The selection of the spectral region (s) coming from the source 12 are to be emitted depends on the liquid component to be measured, the available light sources and suitable detectors, the water absorption and noise considerations. In general, one selects spectral regions in the vicinity of the absorption band (s) of the component to be measured or of the components to be measured. For example, the strongest glucose bands are around 10,000 nm, followed by the bands at about 2240 and 2080 nm, and finally followed by an even weaker band at 1600 nm. Many other blood components show similar absorption patterns. On the one hand, water absorption and device limitations, in particular the availability of diodes and uncooled sensitive detectors, over 2500 nm become very problematic. As discussed above, the photodetector must be 44 have a five decade dynamic range to accurately detect glucose.

at an implantable device that are not cryogenically cooled can reduce the required linearity when using wavelengths above 3000 nm for Low concentration components because uncooled detectors above this wavelength have no more suitable features.

In addition, will the water absorption above 2500 nm significantly more problematic. Light energy at 10,000 nm learns a loss of over 10 dB through a water layer of 0.1 mm, making it very difficult will provide an acceptable signal-to-noise ratio for an implantable device To get device. For the measurement of glucose and most other blood components Therefore, it is generally preferred to have light of one wavelength below 2500 nm, even if it is for most blood components gives significant spectral information over 2500 nm. Even under 2500 nm, there are some water absorption bands that increase at longer path lengths cause excessive weakening; a at 1899 nm and one at 1454 nm.

In 5 can the light emitting device 28 out 1 also a tunable laser 70 which is programmed to give a repeated output at the desired wavelengths. A tunable laser inherently emits collinear light at multiple wavelengths because it has only a single output. The output of the tunable laser 70 can thus be directly connected to the input end of the fiber optic 34 out 1 get connected. As with the laser diode arrangement 28 the entire desired spectrum should be recorded substantially within the same time frame. The calibration detector 40 , the catch detector 16 , the processor 18 and the relevant components are in 5 not shown and can be the same as in 1 ,

Furthermore, the light emitting device 28 a white light source, for example a white light source with a maximum output energy in the near infrared range. A white light source also inherently emits light at wavelengths that are collinear with one another because there is only a single output. The output of the white light source can thus also directly with the input end of the fiber optic 34 out 1 be connected. When using a white light source, the output of the white light source must be spectrally separated to obtain spectral information for the desired wavelengths. In a technique that in 6 is a tunable spectral filter 72 between the white light source 74 and the photodetector 44 arranged, for example between the off passage of the white light source 74 and the input end of the fiber optic 34 , The tunable spectral filter 72 is programmed to repeatedly emit light at the desired wavelengths. As with the laser diode arrangement 28 The entire desired spectrum must essentially be recorded within a single period of time. To conserve energy, the white light is only pulsed when a spectrum is to be recorded. The tunable spectral filter 72 can be implemented either by a turntable with different spectral optical bandpass filters or by a rotating grating that transmits the desired spectra into the fiber optic 34 couples.

In another technique, in 7 is visible, is a solid grid 76 at the entrance of a photodetector 78 arranged to spectrally divide the broadband white light into different spatial packets with respect to the thirteen wavelengths. (The photodetector 78 replaces the photodetector 44 out 1 .) The photodetector 78 consists of a series of photodetectors 80 ₁ - 80 ₁₃ , each photodetector 80 detects one of the thirteen wavelengths. Because every photodetector 80 ₁ - 80 ₁₃ in 7 with the same collection fiber optics 42 is connected, the requirement that light overlaps spatially at each wavelength, is met.

The sensors in 5 and 6 contain similar calibration detector, catch detector and processor components as in FIG 1 , Only the collection fiber optics 42 can be seen in these figures. Also the sensor in 7 contains similar calibration detector and processor components as in FIG 1 , The catching fiber optic 42 in these figures, a multimode fiber, typically 0.5 mm or larger in diameter, is not a problem if non-linearity due to path length deviations across the fiber opening presents no problem.

Light having a plurality of different wavelengths that flow substantially over the same light path (and thus are collinear) is most easily reconciled with the source configuration in FIG 1 where a plurality of single mode fibers are connected to a single mode fiber via a Bragg refraction grid. The extent of the necessary spatial overlap of the light at the different wavelengths relative to each other depends on the strength of the signal generated by the desired liquid constituent and the expected degree of spatial inhomogeneity. Blood and most other human tissues tend to be very inhomogeneous in space, which is mainly due to scattering. For example, a 1 mm blood path with a glucose concentration of 100 mg / dl causes peak absorption on the order of 10 ^-4 at 2273 nm (due to glucose). Two wavelengths separated by 0.1 mm could produce an apparent signal of 3 × 10 ^-4 if the variations in the scattering are of the order of 1% per mm. With a spatial variation in the scattering of the order of 1% per mm and with a separation of the two wavelengths of light by 0.1 mm, the detector would have 16 for example, difficulty seeing the glucose absorption peak over the variation in scattering.

As discussed above represents a white light source in terms of collinearity not a problem, but care must be taken that the spatial Relationship between the source and the detector during the recording time of the Spectra remains constant.

to Generation of the desired Collinear source beam of LEDs or individual laser diodes can also Other techniques without grids or long fiber optics are used. The LEDs or laser diodes can for example physically very close to each other so that the Light path of the light of each wavelength is substantially collinear each other relative to the light receiving liquid. For improvement of collinearity can Also a negative lens or aperture can be used.

In an alternative embodiment of the invention, each individual light source in the light source 12 be frequency modulated with a unique predetermined frequency. When using a frequency modulation scheme, all the individual light sources can be pulsed simultaneously, which guarantees that the output of the light source 12 temporally coincident at the plurality of wavelengths.

The sensors off 1 - 7 are extravascular, but they can also be used intravascularly. However, intravascular sensors are not preferred because they may be associated with clinical and technical complications such as thrombosis and rejection.

alternative could the sensors are placed in a bone. The risk of a Infection is much higher in bones than elsewhere Inside of the body.

The sensors off 1 - 7 are arranged to measure light from the source directly through the vein 14 or another anatomical structure flows. However, the sensors can also be arranged so that they detect reflected or scattered light or combinations of transmitted, reflected or scattered light. At present, it is assumed that a precise coupling between the source 12 and the test sample (Ve ne 14 in 1 ) is more critical when performing reflected measurement rather than transmission measurements.

It It is further assumed that the signal-to-noise ratio, the obtained in the blood glucose measurement with reflectance spectroscopy, lower is the signal-to-noise ratio in transmission spectroscopy in the Near-infrared range.

The sensors off 1 - 7 are suitable for a variety of environments in vivo, such as veins, vascular membranes, muscles, arteries, and the like. In order to obtain accurate measurements of fluid constituents in certain types of environments in vivo, it may be necessary to protect the sensor components (particularly the output of the source and the detector input) and / or the sensing area from contact with undesirable components, particularly in the case of Measurement of fluid components in interstitial fluid.

It It is known that the glucose concentration in interstitial fluid accurately correlated with blood sugar levels and for diabetes control and Treatment can be used. For example, it is known that the glucose levels in interstitial fluid by about 20 minutes lie behind the glucose levels in the blood. When measuring the glucose concentration owns the interstitial fluid numerous advantages over Blood. Interstitial fluid and gel have little or no hemoglobin and red blood cells, which cause both artifacts in spectroscopic measurements. Erythrocytes cause, for example, scattering. By measurement the interstitial fluid can other sources also possible Artifacts in blood measurements are eliminated, including artifacts through pulsating blood vessels and the Confusion between glucose spectra and hemoglobin spectra. Path length variability is included Measurements in interstitial fluid less significant than blood measurements. Interstitial fluid may also have a lower protein concentration than blood, making glucose easier and with higher predictive accuracy from the blood spectrum can distinguish. Thus, for accurate concentration measurements of less wavelengths of the source necessary.

8th shows a sensor 100 responsible for monitoring a fluid component in interstitial fluid 102 suitable. Interstitial regions contain interstitial fluid composed of an interstitial fluid and a solid material mainly composed of collagen fibers. At times, cell debris can float in the interstitial region. The cell debris, adjacent tissues and solid material (hereafter "tissue and cell debris") can reduce the accuracy of the measurements as they enter the measurement zone 100 is designed so that it prevents these components from entering a measuring zone or chamber 104 between a source 106 and a detector 108 penetration. The sensor 100 contains a barrier or a coat 109 with a wall or barrier 110 and end caps or lids 112 . 114 , The wall 110 consists of an inert biocompatible material with a porosity chosen so that the interstitial fluid and gel can pass through it (good fluid exchange) while blocking adjacent tissue and cell debris. Suitable materials for the production of the wall 110 are polytetrafluoroethylene (such as TEFLON ^®) having a plurality of perforations, porous ceramic and a biocompatible mesh material. The lids 112 and 114 may be made of an inert hermetic biocompatible material, such as titanium. The coat 109 is preferably cylindrical, but other shapes are suitable.

A barrier 110 a biocompatible mesh keeps adjacent tissue from passing through the barrier 110 but it does not prevent capsular tissue from growing on the sensor components unless the mesh is extremely fine and prevents the proteins and leukocytes from penetrating.

Alternatively, the barrier 110 consist of a biocompatible wire mesh that forms a shield or cage that is so fine that it prevents adjacent tissue from entering the sensor chamber 104 pushes and disturbs the light transmission. Of course, this scheme does not prevent the growth of capsular tissue on the sensor components.

8th shows adjacent tissue and cell debris molecules 116 that are not through the barrier 110 can penetrate while glucose molecules 118 in the interstitial fluid 102 through the barrier 110 and in the chamber 104 can penetrate. Because the interstitial fluid 102 free through the barrier 110 penetrates, is the glucose concentration in the sensor 100 and in the area around the sensor 100 essentially the same. In 8th can the source 106 contain one of the direct light emitting sources described above or the source 106 may be the output end of a fiber optic connected to the light emitting source. The detector 108 may also be one of the photodetectors described above, which are compatible with the respective source, or the detector 108 may be the input end of a trap fiber optic connected to one or a plurality of photodetectors. if the source 106 and the detector 108 The fiber optics are the light emitting sources and the photodetector (s) can be from the cladding 109 away lie. The source 106 and the detector 108 are at the respective ends of the mantle 109 attached.

Regardless of whether the source 106 and the detector 108 in 8th coupled directly or via a fiber optic delivery system, it is to minimize the degradation of the source 106 and the detector 108 necessary, they with hermetic coats 126 respectively. 128 and biocompatible, hermetically sealed transparent windows 130 and 132 to provide. The hermetic coats 126 and 128 consist of a suitable inert, biocompatible material, such as titanium. The coats 126 and 128 prevent all substances and components of the interstitial fluid from coming into contact with the sensor components. To further minimize the growth of capsular tissue, a hydrogel or similar material layer can be used on the windows 130 and 132 and the coats 126 and 128 to be added.

TEFLON ^® is not wetted. So if TEFLON ^® for the production of the coat 109 can be used, the sensor chamber 104 before the introduction of the sensor 100 be filled into the interstitial fluid with alcohol or saline. Alternatively, the sensor chamber 104 immediately after insertion with the solution pressurized. In either case, the solution gradually self-exchanges with the liquid outside the chamber and the interior of the chamber attains equilibrium with respect to the outside of the chamber. An insertion mouthpiece 119 can through one of the lids 112 or 114 be guided to facilitate the filling of the chamber.

9 shows another embodiment of an open sensor 120 which is suitable for use in an interstitial fluid. This sensor 120 contains a source 106 and a detector 108 similar to those in 8th that of similar coats 126 and 128 and optically transparent windows 130 and 132 are protected. The output of the source 122 and the input of the detector 124 are separated by a predetermined distance. A rigid coupler 125 who is wearing the coats 126 and 128 is fixed, keeps the predetermined distance between the source 122 and the detector 124 upright.

In 9 can the source 122 a direct light-emitting source, or a variety of sources as described above, and the detector 124 For example, a single photodetector or a plurality of photodetectors may be as described above. Alternatively, the source 122 and the detector 124 Be fiber optics, which are connected to respective light-emitting sources and photodetectors, the fiber optics acting as a light delivery system. In the latter case, the light emitting sources and the photodetectors may be from the open sensor 120 be located away.

In each of the in 1 - 9 In other embodiments, other techniques for preventing or suppressing the growth of capsular tissue on or about the sensor components may also be used. Examples of possible suitable schemes are the use of heat, electric fields or electric current, either constant or sporadic. For example, protein denatures at 60 ° C. A daily heat radiation over a few seconds may therefore be sufficient to prevent growth. An alternative physical barrier technique that prevents adjacent tissue from getting between the source and the detector is placement of the sensor in the interstitial fluid so that the measurement chamber is significantly removed from adjacent tissue.

The at the output of the photodetector 44 or its equivalent, spectral information is sent to the processor for spectral analysis 18 forwarded. As discussed above, the processor receives 18 also a signal from the calibration detector 40 , When measuring blood components with the sensor 10 out 1 contains this from the photodetector 44 received signal spectral information from all substances, the light from the source 36 have received. Therefore, there is spectral information from the fluid of interest (eg blood) and spectral information from foreign tissue, including the tissue wall of the vein 14 and capsule tissue 48 , Additional spectral information may be from the interstitial fluid and the gel between the sensor 10 and the vein 14 available. Hereinafter, the spectral information relating to the liquid of interest is referred to as "spectrum of liquid" or "liquid spectrum", while the spectral information relating to the foreign tissue, including the vein wall of the vein 14 and capsule tissue 48 and all solids in the interstitial gel are referred to as the "spectrum of tissue" or "tissue spectrum". The combination of both spectra is referred to as "composite spectra." A goal of the processor 18 is the extraction of the liquid spectra from the composite spectra and the use of the liquid spectra to measure the concentration of the liquid constituent of interest or alternatively to correct for the contributions of the tissue spectra to obtain accurate predictions of the concentration of the constituent.

10 is a flow chart of the steps using the sensor 10 out 1 in the processor 18 be performed. In general, the processor takes 18 the one from the detector 16 out given spectral data and creates an in vitro algorithm to determine the liquid constituents of interest. The in vitro algorithm is generated by correlating Fourier Transform Infrared (FT-IR) spectroscopy transmission measurements of fluid samples of a statistically significant number of persons with the fluid samples measured by conventional techniques (ie, techniques other than the techniques of the invention) , A statistical model is applied to the FT-IR data and the in vitro algorithm is optimized for the least squares prediction error and the least number of spectral points. The in vitro algorithm therefore determines the spectral points to be measured and their relative weighting of the absorption capacity. The algorithm is then applied to the in vivo measured spectral data to determine the concentration of the selected liquid components.

In 10 the processor calculates 18 the absorption spectra by means of the log of the quotient of the output of the detector 16 to the output of the calibration detector 40 for each wavelength (step 100 ) and adding a correction factor (step 200 ) for adjustment for (1) variability in response between detectors 16 and 40 and (2) chromatic differences in the optical systems. The resulting signal is the composite absorption spectrum or simply the "composite spectrum" as described above.

Next, the processor sets 18 an in vitro algorithm as in steps 300 - 600 in the dashed lines in 10 defined to the composite spectrum. The first step in the in vitro algorithm is to apply a pathlength correction to the composite spectrum (step 300 ). One method of performing this step is to divide the composite spectrum by the difference in absorbance of the composite spectrum at two wavelengths chosen for path length normalization. Another method is to divide the composite spectrum by the "calculated path length" as determined by the partial least squares algorithm.

After completion of the path length correction, the processor sets 18 a baseline correction to the composite spectrum (step 400 ). One method of performing a baseline correction is to subtract the absorption (also path length correction) at a third wavelength from the path length corrected composite spectrum. The resulting composite spectrum is now corrected for the path length and the output value. Alternatively, the baseline correction may be made before the path length correction by the order of the steps 300 and 400 is reversed.

After that, the processor calculates 18 an "uncorrected" concentration of the desired constituents by multiplying the compound spectrum with beta coefficients corrected by the path length and the output value (step 500 ) and adding a constant to the result (step 600 ). The concentration is said to be "uncorrected" because the beta coefficients are applied to the composite spectrum containing foreign tissue spectra and not just the liquid spectra of interest.

DETERMINATION THE IN VITRO ALGORITHM

The preferred embodiment The invention uses the Norris regression model for creation the optimal in vitro algorithm. In the preferred embodiment of the invention the choice of wavelengths and their corresponding beta coefficients also on an in vitro basis and ideally determined on a cross-population basis. Liquid samples suitable for the measuring liquid representative are, by a statistically significant group of people, for example It should be carefully followed by a few hundred people Care should be taken that the chemistry of the fluid samples is not through Contact with air, time and the like is changed. Ideally, span the chemical components of the desired constituents and others Ingredients having influence on the spectra essentially the whole physiological area for healthy and sick persons. It is also important that other physiological Variables, such as Temperature and pH, the entire physiological range cover, so that the calibration model over a wide fluctuation range of health conditions true.

To Taking the fluid samples becomes a spectrum of the liquid with a conventional one spectrophotometric technique, for example, by a scanning or FT-IR spectrophotometer with a white light source. Furthermore Also, the concentrations of the desired ingredients (e.g. Glucose), with a conventional one non-spectroscopic means, e.g. with a chemical Assay. For the path length correction should be cuvettes with several path lengths used, which measured the range of ultimately in vivo path lengths cover.

When creating the in vitro algorithm for use in step 300 . 400 and 500 For example, the wavelengths used for the path length correction, the baseline correction, and the beta coefficients are preferably reiteratively optimized to provide the Least Squares get customization. Before this optimization, however, the wavelengths must first be estimated.

For the path length correction, for the at least two wavelengths necessary, the first wavelength can be calculated by calculating the correlation coefficient for the from the in vitro samples obtained spectrum with the measured path length and Choice of the coefficient, which gives a correlation as close as possible one obtained become. The estimation the second wavelength can be calculated by calculating the correlation coefficient at each wavelength of the Spectrum divided by the difference between the absorption at every wavelength and the absorption at the first wavelength with the measured path length and again choosing a correlation coefficient as close to unity as possible become.

The actual Path length correction involves the division of the absorption spectrum by the difference in the absorption at the first and second wavelength, like described above.

For the baseline correction can be the first estimate for the Offset Wavelength be determined empirically by checking the spectra and then a wavelength in the range the greatest flatness chosen becomes.

For the beta coefficients can the first estimates for the wavelength by calculating the correlation of the measured components with the spectrum and choice of wavelengths with correlation coefficients, the possible close to +1 or -1 be set. The first wavelengths can be guessed even better if wavelengths which are the largest beta coefficients (in absolute values), as with the partial least squares (PLS) Routine determined.

The Beta coefficients are chosen by finding coefficients that are a best least square Adaptation of the spectra at the initially estimated wavelengths measured concentration of the components. Once the beta coefficients for the optimized "estimated" wavelengths The process is repeated, sequentially varying the wavelength and the beta coefficients are optimized again to be the best Least Square customization, with an optimized set more discrete wavelength and corresponding beta coefficients is obtained. Finally approaches the standard error of the calibration is asymptomatic to a minimum value.

The method described above provides an in vitro algorithm for calibrating the measured spectra at a finite number of wavelengths to the measured concentration (step 300 and 400 ) and for calculating an uncorrected concentration of the desired constituents (step 500 and 600 ). After these steps, it is still necessary to correct tissue spectra and obtain "corrected" concentrations.

at an implanted sensor becomes the fluid that the desired Contains ingredients and provides the spectrum to be measured, inevitably surrounded by foreign tissue. There are variations from person to person in terms of chemical composition and the thickness of the capsular tissue and in other foreign tissues. in the Changed in general the foreign tissue is the spectrum recorded by the sensor. Unless one or the other method is used to add the foreign tissue compensate, leads it made a mistake in the calculated concentrations of the chemical Ingredients because of the in vitro algorithm using fluid samples in vitro (containing no foreign tissue) has been developed. So long but the capsule tissue remains stable over time, this can Problem completely be resolved.

METHOD FOR GENERATING AN ALGORITHM FOR COMPENSATING FOREIGN FABRICS

In a preferred compensation method, for each individual patient, a best linear fit of the uncorrected concentration of the ingredients (output from step 600 ) to the concentrations measured in vitro by conventional means, whereby a set of best linear coefficient of adaptation is obtained. This measurement should be taken after a sufficient period of time for the capsule tissue to reach a steady state (usually more than four weeks). These coefficients are in the processor 18 for step 700 saved. The coefficients are applied to the uncorrected concentration of the components to obtain predicted concentrations that are "corrected" for foreign tissue (step 700 ).

For a very accurate Calibration of the uncorrected concentrations to the corrected ones Concentrations should be the concentrations of the desired Ingredients for a particular patient span the probable range of measurement.

If For example, if the sensor is a blood glucose sensor, the calibration should be done via a Range of glucose levels from 25 mg / dl to 300 mg / dl.

ALTERNATE EMBODIMENTS FOR DETERMINING THE IN VITRO ALGORITHM

One alternative way of determining the values for the in vitro algorithm is the application of a multivariate analysis with PLS optimization on the data obtained in vitro. This will be exactly the same data in vitro (spectra, concentrations and path lengths) on a cross-population basis as described above in the section "DETERMINATION OF THE IN VITRO ALGORITHM " collected. In one example, a Galactic PLS routine corrects Industries, Salem, New Hampshire, by the path length variability through Use of the measured path length as the first component and the other measured components as second and further components. First, the entire spectral range be used. Thereafter, the routine determines the standard prediction error Cross calibration base and the beta coefficients for the path length and for each chemical ingredient.

Using this alternative route will take steps 300 . 400 and 500 in 10 changed as follows: step 300 (Path Length Correction) - (1) Determine the path length by applying the beta coefficients for the above-determined path length to the composite spectrum, and (2) Divide the composite spectrum by the calculated path length.

step 400 (Baseline Correction) - This is an optional step before or after step 300 can be carried out.

step 500 Apply the beta coefficients in the PLS algorithm as determined above to the composite spectrum corrected for the path length.

at the PLS technique results in the use of the entire spectral range, measured with a spectrophotometer, usually not the best Prediction error or standard error of calibration, and it is also not the most efficient method. Improvements in accuracy and efficiency of the in vitro algorithm are possible by choice of spectral regions, where the beta coefficients are greatest and by elimination of spectral regions where the beta coefficients are smallest. This Method includes the reiterative optimization of selected spectral ranges and the search for a limited number of spectral points, the sufficient cross-calibration and reasonable prediction error result. Ideally, it should be as possible selected few spectral points be, corresponding to a low cross calibration error the cost and complexity to minimize the sensor. For example, each spectral point correspond to another diode.

Other statistical models that are alternative to creating the optimal Algorithm can be used are the regression of the major components, neural networks and Combinations of the models described above. Second derived data be used to minimize problems with baseline shift and Spektralstrukturverstärkung used by scattering and path length variability.

ALTERNATIVE PROCESS FOR CORRECTION TO FOREIGN FABRICS

The most obvious way to correct for extraneous tissue is to select wavelengths for the in vitro algorithm where the composite spectrum is not particularly affected by foreign tissue. In many cases this is not possible. Therefore, a correction as in step 700 in 10 be performed.

Another possible method for correcting for extraneous tissue is to subtract a "representative spectrum" from the foreign tissue, an additional step between step 200 and 300 in 10 inserted. The representative spectrum of the foreign tissue would be subtracted from the composite spectrum to obtain a spectrum representative of the liquid alone. The "representative spectrum" (obtained as described below) would be specific to the patient 300 until step 600 would not be in the composite spectrum then 10 , but for the "liquid spectrum" are used 700 would be omitted because it would not be necessary.

The representative spectrum can be obtained in several ways. One way is to take a number of fluid samples of each patient in vitro, after the sensor has been implanted for a sufficient time for capsule tissue to reach steady state. Simultaneously with taking the fluid samples in vitro, the spectra of the corresponding composite spectrum would be picked up by the implantable sensor. The spectrum of fluid samples in vitro would be picked up with either a FT-IR, a scanning spectrophotometer or a special spectrophotometer similar to the implanted one. It can be assumed that the representative spectrum of foreign tissue from sample to sample is constant for each patient. The representative spectrum of foreign tissue can thus be found by looking for the spectrum that best suits the difference between the composite spectrum and the pro a constant times the spectrum of the corresponding fluid sample fits in vitro. The constant, which is multiplied by the spectrum of the liquid sample in vivo, is intended to account for variations in liquid path length in vivo due to, for example, changes in blood flow. The representative foreign tissue spectrum is then stored in the processor and used for the further step.

The Difficulty with this way is that extremely carefully ensured that the temperature at which the in vitro sample is taken, the same as that of the corresponding in vivo sample.

A Modification of the determination of the representative foreign tissue spectrum is the recording of the spectrum of the representative components of the Foreign tissue, such as Venous wall, collagen and the like, and the determination of the best linear combination of these components, the best on the difference between the compound Spectrum and the product of a constant times the spectrum of the corresponding fluid sample fits in vitro. Every other step is as described above.

Yet another way to compensate for the contributions of tissue spectra in the Measurement over a pulsating blood vessel, e.g. an artery, is the recording of the composite spectrum at the maximum and minimum of the pulse, and the subsequent subtraction the two Spektra. The subtracted spectrum is essentially that from blood alone, but there is also a small component due the change the thickness of the vessel wall.

DETAILS TO THE PROCESSOR

In 1 may be the corrected concentrations of the ingredients on the display 20 are shown, which trigger the alarm 22 or to control the medication dispenser 23 is used. The corrected concentration values and the date and time of the measurement may optionally be stored in a memory (not shown) of the processor 18 get saved. The processor 18 may be the implantable sensor 10 via telemetry signal that he should take measurements more frequently if the trend or the measurement of the components are problematic. Ideally, a clinician would set the parameters for alarm and measurement frequency.

The processor 18 also captures "outlier data", ie data that is out of limits due to sensor malfunctions, and when outlier data is received, the processor signals 18 the sensor 10 via telemetry, to perform another measurement, and he triggers the alarm 22 off, even if the following measurements are outside the specified limits.

The processor 18 can be sporadically calibrated against conventionally obtained measurements of liquid components to avoid long-term shift. If the sensor 10 For example, for blood glucose monitoring, the processor can 18 sporadically calibrated against blood glucose levels measured by a chemical assay. One form of this calibration is the supply of a new set of coefficients for the best linear fit in step 700 ,

If the processor 18 Used for feedback control, the output signal of the processor can be used for direct control of the dispenser 23 be used, or the donor 23 may have its own processor (not shown) for receiving the output signal and controlling the output dose.

The The present invention is particularly useful for detecting the glucose concentration in the blood and in interstitial fluid. The invention can but also to measure other body fluid components can be used, e.g. Proteins, cholesterol and other blood components.

It is for those skilled in the art will appreciate that the embodiments described above changed could become, without departing from the rough concept of invention. It goes without saying Therefore, this invention is not limited to the disclosed embodiments limited is, but that they are also modifications within the present Invention as in the appended claims includes defined.

Claims (10)

Implantable sensor for use in an in vivo spectroscopy device with a processor ( 18 ) for evaluating signals from the sensor and for determining the concentration of a constituent of a body fluid, the sensor comprising: a light source ( 12 ) for emitting light, and a detector ( 16 ) for detecting the emitted light, wherein the light source ( 12 ) and the detector ( 16 ) are arranged so that the emitted light interacts with the body fluid before it is removed from the detector ( 16 ) while the detector ( 16 ) emits a signal containing a spectrum of body fluid, wherein the light source ( 12 ) for emitting light of a variety unterschiedli wavelength, wherein at least one wavelength is in the infrared range, characterized in that the light from each of the plurality of different wavelengths has a substantially collinear optical path length through the body fluid relative to each other, and that the detector ( 16 ) can obtain a composite spectrum including spectra of the body fluid and spectra of the foreign tissue when the infrared light penetrates through the body fluid containing the component and foreign tissue, while the processor (FIG. 18 ) can perform a cleanup around the spectra of the foreign tissue. Sensor according to claim 1, characterized in that the processor does a cleanup around the spectra of the foreign tissue carry out can by doing a path length correction applies in the composite spectra, an output value correction at the composite spectra, the path length of the composite spectra and the output values of the composite Spectra multiplied by beta coefficients and a constant added to a concentration of the at least one component which is not adjusted to the tissue spectra, and by giving coefficients for the best linear adjustment to the unadjusted concentration of the at least one component of the body fluid applies to a concentration of the at least one constituent that is around the tissue spectra is cleaned up. Sensor according to one of claims 1 to 2, characterized that the light source for emitting light with different Wavelengths in a substantially single period is suitable. Sensor according to claim 3, characterized in that the essentially single period is less than about 100 microseconds is. Sensor according to Claims 3 to 4, characterized that the light source will light in a variety of pulses of the essentially single period. Sensor according to one of claims 1 to 5, characterized that the body fluid is blood is and the at least one component is glucose. Sensor according to one of claims 1 to 6, characterized that the implantable sensor is such that it is next to a vascular membrane can be implanted. Sensor according to one of claims 1 to 7, characterized that the light source is light with wavelengths in the near-infrared range sending out. Sensor according to one of claims 1 to 8, characterized that the detector detect reflected and scattered light can. Sensor according to one of claims 1 to 9, characterized that the processor is a record of a ratio of the output signal of the detector and the output energy of the light source of each wavelength to disposal is provided.