DE19850350C1 - Method and device for generating data for the diagnosis of the degree of damage to a patient's skin tissue - Google Patents

Abstract

The invention relates to a method and a device for generating data for the diagnosis of the degree of damage to a patient's skin tissue. DOLLAR A According to the invention, the task of finding a new possibility for assessing the degree of damage to human skin on the basis of spectrally decomposed video recordings, which allows an objective and reproducible evaluation of the skin damage regardless of the skin type and time of the skin damage, is achieved by illuminating with white light from the Surface of the skin tissue remitted light is simultaneously divided into at least four separate light paths, with successively spectrally selective edge filters coupling out spectrally defined wavelength ranges for the spectral channels, after the gradual spectral decomposition a narrow-band spectral filtering is carried out in each spectral channel before the optoelectronic image conversion, a cluster analysis the remission values are carried out in a parameter space, the centers and radii of the clusters being given degrees of damage by teaching with remission values Determined and stored under skin tissue patterns and an assignment of reflectance values of unknown skin tissues to these learned clusters of degrees of skin damage occurs, and areas of different skin damage degrees in the area of the absorbed skin surface are represented on the basis of the cluster assignment.

Description

The invention relates to a method and an apparatus for generating Data for the diagnosis of the degree of damage to skin tissue Patient. It is used in particular for medical purposes Initial diagnosis, follow-up, documentation and archiving of Burns in surgical clinical use.

In clinical diagnostics, abnormal skin changes, especially skin burns, today like 200 years after Debridement visually with the examining doctor's eye, i.e. H. subjective and diagnosed based on experience. Especially in the case of burns due to patient and environmental influences the diagnostic errors in general and depending on the experience of the doctor in the order of 30 to 50%. Very the extent of the damage often only becomes apparent after a few days or The time of the first operation can be seen.

In the search for objective processes, with ultrasound, Thermography, Staining methods, isotopes, NMR (Nuclear Magnetic Resonance) and laser doppler techniques experimented without the corresponding clinical success to be able to achieve, since the methods are mostly invasive (contamination of the Patients), too risky for the intensive care patients, too time-consuming or too are expensive and the reproducibility of the results is often inadequate. The previous work on reflection-optical methods was also not sufficiently successful and were therefore not put into practice. The Experiments in the visible wavelength range (Afromowitz et al., 1988)  failed in clinical practice, u. a. in that the error rate was too big due to the different skin types.

In US 5,701,902 an attempt was made to use fluorescence excitation (in the UV or visible spectral range) and simultaneous IR spectroscopy to characterize burns. However, this solution is also invasive (intravenously administered fluorescent dyes) and limited to spot measurements (≦ 1 mm ² ). Neither a progress check (reproducibility of the measuring point) nor large-scale recordings (an area of 10 cm. 10 cm would take about 6 hours) can be carried out.

From the US Pat. No. 4,693,255, large-area, pictorial is known Video recordings of skin burns by the Kinetics of changes in the appearance of the burned tissue by means of a marking dye introduced shortly before, preferably a fluorescent dye used for analysis. Because the procedure is invasive and the dye dynamics change within a time window of 1-20 Minutes, the procedure is already for instant diagnosis, and Follow-up checks are unsuitable.

US 4 170 987 describes a medical skin diagnostic system and ver drive with a swivel mirror and three detectors, on the simultaneous the same pixels of the patient's skin scanned in a line grid be mapped. There are black and white detectors behind different color filters (e.g. green, red, IR). The gray values of the detectors are after pre-processing contrast increase or dynamics area adjustment digitized and saved. From each of the three related stored numerical values per pixel  then ratios are formed which are shown on a color monitor False color image can be displayed or printed. Besides the fact that the scanning method of scanning basically The main disadvantage of mobile use is the high optical quality Adjustment effort and that due to the multiple intensity distribution to complain about the weakening of light. Because the adjustment effort actually only for beam splitting within a level to a manageable extent remains and the unwanted light attenuation with two divider mirrors already is considerably high, the system presented in US 4 170 987 actually only intensity images for up to three different wavelengths record if a pixel-synchronous image recording with sufficient Signal-to-noise ratio should be achieved

Another general disadvantage of all known solutions for the diagnosis of Degree of skin damage is that the measured values depend on the individual Skin type, from the location of the burn, from the time of measurement after the combustion, the lighting situation and the eventual Pretreatment (ointments) depends. Therefore, the absolute readings only a limited meaningfulness, so that most of the investigations carried out in described in the literature, using a corresponding standardization and the interpretation model have failed. The result is that the Measuring method by adding additional, partly alien species Measured variables were complicated, e.g. B. by additional measuring channels and / or invasive steps (dyes).

The invention has for its object a new way to Assessment of the degree of damage to human skin on the basis to find spectrally decomposed video recordings that are objective and reproducible evaluation of skin damage regardless of skin type and time of skin damage allowed. It is a special task of the  Invention to realize an objective diagnostic tool that is safe between degrees of combustion 2a (2, superficial) to 2b (2, deep) and 3 and healthy skin and scenes in the image that do not affect the patient, differs.

According to the invention, the object is achieved in a method for producing Data for the diagnosis of the degree of damage to skin tissue Patients with simultaneous imaging of the surface of the Skin tissue takes place in several spectrally different channels, derived image data generated from the spectrally different images and differently damaged areas of the skin tissue are visualized are displayed, solved that the skin tissue with a large area white light is illuminated in such a way that essentially only remitted Light from the surface of the skin tissue into the spectral channels Image acquisition arrives that the remitted light in at least four separate Light paths are split, being by successive spectrally selective Edge filters applied to optical surfaces of prisms, stepwise spectrally defined wavelength ranges for the spectral channels be coupled out that after the gradual spectral decomposition a narrow band spectral filtering in each spectral channel before optoelectronic image conversion is carried out, this Bandpass filtering characteristic changes in reflectance of different layers of skin when damaged takes into account that a cluster analysis in a parameter space of the reflectance values is carried out, the centers and radii of the clusters by teaching with remission values of degrees of damage of known skin tissue patterns can be determined and saved and an assignment of remission values unknown skin tissue to these trained clusters of Degrees of skin damage occurs, and that at least one representation of the Areas of different degrees of skin damage in the area of  recorded skin surface is displayed based on the cluster assignment becomes.

It proves to be advantageous for the spectral decomposition of the remitted Selective reflective edge filters are used in light in wavelength ranges in order to spatially couple out spectral components step by step. Preferably are successive reflective low-pass filters with gradual higher wavelength of the low-pass edge used. Likewise, as successive edge filters also high-pass filters with gradual lower wavelength of the high-pass edge can be used.

It proves to be particularly advantageous for reducing the influence of different skin types, different lighting and other varying factors in the image acquisition if, before the cluster analysis for the individual image points assigned to one another, ratios for the generation of derived remission values according to the relationship from the simultaneously recorded spectral images

are formed, wherein r _{i means} digitalized reflectance values of the n different spectrally narrowband reflectance values recorded for a single pixel of the imaged skin surface.

It turns out to be sufficient for many applications, as the number n to select spectrally narrowband reflectance values four and with parts of the blue, green, red and near infrared (NIR-) Spectral range to use.

For the classification of degrees of skin burns, preferably as central wavelengths of narrow-band filtering 450 nm, 550 nm, 650 nm and 800 nm are used, the half widths of the Bandpass filtering in the range between 5 and 20 nanometers can be selected.  

For a clearer representation of the different degrees of skin damage a false color representation of the surfaces turns out different Skin damage as appropriate.

In addition to showing the different degrees of skin damage also a true color image from the narrowband Spectral channels red, green and blue are composed, whereby can be switched between the two representations.

The above object is achieved in a device for generating Data for the diagnosis of the degree of damage to skin tissue Patients with one image acquisition unit for simultaneous acquisition of several spectrally different images of the surface of the skin tissue using a Beam splitter, suitable spectral filter and image sensors, with one Evaluation unit, the means for generating from the spectral image data derived from different images and one Output unit for displaying differently damaged areas of the Skin tissue, solved in that a lens in the image recording unit for large-scale detection of remitted white light from the skin tissue is provided to supply the detected light to the beam splitter that the Beam splitter is a wavelength division multiplexer made of prisms lined up, which divides the light remitted from the surface of the skin into allows at least four different pictures, the prisms on their from A spectrally selective edge filter layer in each case on the surface facing away from the objective for spatial decoupling of a light beam with limited Have wavelength range and each of these selectively coupled Beams of light before hitting the assigned image sensor is filtered narrowband, and that means for in the evaluation unit Carrying out a data analysis in one of the spectrally different ones Reflectance values of the image sensors generated parameter space are present,  the data analysis is a comparison of current data with already in data stored in the same parameter space from taught-in known ones Includes degrees of skin damage and the current data the known Assigned degrees of skin damage.

Each edge filter layer is advantageously a reflection filter which acts as a low-pass filter with a steep edge, the edge of the low pass from prism to prism is gradually shifted to longer wavelengths. On the other hand, it is too possible that each edge filter layer acts as a high pass Is reflection filter and the edge of each high pass from prism to prism is gradually at a smaller wavelength. With this kind Step edge filter arrangement is a great selectivity in beam splitting, high optical transmission and good image transmission quality for preserve all spectral channels.

It turns out to be used for the adjustment of the beam splitting and image acquisition expedient that all edge filters perpendicular to a common Plane are arranged which are the uniform plane of the spectral spatial Beam splitting is.

The evaluation unit advantageously contains means for forming the ratio of the individual pixels of the spectrally different images recorded at the same time for the generation of derived reflectance values according to the relationship

where r _i mean the digitized reflectance values of the n different reflectance values measured spectrally narrow-band for each individual pixel of the imaged skin surface.

Furthermore, in the evaluation unit for the classification of Skin damage, especially skin burns, advantageous means for  Carrying out a cluster analysis in the spectral parameter space different reflectance values exist, with the centers and radii the cluster by teaching with reflectance values of degrees of damage known skin tissue patterns are determined and stored and Remission values of unknown skin tissue to these learned clusters attributable to degrees of skin damage and thus with regard to their Damage can be classified and are expedient in the output unit at least means to represent the areas of different Degrees of skin damage in the range of the currently recorded Skin surface based on cluster assignment and classification intended.

Exactly be advantageous for the classification of skin burn degrees four different spectral reflectance values recorded in narrowband from parts of the blue, green, red and near infrared (NIR-) Spectral range used. It turned out that as the central Wavelengths of the narrow band filtering 450 nm, 550 nm, 650 nm and 800 nm can be used advantageously, the half widths of the Bandpasses are between 5 and 20 nanometers.

The wavelength division multiplexer of the device according to the invention consists in in this case preferably from three wedge-shaped prisms to the side Coupling of the first to third image sensors and one of these prisms connecting four-sided connecting prism for straight coupling of the fourth image sensor, with respect to the light entering from the lens the wedge-shaped prisms on the back are reflective Edge filter layer is arranged and at least on the front there is an air layer for internal reflection of the edge filter of the prism coupled light in the direction of the short prism side, on the the image sensor is mounted orthogonal to the direction of internal reflection, and  all prisms are dimensioned so that the optical path lengths are the same within each of the prisms up to the image sensor.

To map the skin surface over the wavelength division multiplexer the lens exhibits spectrally different illuminated image sensors Expediently a short focal length on the object side and a large focal length on the image side. The image recording unit is preferably with a lens, wavelengths multiplexer, edge filter layers, bandpass filters, image sensors and one Image storage arranged in a compact video camera, and only for Cluster analysis and classification of skin damage is a data transfer too provided a personal computer.

To reduce the influence of errors in the lighting is a strong white Light source firmly connected to the video camera, the lighting being such is to align that light reflected directly on the skin surface is not in the camera lens can get.

The invention is based on the basic idea, in particular for the diagnosis of Skin burns a methodically simple multispectral detection process through a mobile, easy-to-use and menu-driven digital Video camera in CCD technology with active lighting of the skin tissue with white light and the use of adaptive cluster Evaluation method based on a medically verified learnable Patient database for automated evaluation of the data with output a true color image of high brilliance and a false color image that the Assigns a defined color to areas of the same combustion depth.

Extensive preliminary investigations have shown that a more reliable determination of the combustion depths 2 a (superficial), 2b (deep dermal) and 3 (full thickness) is possible with more than three wavelengths. The invention is further based on the knowledge that the spectroscopic data of healthy and (homogeneously) burned skin form cluster structures in the above-mentioned color space, so that with a clinically pre-verified database which describes the variety of skin types and their degrees of burns by reflectance spectroscopy and also patient and (clinical) treatment data and, using a fuzzy cluster algorithm (e.g. the modified "fuzzy c-means" algorithm from Bezdek, 1981), clearly assign the measured values to the clinically relevant burn depths (without additional reference measurements ) is possible. That is, a specific search is made for clusters of the current measurement data in the color space formed from the reflectance values, and these measurement clusters are assigned to the already learned clusters of known degrees of damage by determining the smallest deviation. It is thus possible to assign each skin pixel with at least three reflectance intensities (which are derived from ratio formation) to a measurement cluster. After filtering and pre-classification of the measurement data, which are to be assigned to the environment and the unburned skin (or areas that are too dark or overexposed), the algorithm then determines the fuzziness, the characteristic topographical shape of the clusters ("pattern recognition") ) and adapts the coordinates of the "measured" color space to that of the database. In addition to the spectroscopic data, the algorithm also takes patient data into account, e.g. B. the type of combustion, combustion process, time after combustion and type of pretreatment.

With the arrangement according to the invention, it is possible to find an assessment of the degree of damage to human skin on the basis of spectrally decomposed video recordings, which permits an objective and reproducible assessment of the skin damage regardless of the skin type and time of the skin damage. In particular, the device according to the invention creates an objective diagnostic aid that reliably differentiates between the degrees of burn 2 a (2, superficial) to 2b ( 2 , deep) and 3 and healthy skin.

The invention will be described in more detail below using an exemplary embodiment are explained. The drawings show:

Fig. 1: a diagram of the essential functional elements of the method according to the invention

Fig. 2: the image pickup unit according to the invention with wavelength multiplexer for wavelength-4

Fig. 3: the color splitting concept of the 4-channel image pickup according to the invention

FIG. 4 shows the inventive principle of the image data analysis, applied to skin burn Grade

Fig. 5 is a representation of the reflectance values of the skin in response to the combustion rate of the skin for the selected spectral bands according to the invention

FIG. 6 shows the inventive function of the software for the multi-spectral classification adaptive image data

The method according to the invention essentially consists of the steps

• The skin tissue is illuminated with white light in such a way that that essentially only remitted light from the surface of the Skin tissue enters the spectral channels for image acquisition,
• - At least division of the light remitted by the surface of the skin four separate light paths, with successive spectral selective edge filter spectrally defined step by step Wavelength ranges for the spectral channels are coupled out,
• - narrow band spectral filtering in each after the gradual Spectral decomposition created light path before optoelectronic Image conversion, this bandpass filtering being characteristic Changes in the skin's reflectivity when the skin is damaged Takes into account
• - Cluster analysis in a parameter space of the reflectance values is carried out with the centers and radii of the clusters through Teaching with remission values of degrees of damage known Skin tissue patterns are determined and saved and one Allocation of remission values of unknown skin tissue to these learned clusters of degrees of skin damage occurs, and
• - Representations of the areas of different degrees of skin damage in Area of the absorbed skin surface based on the Cluster mapping.

This basic concept is shown schematically in FIG. 1. A light source 1 , preferably a halogen lamp, illuminates the skin tissue to be examined, e.g. B. Burnt skin, intense with white light. Light remitted from the surface of the skin (directly reflected light is suppressed by the illumination geometry for the image acquisition) is recorded in at least four spectral channels with a multispectral camera, which is described in more detail below, as outlined in part 1 (image acquisition) of FIG. 1. The recorded spectral images 4 are shown in FIG. 1, Part 2 (image transfer) to an evaluation unit 5, preferably a personal computer (PC), labtop or other computer unit is transmitted without the recorded spectral images preprocessed, that is compressed or were otherwise processed . According to Part 3 of Fig. 1 (Image Analysis and Classification), the spectral images 4 - as will be described in more detail below - are analyzed pixel by pixel using suitable software, whereby a classification of the skin damage of the currently recorded skin damage is performed by means of certain relationships and comparison with reflectance values of degrees of damage of known skin tissue patterns Skin surface is made. From this, the damage areas can be displayed in a display unit 6 in various ways. On the one hand, a true-color image (true color image) is suitable, in particular for visualizing the burn for assessments by not directly inspecting the person injured in the fire. On the other hand, the false color display is particularly preferred as the display mode in order to be able to clearly separate the different degrees of combustion in certain areas. Both forms of representation can be meaningfully output as paper printouts or monitor displays and selected as required.

In the following, without restricting generality, due to the the invention is easier to present and the technical clarity described using a four-channel structure, which is particularly suitable for Classification of skin burns has proven to be beneficial.

In the case of exactly 4 wavelengths, the selection of the wavelengths using a simple layered model of the skin (Anselmo et al., 1977, Afromowitz et al. 1987, 1988).

The light remitted by the human body (and the burn), as shown in FIG. 2, is collected by a special lens 31 with a long image-side focal length and motorized object-side focal length adjustment and via a block filter 32 , which blocks against boring infrared light (IR) a wavelength division multiplexer 33 in the form of a compact prism block. The wavelength division multiplexer 33 is constructed in the most general case from a multiplicity of rigidly cemented prisms in such a way that the different wavelength regimes are split into separate beam paths within a plane without significant losses, and only immediately before the corresponding CCD matrix sensors are encountered selected central wavelength and associated half-width with a suitable bandpass filter (which can also be electrically controlled for other applications). This ensures that at the same time the selectivity is high, the optical transmission is high and good image transmission quality is maintained. With a planar arrangement (from <4 wavelengths an extra-planar arrangement is required) and by using modern adhesive technology, an extremely precise and permanent adjustment of the CCD modules relative to each other is achieved, thus ensuring pixel synchronization between the individual pixels of the CCD sensors.

In the specific case, as shown in FIG. 2, the wavelength multiplexer 33 consists of three three-sided prisms 331 to 333 and a (four-sided) connection prima 334 . On the surface of the three-sided prisms 331 to 333 facing away from the objective 31 (and thus the light entry) there are color-dividing layers 341 to 343 , which are reflection filter layers and have high-pass filter characteristics in reflection, so that high-frequency light is almost completely reflected up to the predetermined filter edge and low-frequency light below the edge is let through almost without weakening. As a result of the different inclinations of the surfaces of the prisms 331 to 333 carrying the color splitting layers 341 to 343 , the reflected spectral components of the white light entering through the objective 31 are coupled out of the incident light path (shown in broken lines). In order to increase the spatial distance required for the detection, the surface of the prisms 331 to 333 located in the front in the light path is used as an internal reflection surface. For this purpose, a large change in the refractive index is required on this surface, which is realized by an air gap between prisms 331 and 332 and between prisms 332 and 333 and the exposed front surface of prism 331 . The two air gaps are preferably permanently configured by cementing using a shadow mask that leaves the light path free. The special, sometimes idiosyncratic shape and size of prisms 331 to 334 is predetermined within narrow limits by the fact that it proves to be very useful for multispectral recording technology to let the light pass through the same optical path length up to the sensors in all spectral channels. A compromise can be found between the smallest possible angle of the spectrally selective reflection at the color splitter layers and the greatest possible gain in space for the coupling of the sensors in the form of the CCD sensor blocks 361 to 364 .

The spectrally selected light reflected on the inside of the front surfaces of the prisms 331 to 333 is coupled out on the third, short side of the prisms onto the CCD sensor blocks 361 to 364 , in such a way that the short side of the prism in question is penetrated perpendicularly to and against each other Connect a narrow band spectral filter (SF) 351 to 353 and the CCD sensor blocks 361 to 363 . The spectral filters 351 to 353 and CCD sensor blocks 361 to 363 are cemented directly to the short prism side. The same also applies to the elements connecting prism 334 , spectral filter 354 and CCD sensor block 364 located in the direct light path.

The CCD sensor blocks 361 to 364 have at least video standard with regard to their CCDs in terms of the number of pixels. The size and number of pixels is determined by the spatial resolution of the burn from <1 to approx. 3 mm and the measuring distance of typically 30 cm (close-up) to 200 cm (full-body image).

With regard to the color splitter layers 341 to 343 , the desired light division by color selection by means of step-by-step differentiated high-pass filtering is specifically shown in FIG. 3. The edge filters used are characterized by a steep edge and almost complete reflective coupling of the bandpass wavelength range. Next, the 3 Bandpaßauswahl the narrow-band spectral filter is shown 351 to 354 versus wavelength in FIG.. On the basis of the calculated and experimentally more precisely determined values, the device according to the invention operates with the wavelengths (450 ± Λ ₁ ) nm (blue), (550 ± Λ ₂ ) nm (green), (650 ± Λ ₃ ) nm (red) and ( 800 ± Λ ₄ ) (NIR), where the half-widths Λ _{n are} typically in the range from 5 to 15 nm and have been optimized with regard to a stable evaluation algorithm and the technical feasibility by choosing the blue spectral range that is significantly broadband (20 nm). The bandwidths are a compromise between camera sensitivity (bandwidths as large as possible) and the specificity of the skin's absorption curves (bandwidths as narrow as possible). Although there are no strong spectral changes in the reflectivity of the burnt skin in the blue, there is a decreasing intensity of the conventionally available light sources (halogen spotlights), so that this is compensated for with a wider bandwidth of the bandpass filter compared to the other bandpass filters. Otherwise, the choice of the central wavelength of the bandpass filter depends on the absorption and scattering curves of relevant skin components, especially the hemoglobin:
450 nm: relatively strong portion of the top layer
550 nm: strong dependence on the volume fraction of blood in the second layer (absorption of hemoglobin)
650 nm: strong relative difference between the absorption of HbO ₂ and Hb
800 nm: isobestic point of HbO ₂ and Hb, hardly any scatter, but empirically a strong effect for third-degree combustion areas is shown.

When expanding the skin model used up to now and including further optically effective, combustion-relevant skin components, the use of further wavelengths can make sense. Although these then lead to the expansion of the dimensions of the parameter space (color space) 52 , the subsequent analysis and classification algorithms do not change (mathematically) as a result.

In order to record the remitted values of a burn with 4 (or more) CCD (or other image sensors, such as CMOS Active Pixel) matrix sensors, the damaged skin tissue (and the healthy skin surrounding it) is scanned with the white light source 1 (separately or connected to the image acquisition unit (hereinafter referred to as a "camera" because of its basic functional similarity with a video camera), irradiated continuously or pulsed (saving battery energy). This provides intensive illumination in the above-mentioned wavelength intervals, the surrounding radiation sources Due to the strong scattering of the light in the (burned) tissue, the reflectance intensity is independent of the angle of illumination. Avoiding direct reflections (illumination of the wound at an angle to the surface normal) ensures that the The reflectance intensities are much greater than the reflection intensities ten are. The camera is in turn black by an internal and an external white balance (Standard) automatically "color calibrated" to the used illumination. With (adjustable) large-area lighting and a somewhat smaller but comparable field of view of the camera lens, large-scale and even whole-body shots are possible in addition to local ones.

A mobile version of the camera comes with a rechargeable battery operated, which can be replaced at any time and at full operation and full lighting (100 W) an operating time of at least 20 min. allowed, which experience has shown is more than sufficient for the recordings. The camera electronics is designed so that a video mode is realized, the feeds the color image into the camera's viewfinder, just like a professional Video camera. At the same time, a menu is displayed in the viewfinder (as required) Control and monitoring of the camera shown. Because the camera itself the lighting is monitored itself in the event of a recording break switched off with a time delay (to protect the battery) or the charge status the battery in the viewfinder and indicated by LED on the camera.

The doctor recognizes in the viewfinder that, in his opinion, optimal Recording scene, this scene is frozen at the push of a button in the viewfinder (Switch position "Freeze"), the viewer decides at short notice whether the image is usable, d. H. saved (switch position "Safe") or whether the Discarded, d. H. can be switched back to video mode should (release switch). Since the mobile camera with a PC chip and flash Memory is equipped, you can choose approximately 10 (à 4. 1.5 Mbit b / w pictures) and more pictures are saved in the camera. Already done In principle, recordings are not lost.

After a recording session has ended or when the memory is full (which is automatically displayed in the viewfinder), the data is transferred from the mobile camera to the hard drive of the PC (labtop), which can be off-site (second step in Fig. 1 ). This transfer takes place in serial data exchange and is only released if the patient identification number has been entered beforehand.

The recording of a burn consists of (at least) 4 b / w pictures, each picture being assigned to a narrow spectral channel λ ± Λ. Each "skin pixel" (a few mm ² ) is assigned four reflectance intensities l _xy 1 to l _xy 4 via the pixels of the CCD sensors through the optical imaging and wavelength division, which, when calibrated to the total reflectance intensity, result in three independent variables and thus a 3-dimensional one Span color space 52 .

The problem now is the assignment of these measured values to the "normal" Skin and the degree of burns as there is no normal skin that Skin (layers) models compared to the actual biological tissue are greatly simplified and spectroscopic in vivo measurement data for the different combustion depths at different (individual!) Skin types, the different body regions and different Pretreatment methods are missing. Consequently, the coordinate axes of the 3-dimensional color space cannot be regarded as absolute.

The invention is based on the knowledge that the spectrocopical data of healthy and (homogeneously) burnt skin in the above-mentioned color space form 52 cluster structures, so that with a clinically pre-verified database which describes the variety of skin types and their degrees of burn by reflectance spectroscopy and also patient and contains (clinical) treatment data, and using a fuzzy cluster algorithm (e.g. the modified "fuzzy c-means" algorithm from Bezdek, 1981) clearly assigns the measured values to the clinically relevant burn depths (without additional reference measurements) is possible. It is thus possible to assign a measurement cluster to each skin pixel with its (at least) 3 derived reflectance intensities r _xy 1 to r _xy 3. The derived reflectance intensities are obtained as shown in Fig. 5 and according to the relationship

where r _i signifies digitized reflectance values of the n different reflectance values recorded for a single image point on the imaged skin surface.

As is schematically supported in FIG. 4, the spectral reflectance intensities recorded pixel by pixel by the burned skin by the camera 3 are recognized as clusters in the color space 52 in certain intensity ratios. After filtering and preclassifying the measurement data, which can be assigned to the environment and the unburned skin (or areas that are too dark or overexposed), a software algorithm then determines the fuzziness, the characteristic topographical shape of the clusters (pattern recognition ") and adapts the coordinates of the" measured "color space to that of the database. This corresponds to a two-stage process, which in the first step determines the currently existing various states in parameter space 52 (cluster), assigns the image pixels to these clusters and in the second step classifies these state clusters 53 by means of a pattern comparison, which then also retrospectively classifies the pixels. Fig. 5 shows for the cluster formation of the qualitative dependencies in the model in the upper part and the lower part of the intensity relationships empirically found by forming appropriate ratio to the combustion degrees of the human skin.

The basic sequence of the evaluation is shown in FIG. 6 in the whole. It begins with the acquisition of measurement data in four spectral channels, continues through the transformation of the spectral reflectance values into an (n - 1) parameter space (through ratio formation, here at four wavelengths: 3 ratios) as well as cluster formation and cluster assignment up to the classification of the clusters in Degrees of skin burn by comparison with samples already classified in a sample database.

In addition to the spectroscopic data, the algorithm also takes into account Patient data, such as B. the type of combustion, combustion process, time after combustion and type of pretreatment.

In the learning process of the method, a representative distribution of the Degrees of combustion in the color space as the basis for the classification of the Determines the cluster and trains the assignment algorithm (axis scaling). The learning phase also serves to statistically secure the classification by comparing the evaluation with the results of histo-morphometric Investigations.

The training happens with a large number of burns histological examinations of samples from the affected Tissue areas and histological assignments of the biopsies different degrees of combustion. This will be done in the first 3 days after burning the debrided and topically treated areas in solid time intervals measured and evaluated. In addition, on the burned tissue biopsies taken on the day of surgery and histologically determined the burn result.

The evaluation method is also characterized by its adaptivity out. It does not have, as usual, defined segment boundaries and therefore one rigid classification of the individual pixels, but is continuous customized. This results from a two-step process, the first step the current various states in the parameter space are determined (Cluster), which assigns image pixels to these clusters and, in the second step, these State clusters classified using a pattern comparison, which then the pixels are also classified retrospectively. This will make the the strong variability of the skin tint and fixed segment boundaries uncertain classification avoided.  

The invention has advantages above all in clinical diagnostics. So delivers objective diagnostic tool that is safe between the Degrees of burns 2a (2, superficial), 2b (2, deep) and 3 and more healthy Skin and scenes in the image that do not affect the patient are different. With The visual representation of true and false colors becomes an objective aid for operations planning and the associated logistics. The The depths of a superficially severe skin burn can already can be determined immediately after debridement. With the objective Detection method is a follow-up check of the burn in the first 3-4 days (after that the wound begins to change so that new ones Algorithms should be used) possible. In particular, it stands Medical professionals now have a diagnostic system available that enables the limit of conservative wound care the degrees of burn of 2a in the 2b area, with the related above positive effects already mentioned.

The doctor is thus offered a procedure that is highly diagnostic and Reproduction security the use of complementary databases as well taking into account the influences of different treatment methods allowed. A reliable objective diagnosis result is achieved under the automatic consideration of the different human skin types and the differences between the different parts of the body due to the Detection of scenes that also include unburned skin areas around the skin Show combustion in the picture. This sets up a separate calibration healthy or fictitious reference skin is unnecessary.

The doctor has a mobile image acquisition system at his disposal allows maximum mobility with simple operation and thus the Protects patients. It basically offers the option of being local Burns, also to take whole body pictures. The necessary full-body images and the possible objective  Determination of the ratio of skin areas burned to different depths to the entire body surface are a crucial aid for that Initiation of the necessary intensive care measures Delivery of the severely burned person to the clinic and the planning of the (gradual) required surgical and logistical measures. This creates the possibility of only the absolute early on necessary to carry out skin transplantation measures and thus the To create the basis for better chances of healing, shorter ones Treatment times for skin damage, especially burn injuries.

Further refinements of the method according to the invention and the associated device are, the image evaluation not only in separate, (electrically) not connected PC, but due to the PC chip included in the camera among uses additional memory modules to perform the analysis in real time and alongside the examining doctor simultaneously the result in the real color image False color display as described above in the viewfinder or another Display. This means that the online transmission of this data is also possible Real-time for training purposes and as part of on-line tele-consulting possible. For this purpose, the data can also be viewed on the camera installable display, the people present in the area shown or via a cable connection or a free space data connection transferred to any receiving module inside and outside the OR and to be followed at the same time and worldwide.

By expanding the image processing software with modules from Area calculation is also the exact acquisition of the numerical Area of the burned areas (with comparison scale in the picture or by extrapolation from the body dimensions) possible.  

Furthermore, it is possible with the method according to the invention, for. More colorful Use of endoscopic aids, including "burns" to examine within the human body, the z. B. in the lungs have arisen when breathing hot air.

The measurement of vitality and "complexion" in cosmetic surgery Transplants are also one of the possible applications of the Invention. The same applies to the measurement and monitoring of carcinogens Skin changes (in addition to the individual carcinoma, it is also possible that in the case of dense distribution through a large-scale parallel analysis, only those Carcinomas appear to the examining doctor which are positive are / or could be.

Another application is the measurement of the condition and the subsequent follow-up of open wounds.

Monitoring the effect of skin engineering measures (e.g. Laser treatment) with recording of the treated area dimensions realizable by the inventive method (in particular the future possibility of debridement from burns and the Removal of "dead" tissue with the help of those described here Invention using z. B. to automate laser ablation procedures, both in area as well as in depth).  

List of the reference symbols used

1

Light source

2nd

Skin tissue / skin surface

3rd

camera

31

lens

32

IR block filter

33

Wavelength multiplexer / prism block

331-333

(three-sided) prisms

334

(four-sided) connecting prism

34

Color splitter layers

341-343

Color splitter layers / reflection filter layers

35

spectral bandpass filters

351-354

narrow band spectral filters (blue, green, red, NIR)

36

CCD sensors

361-364

CCD sensor blocks

4th

Spectral images

5

Evaluation unit / computer

51

spectral reflectance intensities

52

Cluster in the parameter / color space

53

State cluster

54

Database

6

Display unit

Claims (21)

1.Procedure for generating data for the diagnosis of the degree of damage to a patient's skin tissue, in which a simultaneous image recording of the surface of the skin tissue takes place in several spectrally different channels, derived image data are generated from the spectrally different images and differently damaged areas of the skin tissue are displayed , characterized in that

• the skin tissue is illuminated over a large area with white light such that essentially only remitted light from the surface of the skin tissue reaches the spectral channels for image acquisition,
• the remitted light is divided into at least four separate light paths, with successively spectrally selective edge filters which are applied to optical surfaces of prisms, spectrally defined wavelength ranges for the spectral channels are coupled out,
• after the gradual spectral decomposition, a narrow-band spectral filtering is carried out in each spectral channel before the optoelectronic image conversion, this band-pass filtering taking into account characteristic changes in the reflectance of different skin layers when they are damaged,
• a cluster analysis is carried out in a parameter space of the reflectance values, the centers and radii of the clusters being determined and stored by teaching with reflectance values of degrees of damage of known skin tissue patterns and an assignment of reflectance values of unknown skin tissues to these taught clusters of degrees of skin damage, and
• - At least one representation of the areas of different degrees of skin damage in the area of the absorbed skin surface is displayed on the basis of the cluster assignment.

2. The method according to claim 1, characterized in that for the spectral decomposition of the remitted light into wavelength ranges in Reflection effective edge filters are used to gradually Coupling spectral components spatially. 3. The method according to claim 2, characterized in that for spectral decomposition of the remitted light as successive Reflection filter Low pass filter with gradually larger wavelengths Low pass edge can be used. 4. The method according to claim 2, characterized in that for spectral decomposition of the remitted light as successive Edge filter high-pass filter with gradually lower wavelength High-pass edge can be used. 5. The method according to claim 1, characterized in that before the cluster analysis for the individual pixels assigned to one another from the simultaneously recorded spectral images, conditions for generating derived remission values according to the relationship
are formed, wherein r _{i means} digitalized reflectance values of the n different spectrally narrowband reflectance values recorded for a single pixel of the imaged skin surface. 6. The method according to claim 1, characterized in that as the number n of the spectrally narrow-band reflectance values four is chosen and parts of blue, green, red and middle infrared (NIR) spectral range can be used. 7. The method according to claim 6, characterized in that for the classification of skin burns as central wavelengths the narrow-band filtering 450 nm, 550 nm, 650 nm and 800 nm be used, the bandwidths in the range between 5 and 20 Nanometers can be selected. 8. The method according to claim 1, characterized in that to clearly show the different degrees of skin damage a false color representation is selected. 9. The method according to claim 8, characterized in that in addition to showing the different degrees of skin damage True color image from the narrowband spectral channels Red, green and blue are put together, being between the two Representations can be switched. 10.Device for generating data for the diagnosis of the degree of damage to a patient's skin tissue with an image recording unit for simultaneous recording of several spectrally different images of the surface of the skin tissue by means of a beam splitter, suitable spectral filters and image sensors, with an evaluation unit, the means for generating from the contains spectrally different images derived image data and an output unit for displaying differently damaged areas of the skin tissue, characterized in that

• a lens for large-area detection of remitted white light from the skin tissue is provided in the image recording unit in order to feed the detected light to the beam splitter,
• - The beam splitter is a wavelength division multiplexer of prisms lined up, which allows the light reflected from the skin surface to be divided into at least four different images, the prisms each having a spectrally selective edge filter layer on the surface facing away from the lens for spatially coupling out a light beam with a limited wavelength range, and each of these selectively coupled out light beams is narrow-band filtered before hitting the assigned image sensor, and
• - Means for performing a data analysis are present in the evaluation unit in a parameter space generated from the spectrally different reflectance values of the image sensors, the data analysis including a comparison of current data with data of degrees of skin damage of known skin tissue patterns already stored in the same parameter space, and the current data relating to the known ones Assigned degrees of skin damage.

11. The device according to claim 10, characterized in that each edge filter layer acts as a low-pass reflection filter with a steeper Edge is where the edge of the low pass from prism to prism is gradual is shifted to longer wavelengths. 12. The apparatus according to claim 10, characterized in that  each edge filter layer with a high-pass reflection filter is steep edge and the edge of each high pass from prism to prism is gradually at a smaller wavelength. 13. The apparatus of claim 11 or 12, characterized in that all edge filters arranged perpendicular to a common plane which is the uniform level of spectral spatial beam splitting. 14. Arrangement according to claim 10, characterized in that the evaluation unit has means for forming the ratio of the individual pixels of the simultaneously recorded spectrally different images for the generation of derived remission values according to the relationship
where r _i mean the digitized reflectance values of the n different reflectance values measured spectrally narrow-band for each individual pixel of the imaged skin surface. 15. The arrangement according to claim 10, characterized in that

• - Means for performing a cluster analysis in the parameter space of the spectrally different reflectance values are provided in the evaluation unit, the centers and radii of the clusters being determined and stored by teaching remission values of degrees of damage of known skin tissue patterns and remission values of unknown skin tissues being assignable to these taught clusters of skin damage degrees and thus are classifiable with regard to their damage, and
• - The output unit has at least means for displaying the areas of different degrees of skin damage in the area of the currently recorded skin surface on the basis of the cluster assignment and classification.

16. The arrangement according to claim 15, characterized in that Four different degrees of skin burn classification spectral reflectance values recorded from parts of the blue, green, red and near infrared (NIR) spectral range are provided. 17. The apparatus according to claim 16, characterized in that for the classification of skin burns as central wavelengths the narrow-band filtering 450 nm, 550 nm, 650 nm and 800 nm are provided, the half-widths of the bandpass in the range are between 5 and 20 nanometers. 18. The arrangement according to claim 16, characterized in that the wavelength division multiplexer consists of three wedge-shaped prisms for lateral coupling of the first to third image sensors and a four-sided connecting prism connecting these prisms for straight-line coupling of the fourth image sensor, with respect to the light entering from the lens the wedge-shaped prisms

• the reflective edge filter layer is arranged on the back,
• - On the front there is at least one air layer for internal reflection of the light coupled out from the respective edge filter of the prism in the direction of the short prism side, on which the image sensor is located orthogonally to the direction of the internal reflection, and
• - The prisms are dimensioned so that the optical path lengths within each of the prisms are the same.

19. The arrangement according to claim 10 or 18, characterized in that for imaging the skin surface using the wavelength division multiplexer the spectrally different illuminated image sensors the lens has a short focal length on the object side and a large focal length on the image side. 20. The arrangement according to claim 10, characterized in that the image acquisition unit with lens, wavelength multiplexer, Edge filter layers, bandpass filters, image sensors and an image memory is arranged in a compact video camera, and only for Cluster analysis and classification of skin damage a data transfer too a personal computer is provided. 21. The arrangement according to claim 20, characterized in that a strong white light source is firmly connected to the video camera, the lighting is aligned so that directly on the Skin surface reflected light does not get into the camera lens.