DE102008040838A1 - Dermal tissue state detecting method for e.g. human body, involves morphologically classifying spectrally classified results, determining evaluation results from morphologically classified results, and outputting evaluation results - Google Patents

Abstract

The method involves compensating inhomogenity of illuminations and/or variations of color responses or dark current of an individual pixel of a camera. A reflection spectrum is calculated for each image point from energy raw data of a tissue area, such that raw image data is spectrally preprocessed. The preprocessed raw image data is spectrally classified, and the spectrally classified results are morphologically classified. Evaluation results are determined from the morphologically classified results, and the evaluation results are output to a user interface. An independent claim is also included for an arrangement for detecting a state of a dermal tissue.

Description

The The invention relates to a method and an arrangement for detection and subsequent evaluation of skin tissue, in particular for detection of skin changes.

With increasing life expectancy and at the same time increasing burden by chemical and / or physical environmental influences the number of diseases of human or even animal body tissue increases. A typical example of this is malignant, d. H. malicious, Changes in the skin, which include melanoma.

For the effective treatment of such tissue changes it is of high importance, the changes already to diagnose at an early stage. This is done after the current, established state of the art by a purely visual Consideration of suspected sites by a qualified specialist. If a tissue area proves to be suspicious Tissue removed, prepared as a microscopy sample and, again by a qualified, experienced specialist, histologically analyzed. This process is both time consuming and expensive as well as stressful for the patient, being invasive Tissue samples are taken. The assessing specialist is present forced a compromise between the one largely secure Diagnose enabling, but expensive and the examined Individual burdening removal of a tissue preparation and the danger of a vicious change of To overlook tissue to enter. On top of that, the visual In vivo diagnosis strongly dependent on subjective impressions is, such as the color impression, not only of person to person, but also can vary from day to day.

Consequently would be a technical device or process, which autarkic and rapid acquisition of data from tissue in vivo, d. H. at the living human or animal body and without taking a sample, and a preparation of the recorded Data outside the body and independent from the presence of the body and, where appropriate an evaluation of the processed data according to objective criteria of great practical interest. With help such methods and devices it would be possible for example, also in private practitioners or veterinarians, which may not the necessary specialist knowledge and experience for visual tissue diagnostics own, screening for. B. in the context of routine To carry out check-ups. Patients with (potential) Tissue changes could then be targeted to one Specialists are forwarded for diagnosis. Thus would be an efficient "screening" with little effort possible, which also (still) inconspicuous tissue changes in early stages of a disease with significantly higher Probability as detected by the current system and accordingly can be treated.

conditions to such a method or device would be, in addition to compliance with general medical regulations, for example concerning the radiation exposure of the patient etc. or compliance with health regulations, a quick, straightforward Data acquisition of the tissue on the living, preferably awake individual. For this it is desirable to have the measurement data within preferably to absorb less than 60 s, to distortions to prevent movement of the patient. A subsequent one Preparation of the recorded measurement data and their evaluation, for which does not require the patient's presence while being able to (potential) tissue changes with the highest possible degree of safety as such (avoidance false-negative evaluations). At the same time, also for reasons of acceptance, the number of false-positive evaluations should not be too high. moreover It would be beneficial if the meter was not necessarily be operated by a doctor, but similar X-ray devices, for example, a corresponding trained medical technician can perform the measurement. The doctor can thereby focus on the actual diagnosis, whereby the automatically generated, objectifiable evaluation results the measurement, processing and Evaluation constitute a substantial support.

From the literature, a number of corresponding approaches are known. The classical method for the automated evaluation of tissue changes is imaging by means of a camera, followed by a computer-aided classification according to general evaluation rules, which can be used for malignant skin changes, such as melanomas, for example by means of electronic image processing according to the so-called ABCD rule and / or other morphological parameters, such as out Ganster et al., IEEE Transactions on Medical Imaging, 20 (3), 233-239 (2001) known. In this case, changes in the skin are recorded by means of a color camera and classified according to up to 33 different morphological parameters by means of computer-assisted image processing.

additional for the computer-aided classification of RGB color images is the application of such morphological classification and Evaluation models also on images at other wavelengths known. The literature mentions approaches in which In vivo imaging of tissue changes in a wide variety of Wavelengths, from the ultraviolet (UV) to the near infrared (NIR). Additionally or alternatively also evaluated fluorescence images. In all cases will be Images at certain, more or less characteristic wavelengths recorded, computer-aided evaluated and from it evaluations over Type and / or extent of tissue change derived.

Problematic is that when lighting with only one or a few Wavelengths the classification accuracy, both false-positive as well as false-negative evaluations tend to be insufficient is.

As an improvement, in a number of publications ( WO 99/44010 . WO 98/37811 . US 6,208,749 . US 6,081,612 ) proposed to perform such a morphological evaluation for a number of different wavelengths initially separately for the diagnosis of skin tumors. The tissue site to be examined is irradiated sequentially with light of different wavelengths, the remitted light is detected spatially resolved by a camera and, after correction of interferences such as hair, for each of typically up to 15 different wavelengths according to morphological parameters such as texture, asymmetry, mottling and Boundary classified. These individual classification results are combined in sequence, weighted if necessary, and derived from this an overall rating.

A fundamentally different approach is based on spectral classification, ie the evaluation of the spectral properties of tissues. This method is based on the fact that the spectral properties of a tissue usually change significantly with changes in tissue histology. This can be associated with a color change, but also with altered absorption, remission and / or fluorescence characteristics in the ultraviolet and / or (near) infrared. Again, there are a number of previously known publications, eg. B. McIntosh et al., Vibrational Spectroscopy 28, 53-58 (2002) or Dwyer et al., Cancer Epidemiol. 7, 203-206 (1998) ,

A Such evaluation is based on the measurement of the spectral Characteristics and their comparison, mostly using statistical Evaluation method, with known spectra benign or malignant Tissue. The problem with this is that tissue usually does not is homogeneous and healthy tissue with benign and / or malignant changed tissue areas alternates. This leads to, especially in the early stages of a change, in classical point measurements often problems in the evaluation, there is a small number of (malicious) within the detected area altered tissue areas of a variety of benign tissue areas opposite. For a successful detection it would be a vicious change required, the investigation with very high local Resolution perform what with point probes even for temporal reasons on living objects, d. H. Patients, is not practical.

As an alternative to the attempt to assign spectral characteristics to specific changes in the tissue, a number of publications have suggested that specific substances in a tissue should be detected and their distribution determined. Such an approach is for example from the EP 1 185 853 known. Spatially resolved spectral information is recorded and spatial distributions of specific substances within the tissue are determined. Of particular relevance are distributions of melanin, especially in skin tissue, and / or blood, ie the degree of blood flow to the tissue. This distributional information of individual constituents allows in consequence also limited conclusions about possibly pathologically changed tissue areas.

Alles Common to these methods is that the requirements in particular for the correctness of the tissue assessment mostly only be met restricted. With a narrow choice of Evaluation parameters become indistinct tissue changes, as they are typical for early stages, only very unreliably recognized. When "softer" parameter is selected although the proportion of correctly recognized tissue change increases, the proportion of erroneously classified as malignant At the same time, areas quickly reach unacceptable levels. In addition, many of the known models are very limited applicable to people of a different skin color, for example. This applies in particular to valuation models that specifically target the Evaluation of skin colorants, such as melanin. All in all Thus, there is no really and completely satisfactory solution for the comprehensive, flexible and reliable automated in vivo evaluation of biological tissues.

It is therefore an object of the invention to provide a method or an arrangement that / the meets the requirements initially set and overcomes or at least alleviates the disadvantages of the prior art.

The Task is solved by a method with the characterizing Features of claim 1 and by an arrangement with the features of claim 16.

According to the invention provided that the method for status detection of skin tissue the the following method steps comprises: in-vivo acquisition of raw image data a tissue part to be examined; outside the body and regardless of the presence of the body: spectral preprocessing of the raw image data, spectral classification preprocessed raw image data, morphological classification the spectral classification result, determining an evaluation result from the classification, and outputting the evaluation result to a user interface.

Thereby In particular, it is possible to use both the spectral and the spectral also the morphological information content of the recorded spatial and evaluated spectrally resolved measurement data coordinated and thus significantly improved correctness of the evaluation to reach. This is the process of the invention not to specific tissue types and / or spectrally well detectable Substances, such as melanin, bound but can be flexible to specific requirements and also different tissue types be adjusted.

Further Advantages and advantageous developments of the invention are in specified in the dependent claims.

Especially is beneficial to the accuracy of the prediction of a judgment can be significantly improved by the tissue area to be examined first with optical radiation in at least one of Areas of the visible, ultraviolet and / or near infrared lit and the remitted, d. H. (diffused) reflected and / or radiation emitted by fluorescence effects resolved spatially and spectrally is detected. This data will be used first for every single pixel according to their spectral information classified, d. H. assigned to a specific (tissue) class. The resulting classification image is, preferably after performing a step to clean up the image of artifacts, a morphological classification applied.

The Data analysis can optionally be spatially evaluated and spectrally resolved data from a tissue region diffused reflected radiation and / or used by emitted fluorescence radiation become. Likewise, this procedure can be used both for evaluation continuous spectral data within one or more optical Wavelength ranges or for the evaluation of discrete spectral data at several, typically 5 to 15 different wavelengths from one or more wavelength ranges, in particular the visible light, the UV and / or the near infrared become.

Preferably In the case of in vivo acquisition of the raw data, the tissue part to be examined becomes from one or more optical radiation sources with the required Wavelength range of adapted radiation characteristics illuminates the radiation remitted by the considered tissue area an optic and a slit diaphragm in an imaging spectrograph discharged, the remitted radiation in the spectrograph spectrally decomposed, and the spectrally decomposed data on a camera chip with sensitivity adapted to the optical wavelength range projected.

By linear displacement of the measuring device relative to the tissue sample can the process is repeated line by line, creating a fast and comprehensive scanning of the tissue parts to be examined possible is.

In an alternative embodiment is to be examined Tissue particle with one or more optical radiation sources adapted to the required wavelength range Radiation characteristic illuminated and that of the considered Tissue area remitted radiation through an optic and a wavelength selection on a camera chip with the optical wavelength range adapted sensitivity projected. The wave selection is preferably carried out by means of a tunable filter or a series of discrete single filters.

By tuning the transmission range of the tunable filter or sequential swinging in and out of the individual filters, a number of spatially resolved images are generated in the jewei At the selected wavelength of the considered tissue part, the radiation received was remitted. The combination of this sequentially recorded spectral information provides the basis for the subsequent spectral classification.

In Another alternative embodiment is the examining tissue section selectively with narrowband radiation sources Emission behavior or a broadband radiation source with appropriate optical filters between the radiation source and to be examined Tissue part, or a combination of both, illuminated. The Radiation remitted from the considered tissue area is transferred over an optic on a camera chip to the optical wavelength range adapted sensitivity projected.

By sequential activation and deactivation of the individual radiation sources or pivoting in and out of the filters between the radiation source and considered tissue part become a number spatially resolved images of the lighting with the respective selected wavelength from the considered tissue part received remitted radiation. The combination of these sequentially recorded spectral information provides the basis for the subsequent spectral classification.

in the Spectral pre-processing of the raw data can Inhomogeneities of lighting and / or variations of Sensitivity or dark current of each pixel of the camera and from raw energy data of the considered tissue site and those a constant spectral standard for each pixel Reflection spectra are calculated.

Preferably is the spectral classification by a linear discriminant Analysis (LDA) or a Quadratic Discriminant Analysis (QDA) or by a multivariate regression method (partial least squares Regression, PLSR).

In front The spectral classification can be a reduction in dimensionality of the data, with the spectral characteristics of the spectral data preferably by a Principal Component Analysis (PCA) or an Independent Component Analysis (ICA). Thereby the amount of data can be reduced and the evaluation speeded up. Subsequently, one can use the spectral features individual or derive several wavelengths for a classification can be used advantageously.

The Number of spectral features will be beneficial to five to 15 spectral features reduced, which in one or more spectral ranges can lie, leaving the automated assessment free from unnecessary information to be done quickly and efficiently can.

A Artifact correction using computer-based image processing algorithms allows the removal of artifacts from recording and / or Classification, in particular pixel errors, speckle noise or scattered light artifacts. This can be done advantageously by using morphological filters and analysis of contiguous (pixel) surfaces the classification pictures happen. The algorithms for this are characterized by a high efficiency.

In In a preferred embodiment, the method steps Spectral Classification, Artifact Correction and, if applicable Sub-steps of morphological classification to a process step summarized, preferably by applying a Contextual Classification.

in the Process step of the morphological classification are the Classification images according to corresponding morphological criteria, in particular texture, homogeneity and / or size classifies certain areas, thereby reducing the final Classification and evaluation can be done.

The thus determined evaluation result contains advantageously the evaluation of the examined tissue part as well as information about the location and geometric dimensions of the tissue changes, so that, for example, at a follow-up a fast and easy retrieval of the tissue part possible is.

advantageously, The method step comprises outputting the evaluation result to a user interface, the visualization of the evaluation results and / or storage in a data store, so that in easier and clear the data for more Appraisal will be available.

Further Details, features and advantages of the invention Arrangements will become apparent from the description below Reference to the accompanying drawings, in which also preferred Embodiments are shown. Show it:

1A a flow chart of the evaluation of a tissue site during measurement of continuous spectral data,

1B a flow chart of the procedure when measuring discrete spectral data at predefined wavelengths,

2 a flowchart of a further embodiment of a measurement sequence according to the invention,

3 an exemplary result of the evaluation achievable with the invention,

4 another exemplary result of the evaluation achievable with the invention,

5 a possible basic arrangement for measuring spatially resolved, spectrally continuous data, and

6 a possible basic arrangement for measuring spatially resolved data at defined discrete wavelengths or wavelength ranges.

Looking first at the in 1A In the form of a flow chart shown first embodiment of the method according to the invention, first the tissue site to be examined in vivo with a suitable measuring arrangement, for example according to one of in 5 respectively. 6 measured and described in more detail below measurement principles spatially and spectrally resolved. The in this raw data 10 Spectral information contained in a sequence in a suitable manner pre-processed (in 11 ). Purpose of this spectral preprocessing 11 it is, for the subsequent spectral classification (at 13 ) to compensate for disadvantageous properties of the measuring method, such as inhomogeneous illumination and / or variations in the sensitivity or the dark current of the individual pixels of the camera. Furthermore, depending on the selected measuring method, any further required calculations can be carried out. For example, it makes sense to evaluate diffuse reflected radiation during spectral preprocessing 11 calculate reflection spectra for each pixel from the raw energy data of the considered tissue site and those of a constant spectral standard and use these instead of the raw energy data for the subsequent classification.

When measuring continuous spectral data, for example by means of imaging spectrograph, before carrying out a spectral classification (at 13 ) a reduction in dimensionality 12 meaningful or depending on the used classifier possibly even necessary. The following are used in the following spectral classification 13 Wavelengths to be evaluated on the statistically necessary and relevant information and thereby enables an efficient spectral classification. Particularly suitable methods for this are "Principal Component Analysis" (PCA) or "Independent Component Analysis" (ICA).

As an alternative to the software-based reduction of continuous spectral information to a smaller number of relevant components, it is also possible with some tissue modifications to implement this reduction in hardware in the measurement method. For this, typically first continuous spectra are recorded in one or more relevant optical spectral ranges and these are subjected to a corresponding analysis offline, for example once again with PCA or ICA. Based on this, assuming correspondingly pronounced spectral differences, it is possible to select a number of discrete wavelengths and to measure them in a targeted manner. Thus, the dimension ("dimension") of the spectral information is reduced from mostly well over 100 wavelengths to typically five to 15 wavelengths within one or more spectral regions, thus allowing dimensionality reduction in such arrangements 12 according to 1A be waived, as in 1B shown.

The pre-processed spectral data, which may have been reduced in terms of its spectral dimensionality, are subsequently processed with the aid of a statistical classification method, for example LDA (Linear Discriminant Analysis) or QDA (Quadratic Discriminant Analysis) or a multivariate regression method, for example PLSR (Partial Least Squares Regression). evaluated (point 13 ). In the mentioned statistical classification methods (LDA, QDA) are about an estimate of the class centers m _i and the covariance matrix Si from the multispectral measurement data x _{i determines} the probability p (x | w _i ) that the spectrum x belongs to the class w _i . In the Bayesian sense, LDA and QDA determine the asymptotically optimal decision limits for the classification. With the following formulas, the a-posteriori probability p (x | w _i ) can be calculated for a given spectrum x of the dimensionality d. E [x] denotes the expected value of x:

The difference between LDA and QDA is that for LDA, the same covariance matrix Σ = Σ _{i is} used for all classes, resulting in linear decision boundaries in the feature space of x. In QDA, each covariance matrices S _{i are} estimated for each class, which results in decision boundaries of second-order polynomials [1].

In multivariate regression methods, a coefficient matrix K is generally searched for which satisfies the following equation for the given values in a vector y: y = Kx (4)

The Elements of y in this case correspond to the known class memberships the training data. In the case of PLSR this equation system becomes solved in the sense of partial least squares [2]. For an unknown spectrum x can over same class of class membership.

The classes can be defined depending on the application and may include, for example, "healthy tissue" / "benignly altered tissue" / "malignantly altered tissue", but also different classes of biological tissue, etc., and, in particular, relevant to skin considerations, for example hair are spatially resolved classification images, as exemplified in 3 shown, with a very good compared to the prior art spatial resolution and high accuracy of the classification, ie, the assignment of the individual pixels to their actual classes corresponding to them.

• 1. Morphologisches Öffnen mit 3×3 Pixel Kernel und kreisförmiger Maske mit Radius 3 Pixel
• 2. Morphologisches Schließen mit 5×5 Pixel Kernel und kreisförmiger Maske mit Radius 5 Pixel

In a next step 14 For example, computer-aided image processing algorithms remove artifacts from capture and / or classification, such as pixel errors, speckle noises, or stray light artifacts that are often unavoidable by the tissue topology, preferably by using morphological filters and analyzing contiguous (pixel) areas of the classification images For the objective task, the following morphological filter operation has proven itself in this order:

• 1. Morphological opening with 3 × 3 pixels kernel and circular mask with radius 3 pixels
• 2. Morphological closure with 5x5 pixels kernel and circular mask with radius 5 pixels

A Definition of these operations can be found in [3]. The corresponding artefacts are preferably from the classification image calculated and the class assignment of the corresponding areas interpolated from the classification results of the neighboring pixels.

By using novel statistical algorithms for spectral classification, which in the spectral classification also include the environment of the relevant pixel in the probability assessment of the class assignment ("contextual classifiers"), it is an alternative to a separate use of the steps 13 and 14 possible, this in one step 16 to unite, as in 2 is shown by way of example with reference to a further preferred embodiment of the method according to the invention.

The classification images improved by application of corresponding image processing algorithms and freed of (largely) artifacts are proposed as the last step of the invention evaluation chain is a computer-aided morphological analysis 15 applied. The classification images are classified according to corresponding morphological criteria, for example texture, homogeneity and / or size of specific regions.

From the thus obtained, both the full spectral and the contained spatial information of the measured raw data considering classification result can now also automatically an evaluation result 17 the considered tissue site are generated. This can, in addition to an evaluation of the considered tissue part according to established criteria, for example, as a benign (birthmark) or malignant (melanoma) change in the skin, for example, information about the geometric dimensions of the tissue changes, etc. included.

This procedure is based on 3 and 4 for two different, in a purely visual assessment not clearly classifiable skin lesions that would be examined according to the medical arts both by biopsy and subsequent histological analysis.

In 3 is an intensely pigmented skin lesion, which would be classified as melanon suspicious due to its shape and staining according to the ABCD rule. The spectral classification ( 13 ) gives distributional images, which, although an indication of a benign tissue change (nevus or

However, malignancy (melanoma) can not be completely ruled out, for example in the upper right margin of the lesion. Only by subsequent application of the morphological analyzes 15 on the classification image, it is possible to achieve a definitive classification as a benign skin lesion. The automated, typically within 60 s received evaluation result 17 for this skin lesion is thus "UNAUFFÄLLIG", which also corresponds to the biopsy result of this skin site.

In 4 is a little pigmented skin lesion shown, which is also crossed by a vertically running in the image narrow scar tissue (scar tissue). The spectral classification 13 results in distributional images that provide a strong indication of a malignant tissue change (melanoma). By subsequent application of morphological analyzes 15 on the classification image, it is possible to achieve a definitive classification as a malignant skin lesion, wherein the intervening or overlying scar tissue can be clearly distinguished. The automated, typically within 60 s received evaluation result 17 this skin lesion is thus "amazing", which also corresponds to the biopsy result of this skin area.

The evaluation results 17 are preferably to a user-customized visualization system 20 and can be used there to make a diagnosis by a qualified physician. For this purpose, it may be useful to automatically generate the evaluation results according to the invention 17 For example, an RGB image of the same skin area superimposed represent. The evaluation results can advantageously also be stored in electronic form, linked with required information such as patient data, position data of the examined tissue site, etc., and, if appropriate, the measured raw data and, for example, enclosed with the patient's file.

When Measuring arrangements for recording suitable spectral and spatial resolved measurement data is in principle any arrangement in question, such data with for the application required spatial as well as spectral resolution can deliver. For practical use, it is advantageous to use measuring arrangements that store these data within one Time span of typically a maximum of 60 s, since otherwise the danger of significant falsification by movements of the considered tissue part relative to the measuring arrangement and thus, depending on the measuring method, distortions in the spectral or the spatial Dimension can occur, which is a reliable Make classification difficult or impossible.

A principle arrangement for the noninvasive in vivo measurement of spatially resolved, continuous spectral data of tissue areas is in 5 shown. A tissue part to be examined 30 is from one or more radiation sources 31 illuminated with preferably broadband, the required wavelength range adapted radiation characteristic. The tissue area considered by the 30 Remitted radiation is via an optic 32 and a slit 33 into an imaging spectrograph 34 passed, there spectrally decomposed and on a camera chip 35 with sensitivity adapted to the desired wavelength range. By linear displacement of the measuring arrangement relative to the tissue sample are thereby line by line, the raw data 10 for the subsequent evaluation as described above. The measurement data of the camera chip 35 become means of calculation 36 conducted, integrated into the arrangement or separately via a data connection with the camera chip 35 can be connected. In the Re chenmitteln 36 the processing and evaluation of the measured data takes place.

When measuring fluorescence data, it may be useful in addition, according to 6 in the beam path between the radiation source 31 and tissue 30 incorporate optical filters which define the excitation radiation wavelength wise. In addition, it is also possible, in particular between radiation source 31 and tissue 30 optical elements, such as optical fibers to provide.

A principle arrangement for non-invasive spatially resolved in-vivo measurement at defined discrete wavelengths is in 6 shown. The tissue part to be examined 30 is from one or more optical radiation sources 41 illuminated. The tissue area considered by the 30 Remitted radiation is via an optic 43 on a camera 45 projected. In this arrangement, it is possible and may be useful between the radiation source 41 and the tissue 30 to arrange further optical elements, such as optical fibers or optical fiber bundles. Likewise, it is possible and may, for example, for endoscopic applications, be useful in the optics 43 between tissues 30 and camera 45 integrate image-retaining fiber bundles. The measurement data of the camera 45 become means of calculation 46 routed into the arrangement or separately via a data connection with the camera 45 can be connected. In the calculation means 46 the processing and evaluation of the measured data takes place.

The Selection of the wavelengths and thus the spectral information can in arrangements of this principal kind in different ways respectively.

An advantageous possibility sees the use of a plurality of monochromatic or narrow-band radiation sources emitting at the relevant wavelengths 41 , For example, light-emitting diodes (LED) or laser diodes (LD) before. By alternating operation of these light sources 41 images are sequentially obtained at different wavelengths (when measuring the diffuse reflection) or at different excitation wavelengths (when measuring the fluorescence).

Alternatively, it is possible broadband radiation sources 41 to use and the emitted light through variable wavelength selectors 42 to filter. As wavelength selectors 42 In particular, filter wheels with a series of bandpass filters of different transmission wavelengths or tunable optical filters, for example so-called AOTFs or LCTFs, come into consideration.

As an alternative to filtering the illumination radiation, it is also possible for the selection of certain relevant wavelengths, the tissue 30 with one or more optical radiation sources 41 with preferably broadband, adapted to the required wavelength range radiation characteristic to illuminate and of the tissue 30 remitted radiation using a variable wavelength selector 44 to filter. In turn, a filter wheel with a number of different bandpass filters or tunable optical filters can be considered as the wavelength selector.

In particular, for the evaluation of fluorescence phenomena, it may be useful in addition, both between light source 41 and tissue 30 as well as between tissues 30 and camera 45 variable optical filters 42 . 44 provide, so as to be able to select both the excitation wavelength and the emission wavelength accordingly.

References:

• [1] Duda, RO; Hart, PE and Stork, DG, Pattern Classification; John Wiley & Sons (ed.), 2nd ed. (2001)
• [2] Martens, H., Naes, T., Multivariate Calibration, John Wiley & Sons (2002)
• [3] Foley, JD, van Dam, A., Feiner, SK, Hughes JF, Phillips RL, Introduction to Computer Graphics, Addison-Wesley Professional (1993)

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• WO 99/44010 [0009]
• WO 98/37811 [0009]
• US 6208749 [0009]
• - US 6081612 [0009]
• - EP 1185853 [0012]

Cited non-patent literature

• Ganster et al., IEEE Transactions on Medical Imaging, 20 (3), 233-239 (2001) [0006]
• Mclntosh et al., Vibrational Spectroscopy 28, 53-58 (2002) or Dwyer et al., Cancer Epidemiol. 7, 203-206 (1998) [0010]
• - Duda, RO; Hart, PE and Stork, DG, Pattern Classification; John Wiley & Sons (ed.), 2nd ed. (2001) [0070]
• Martens, H., Naes, T., Multivariate Calibration, John Wiley & Sons (2002) [0070]
• Foley, JD, van Dam, A., Feiner, SK, Hughes JF, Phillips RL, Introduction to Computer Graphics, Addison-Wesley Professional (1993) [0070]

Claims (18)

A method for status detection of skin tissue, based on an in-vivo acquisition of raw image data of a tissue part to be examined on a human or animal body and the evaluation outside the body and independent of the presence of the body, characterized in that the evaluation comprises: Spectral preprocessing of the raw image data, comprising - the compensation of inhomogeneities of the illumination and / or variations of the sensitivity or the dark current of the individual pixels of the camera, and - the calculation of reflection spectra for each pixel from the raw energy data of the examined tissue site and those of a constant spectral standard - spectral classifying the preprocessed raw image data, - morphologically classifying the spectral classification result, - determining an evaluation result from the classification, and - outputting the evaluation result to a user interface. Method according to claim 1, characterized in that in that the acquisition of the raw data comprises the following method steps: - Illuminate the tissue to be examined with one or more optical Radiation sources with the required wavelength range adapted radiation characteristics, - Depart the remitted from the considered tissue area radiation over an optic and a slit diaphragm in an imaging spectrograph, - spectral Decomposing the remitted radiation in the spectrograph, and - Project the spectrally decomposed data on a camera chip with the optical Wavelength range of adjusted sensitivity. Method according to claim 2, characterized in that that the method steps by linear displacement of the measuring arrangement be repeated line by line relative to the tissue sample. Method according to claim 1, characterized in that in that the acquisition of the raw data comprises the following method steps: - Illuminate the tissue to be examined with one or more optical Radiation sources with the required wavelength range adapted radiation characteristics, - Depart the remitted from the considered tissue area radiation over an optic and a wavelength selection on a camera chip with sensitivity adapted to the optical wavelength range, wherein preferably the wavelength selection by means of a tunable filter or a series of discrete individual filters he follows. Method according to claim 1, characterized in that in that the acquisition of the raw data comprises the following method steps: - Illuminate the tissue part to be examined selectively with radiation sources with narrow-band emission behavior or a broadband radiation source with corresponding optical filters between the radiation source and tissue section to be examined, or a combination of both, - Depart the remitted from the considered tissue area radiation over an optic on a camera chip to the optical wavelength range adapted sensitivity. Method according to claim 1, characterized in that that a spectral classification of the raw data preferably by a Linear Discriminant Analysis (LDA) or a Quadratic Discriminant Analysis (QDA) or by a multivariate regression method (Partial Least Squares Regression, PLSR). Method according to Claim 6, characterized that the method step of classifying further the sub-step a reduction in the dimensionality of the data before the spectral Classification includes spectral characteristics of spectral data preferably by a Principal Component Analysis (PCA) or an Independent Component Analysis (ICA). Method according to claim 7, characterized in that that the number of spectral features on five to 15 spectral features is reduced, which in one or more spectral ranges lie. Method according to claim 1, characterized in that that the step of classifying the sub-steps spectral classification and artifact correction. Method according to claim 9, characterized that in the process step of artifact correction artifacts from recording and / or classification, in particular pixel error, speckle noise or stray light artifacts by using morphological filters and Analysis of contiguous (pixel) areas of the Classification images are removed. Method according to claim 10, characterized in that that the method step of the spectral classification also parts of the morphological classification, for example by Use of a Contextual Classifier. Method according to one of the preceding claims, characterized in that in the step of the morphological Classify the classification images according to appropriate morphological criteria, in particular texture, homogeneity and / or size certain areas. Method according to one of the preceding claims, characterized in that after the step of the Determining an evaluation result determined evaluation result the evaluation of the examined tissue part as well as information about the location and geometric dimensions of the tissue changes contains. Method according to one of the preceding claims, characterized in that the step of outputting the evaluation result to a user interface the visualization the evaluation results and / or the storage in a data memory includes. Arrangement for the status detection of skin tissue, comprising at least one radiation source ( 31 . 41 ) for the emission of radiation to tissue ( 30 ), an optic ( 32 . 43 ) and a camera ( 35 . 45 ) for detecting radiation emitted by the tissue ( 30 ) is remitted, as well as computing means ( 36 . 46 ) for the preparation and evaluation of the measured values of the camera ( 35 . 45 ) according to the method of any one of claims 1 to 10. Arrangement according to claim 15, characterized by a between the optics ( 32 ) and the camera ( 35 ), imaging spectrographs ( 34 ). Arrangement according to one of claims 15 or 16, characterized by at least one between the optics ( 43 ) and the camera ( 45 ) arranged optical filter ( 44 ). Arrangement according to one of claims 15 to 17, characterized by at least one in front of the radiation source ( 41 ) arranged optical filter ( 42 ).