JP2000139846A - Method and device for creating diagnostic data in degree of wound to skin tissue of patient - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable inspecting a wound objectively by irradiating a wide range of skin with a white light, transmitting only the mitigated light to a spectrum channel, resolving it into a specified number of light paths and expressing at least one of surfaces having the skin wounds after stepwise spectral resolution and cluster analysis. SOLUTION: The white light is irradiated from the light source 1 to permit only the mitigated light from the tissue of skin 2 to reach a four-channel multi- spectrum camera 3. In the camera 3, the light is resolved into at least four spectrum channel light paths, a wave length range is stepwise extracted through the use of a continuous spectrum selecting cut-off filter and it is recorded as spectrum images 4. Then it is transmitted to an evaluation unit 5. The image are analyzed and classified, the wounds are classified as against the skin surface which are recorded at present and analysis concerning pixels are executed and they are displayed in a display unit 6. Thus, the state of a lesion is displayed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

FIELD OF THE INVENTION The present invention relates to a method and apparatus for generating diagnostic data for the degree of wounding of a patient's skin tissue. It applies in particular to the initial medical diagnosis, follow-up, documentation and storage of burn wounds for clinical use.

[0002] In clinical diagnosis, abnormal skin changes,
In particular, skin burns, like 200 years ago, are still diagnosed by the examining physician's eyes, i.e., after subjective and empirical visual debridement. In particular, in the case of a burn wound, due to the influence of the patient and the environment, the misdiagnosis is about 30 to 50% depending on the experience of the doctor. In most cases, the extent of the wound can only be ascertained after a few days or at the first surgical stage.

[0003]

2. Description of the Related Art In search of objective methods, ultrasonic methods such as "Wyllie et al. Burns 17 (1991) 123-128" and thermography "Cole et al. Burns 16 (1990) 60-6"
3 ", color method, the isotope method, NMR method (N uclear M agn
etic R esonance) [Nellelbad, et al., Bur
ns Vol. 22 (1996) No. 2, 117-119 ”and laser Doppler technology“ Essex ”et al., J Biomed En
g Vol. 13 (1991) 189-193; Niazi et al., Burns
19 (1993) 485-489, these methods are usually invasive (contamination of the patient) and are too dangerous, time-consuming or expensive for intensive care patients. Since the reproducibility of the results was also often poor, it was possible to achieve considerable clinical success without them. Even the work up to now on the catoptric method was not so good and was not actually performed. Trial within the visible wavelength range (Afromovitz (Afro
mowitz), besides, (1988)) was unsuccessful in clinical practice, especially because the number of errors was too high for different skin types.

In US Pat. No. 5,701,902, the use of fluorescence excitation (in the UV or visible spectral range) and simultaneous IR spectroscopy was attempted to characterize burns. This solution, however, is also invasive (fluorescent dye administered intravenously) and is limited to local measurements (≦ 1 mm ² ). Neither follow-up (reproducibility of the measurement locality) nor extensive recording (approximately 6 hours for a 10 cm × 10 cm area) is performed.

[0005] US Patent No. 4,693,255 discloses the use of dynamic changes in the appearance of burn tissue by a previously introduced marking dye, preferably a fluorescent dye, for analysis of a wide range of video recordings of skin burns for analysis. Disclose the record. This method is inadequate for immediate diagnosis and follow-up anyway because the method is invasive and dye dynamics occur during the 1-20 minute window.

[0006] US Pat. No. 4,170,987 discloses a medical skin diagnosis using a rotating mirror and three detectors on which the same pixels of a patient's skin sampled in a line scan are simultaneously imaged. The system and method are described. In this case, the black and white detectors are located behind different color filters (eg, green, red, IR). The intermediate (gray) values of the detector are digitized and stored after adjusting for contrast increase or matching dynamic range. From each of the three associated stored digital values per pixel, a ratio number is formed which is displayed or printed on a color monitor as a pseudo-color image.

[0007] In addition to the disadvantages of the scanning sampling method in principle for mobile applications, there are also the disadvantages mainly of high optical alignment costs and light attenuation resulting from multi-intensity decomposition. The high cost of this alignment is, in fact, only manageable in the case of beam resolution in one plane, and the undesirable light attenuation is also quite high with two splitter mirrors, so US Pat. No. 4,170,987. The system proposed in US Pat. No. 6,064,075 can only actually record intensity images up to three different wavelengths if the intent is to achieve pixel-synchronous image recording with a satisfactory signal / noise ratio.

[0008] Another general disadvantage of all known solutions to the diagnosis of the degree of skin wounds is that the measured values are determined by the individual skin type, the location of the burn, the time of measurement after the burn, the lighting conditions, and Expected initial treatment (ointment)
Depends on the fact that Since the absolute measurements are therefore of only limited meaning, most studies described in the documentation prove to be incorrect due to the corresponding standardization and interpretation models. The significance is that the measurement method is complicated by the addition of extra, sometimes unsuitable, measurement quantities, for example by extra measurement channels and / or invasive steps (dyes).

[0009]

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for reconstructing skin wounds that is objective and reproducible regardless of skin type and skin wound time. It is to provide a new method for assessing the degree of wounding on human skin. A specific object of the present invention is to provide a burn degree of 2a (2, superficial) to 2b (2,
Depth) and 3, and to create an objective diagnostic aid that reliably distinguishes healthy skin as well as scenes in non-patient related images.

[0010]

In accordance with the present invention, simultaneous imaging of the surface of skin tissue is performed on a plurality of different channels of the spectrum, and the resulting image data is generated from the different images of the spectrum to obtain various images of the skin. A method for generating data for diagnosing the degree of wounding on skin tissue of a patient, wherein the wounded area is displayed as an image,
Skin tissue is extensively illuminated with white light such that only substantially relaxed light from the surface of the skin tissue reaches the spectral channel for image recording; the relaxed light is broken down into at least four separate light paths And the spectrally defined wavelength ranges for these spectral channels are extracted stepwise using a continuous spectrally selective cutoff filter placed on the optical surface of the prism; after stepwise spectral decomposition, Narrow spectral filtering is performed on each spectral channel prior to photoelectric image conversion, and this bandpass filtering takes into account the characteristic changes in their relaxation ability when various skin wounds occur; Performed in parameter space, the centers and radii of the clusters are defined as the relaxivity of the degree of wounding for a known skin tissue pattern. Assignment of relaxation values of unknown skin tissue to these trained clusters for the extent of skin damage, determined and stored by training, is performed; various degrees of skin within the area of the recorded skin surface This object is achieved in the case of a method in which at least one representation of the surface with the wound is displayed based on cluster assignments.

It can be seen that it is advantageous to use a selectively reflecting cut-off filter for spectral decomposition of light relaxed into a wavelength range in order to spatially extract the spectral components stepwise. Preferably, in this case, a continuous reflection low pass filter with a stepwise increasing low pass cutoff wavelength is employed. Similarly, a high-pass filter having a progressively decreasing high-pass cut-off wavelength may be employed as the continuous cut-off filter.

Before the cluster analysis for the individual pixels assigned to each other from the simultaneously recorded spectral images, the ratios for generating the obtained relaxation values are given by the following equation:

(Equation 3) When formed, the different skin types in the image recording, different lighting, and it found to reduce the influence of other variables factors is particularly advantageous, where r _i
Shows digitized relaxation values of n different relaxation values recorded using narrow spectral bands for individual pixels on the imaged skin surface.

For many applications, it has been found that it is sufficient if the number n of relaxation values recorded using a narrow spectral band is chosen to be equal to four;
Parts of the green, red and near infrared (NIR) spectral range are used.

450 nm, 550 nm, 650 nm, and 8
00 nm is preferably adopted as the center wavelength of the narrow-pass filtering for the classification of skin burn degree, and the full width at half maximum of the band-pass filtering is selected in the range of 5 to 20 nm.

For a different degree of sharper representation of skin wounds, a pseudo-color representation of the surface of different skin wounds has proven advantageous.

[0016] In addition to different degrees of representation of skin wounds, true color images are recorded using red, green,
And blue spectral channels,
It is also possible to switch between the two expressions.

An image recording unit for simultaneously recording a plurality of spectrally different images of the surface of the skin tissue by means of a beam splitter, a suitable spectral filter and an image sensor, for generating image data obtained from the spectrally different images; An apparatus for generating data for the diagnosis of the degree of wounding of a patient's skin tissue, comprising: an evaluation unit including means for performing the measurement, and an output unit for displaying various damaged areas of the skin tissue. A lens for extensive collection of relaxed white light from the skin tissue is provided in the image recording unit to send the collected light to a beam splitter; the beam splitter includes a sequentially arranged prism. Wavelength multiplexer, at least four light mitigated by the skin surface These prisms, which in each case allow to be decomposed into different images, provide a spectrally selective cut-off for spatial extraction of the light beam in a limited wavelength range on their surface away from the lens A filter layer, wherein each of these selectively extracted light beams is narrow-pass filtered before impinging on the associated image sensor; data in the parameter space generated from different relaxation values of the image sensor spectrum The means for performing the analysis is in the evaluation unit, the data analysis comprising a comparison of the current data with data already stored in the same parameter space from the trained known degree of the skin wound, The above-mentioned objects are achieved in the case of a device, which is assigned a known degree of skin wound.

Conveniently, each cutoff filter layer is a reflective filter with a steep cutoff that acts as a lowpass filter, the cutoff of the lowpass filter being shifted stepwise from prism to prism to larger wavelengths. On the other hand, it is also possible that each cut-off filter layer is a reflective filter acting as a high-pass filter, and that the cut-off of each high-pass filter is progressively smaller wavelength from prism to prism. With this type of stage cut-off filter arrangement, high selectivity, high optical transmission and good image transfer quality for all spectral channels are obtained by beam decomposition.

To adjust the beam resolution and image recording, it can be seen that it is advantageous to arrange all the cut-off filters perpendicular to a common plane, which is the uniform plane of the spatial beam resolution of the spectrum. . The evaluation unit is the following equation

(Equation 4) Expediently comprises means for forming the proportions of the individual pixels of the different images of the simultaneously recorded spectra for the generation of the relaxation values obtained by, where r _i is the individual pixel of the imaged skin surface 9 shows digitized relaxation values of n different relaxation values measured using a narrow spectral band for each.

Furthermore, within the evaluation unit for the classification of skin wounds, in particular skin burns, the means for performing a cluster analysis are conveniently located in the parameter space of different relaxation values of the spectrum, the center and the radius of those clusters. Are determined and trained by training with wound magnitude reliefs against known skin tissue patterns, and the relaxation values of unknown skin tissue are assigned to these trained clusters of skin wound magnitudes, and hence those wounds. And within the output unit at least one means is advantageously provided for representing different degrees of surface of the skin wound in the area of the skin surface currently recorded based on the cluster assignment and the classification. .

Advantageously, for classification of skin burn degree, exactly four different spectral relaxation values recorded using narrow bands are used from portions of the blue, green, red and near infrared (NIR) spectral ranges. Is done. In this case, 450nm,
It has been found that 550 nm, 650 nm and 800 nm can be advantageously used as the center wavelength of the narrow-pass filtering, and the full width at half maximum of the bandpass filter is in the range of 5 to 20 nm.

The wavelength multiplexer of the device according to the invention preferably has a fourth wedge prism in the case of three wedge prisms for lateral coupling of the first to third image sensors.
Consisting of four-sided connecting prisms adjacent to these prisms for the linear coupling of the image sensor, in the case of wedge-shaped prisms, in the case of wedge-shaped prisms, in each case the reflection cut-off filter layer is arranged on the rear side At least one air layer is located on the front side for internal reflection of light extracted by the respective cut-off filter of the prism in the direction of the short prism side, and the image sensor is located on the cut-off filter. The prisms are installed so as to be orthogonal to the direction of reflection, and all the prisms are dimensioned such that the optical path length in each prism is the same as that of the image sensor.

The lens advantageously has a short object length and a large image length for imaging the skin surface via a wavelength multiplexer with different spectral illumination on an image sensor. Conveniently, lenses, wavelength multiplexers,
The image recording unit with cut-off filter layer, band-pass filter, image sensor and image memory is located in the compact video camera and data transfer to personal computer is provided only for cluster analysis and classification of skin wounds You.

To reduce the spurious effects of illumination, a strong white light source is securely connected to the video camera, and the illumination is aligned so that light directly reflected from the skin surface does not enter the camera lens.

The basic concept of the invention is to provide active illumination of skin tissue with white light, especially for the diagnosis of skin burns.
Organized simple multispectral detection method using mobile, easy-to-use, menu-driven digital video camera with CCD technology, high-intensity true color image and pseudo-color image that assign defined colors to parts of equal burn depth And an adaptive cluster evaluation method based on a medically proven learnable patient database for automated evaluation of data at the output.

From extensive preliminary studies, at 3 or more wavelengths, burn depths 2a (superficial), 2b (deep dermis)
It has been found that a more reliable determination of and 3 (all layers) is possible. In this case, the present invention is based on the discovery that the spectral data of healthy and (uniformly) burned skin forms a cluster structure in the color space described above, so that a number of skin types and their degree of burn are considered with respect to relaxation spectra. A clinically-proven database that describes and also contains patient and (clinical) processing data, and fuzzy cluster algorithms (eg, modified “fu
zzy c-means "algorithm, Bezdek, 1
981) allows for a unique assignment of measured values to clinically relevant burn depths (without extra reference measurements). In other words, a deliberate search for clustering of the current measurement data in the color space formed from the relaxation values is performed, and these measurement clusters are determined by determining the minimum deviation to determine the known degree of the wound. Assigned to already trained clusters. Thus, it becomes possible to assign each skin pixel having at least three obtained (by forming a ratio) relaxation intensity to a measurement cluster. After filtering and pre-classifying the measured data that should be assigned to the environment and unburned skin (ie, to dark or overexposed areas), the algorithm then proceeds to remove the stain (sharpness), the characteristic surface of the cluster Judge the shape (pattern recognition)
Adapt the "measured" color space coordinates to those of the database. In addition to spectroscopic data, this algorithm can be used to determine, for example, burn type, burn process, time after burn,
Also take into account patient data, such as the nature of the initial treatment.

The arrangement according to the invention makes it possible to obtain an assessment of the degree of wounding on human skin on the basis of the spectrally resolved video recording, and that the skin wound can be irrespective of the type of skin and the time of skin wounding. Objective and reproducible evaluation can be performed. In particular, the device of the present invention provides objective diagnostic aids that reliably distinguish between burn degrees 2a (2, superficial) to 2b (2, deep) and 3 and healthy skin. The present invention will be described in detail below with reference to the accompanying drawings.

[0028]

DETAILED DESCRIPTION OF THE INVENTION The method according to the invention consists essentially of the following steps:-The skin tissue has essentially only reduced light from the surface of the skin tissue reaching the spectral channel for image recording. The relaxed light from the skin surface is decomposed into at least four separate light paths, such that the wavelength range defined by the spectrum of the spectral channel is a continuous spectrally selective cut-off filter A narrow-spectrum filtering is performed on each light path appearing after the step-wise spectral decomposition before the photoelectric image conversion, this band-pass filtering being the ability of the skin to relax when the skin is wounded Cluster analysis is performed in the parameter space of the relaxation values, and the center and radius of the cluster are Assignment of relaxation values of unknown skin tissue to these trained clusters of skin wound degree, determined and trained by training with wound degree relaxation values for the tissue pattern, is performed; Surface representations with different degrees of skin wounds within are displayed based on the cluster assignment.

This basic idea is schematically represented in FIG. The light source 1, preferably a halogen lamp, intensively illuminates the skin tissue to be examined, for example burned skin, with white light. Moderated light from the skin surface (directly reflected light is removed from the image record by the illumination geometry)
Is, as outlined in part 1 (image collection) of FIG.
Recorded by a multispectral camera, described in further detail below, of at least four spectral channels. These recorded spectral images 4 are shown in FIG.
Of the evaluation unit 5, preferably a personal computer (PC), laptop or other computer unit, without pre-processing, ie without being compressed or processed in any way, by means of two parts (image transfer) Is forwarded to According to the third part of FIG. 1 (image analysis and classification), as will be described in more detail below, the spectral image 4 is generated by the appropriate software, the formation of specific ratios, to the currently recorded skin surface relative to the skin surface. Wound classification and analysis are performed on pixels using a comparison of the degree of relaxation of the wound with known skin tissue patterns. From this, it is possible to display the affected part in various ways on the display unit 6. On the other hand, color display in true color (true color image) is particularly suitable for visualizing burn wounds for indirect visual inspection of burned persons. On the other hand, in particular, a pseudo-color display is preferred as a display mode in which different degrees of burns can be clearly separated locally. Both display formats are suitable for output as hard copy or monitor display and are selected as needed. Without limiting its scope, in order to make the expression simpler and technically clear, the present invention is particularly preferred for the classification of the degree of skin burns.
This is described below with reference to the channel structure.

[0030]

Detailed Description In the case of exactly four wavelengths, the choice of wavelength is determined by a simple layer model of the skin (Anselmo et al., 197).
7, Afromowitz and others, 1987, 198
This is done using 8).

As shown in FIG. 2, the light mitigated by the human body (and the affected area of the burn) is collected by a special lens 31 having a long image distance and a motor driven objective focal length adjustment, and a long wave infrared (IR) light. The light passes through a block filter 32 which forms a barrier to light and reaches a wavelength multiplexer 33 in the form of a small prism block. In most cases, the wavelength multiplexer 33 will be configured to split the various wavelength arrangements in the same plane without substantial loss into separate optical paths and to apply the appropriate banding before impacting the corresponding CCD matrix sensor. Consisting of a number of prisms firmly glued together so that they are filtered according to the selected center wavelength and the associated full width at half maximum only directly using a pass filter. This at the same time ensures that the level of selectivity is high, the optical transmission is high, and good image transfer quality is maintained. By using a flat-plate arrangement (exceeding 4 wavelengths, an out-of-plane arrangement is required) and using advanced joining technology, CC
Extremely precise and robust alignment of the D modules with respect to each other is achieved, thereby ensuring pixel synchronization between the individual pixels of the CCD sensor.

In the particular case as shown in FIG. 2, the wavelength multiplexer 33 has three three-sided prisms 331-331.
333 and a (four-sided) connecting prism 334. Three-sided prism 331 to 331 away from lens 31 (hence the incident light)
The color separation layer 34 in each case is
There are 1 to 343, which are reflection filter layers,
Since it has a high-pass filter characteristic of reflection, high-frequency light up to a predetermined filter cutoff is almost completely reflected,
Low frequency light below that cutoff is transmitted with little attenuation. Due to the different slopes of the surfaces of the prisms 331 to 333 carrying the color separation layers 341 to 343, the reflected spectral components of the white light incident through the lens 31 are extracted from the incident light path (shown by dotted lines). You. To increase the spatial separation required for detection, the surface of the preceding prism 331-333 in the optical 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 between the prisms 331 and 332 and between the prisms 332 and 333
And the exposed front surface of the prism 331. These two voids are preferably
It is configured by an adhesive bonding method using a mask with a hole that does not obstruct the optical path. The special, somewhat peculiar shapes and sizes of the prisms 331-334 have shown that directing light at the same light wavelength to sensors in all spectral channels is very useful for multispectral recording techniques. Therefore, it is determined within the narrow limit. in this case,
Minimum possible angle of spectrally selective reflection on color separation layer and
A compromise must be found between the maximum possible space savings for combining sensors in the form of CCD sensor blocks 361-364.

The light selected in the spectrum reflected internally on the front surfaces of the prisms 331-333 is applied to the CCD sensor blocks 361-361 at the third short side of the prism.
64, in particular, the short prism faces are vertical in each case, and then the narrow-spectrum filters (SF) 351-351.
353 and the CCD sensor blocks 361 to 363 are extracted so as to penetrate. In this case, the spectral filters 351 to 353 and the CCD sensor block 361
363 are directly adhesively bonded to the short prism surface.
The same is true for the elements located in the optical path direction: the connecting prism 334, the spectral filter 354, and the CCD sensor block 364.

For those CCDs, the CCD sensor blocks 361-364 are at least video standard with respect to pixel count. The size and number of pixels is determined by the spatial resolution of the burn, <1 to about 3 mm, and the measurement distance, typically 30 cm (local recording) to 200 cm (whole body recording).

With respect to the color separation layers 341-343, the desired photo separation by color selection is specifically shown in FIG. 3 by stepwise high pass filtering. The cutoff filters used are distinguished by a sharp cutoff in the bandpass wavelength range and almost perfect reflection extraction. FIG. 3 also illustrates the bandpass selection of the narrow spectrum filters 351-354 for wavelength. Is calculated, based on more precise experimentally determined values, the apparatus of the present invention, the wavelength (450 ± lambda ₁₎ nm (blue),
(550 ± lambda ₂₎ nm (green), (650 ± lambda ₃₎
Operating at nm (red) and (800 ± Lambda ₄ ) (NIR), full width half maximum lambda n is typically in the range of 5 to 15 nm and is optimal for what is achievable with respect to stability evaluation algorithms and engineering Where the blue spectral range was selected with a fairly large bandwidth (20 nm). These bandwidths are a compromise between the camera sensitivity (maximum possible bandwidth) and the characteristics of the skin absorption curve (minimum possible bandwidth). In the blue range, there is virtually no strong spectral variation in the soothing ability of burned skin, but there is a reduction in intensity from conventional available light sources (halogen emitters), so this is compared to other band filters. It should be noted that the compensation is made with the larger bandwidth of the band filter. Elsewhere, the choice of the center wavelength of the bandpass filter depends on the absorption and dispersion of the relevant skin components, especially hemoglobin. 450 nm: relatively strong component in upper layer 550 nm: strong dependence on blood volume component in second layer (hemoglobin absorption) 650 nm: strong relative difference in absorption between HbO ₂ and Hb 800 nm: rarely dispersed but strong The isosbestic point between HbO ₂ and Hb, which is empirically seen in the third degree burned area.

The skin model used so far has been expanded,
Where other optically active skin components for burns are considered, it may be logical to use other wavelengths.
These will increase the size and shape of the parameter space (color space) 52, but nonetheless, the following analysis and classification algorithms will not (strictly) change.

The four (or more) CCD (or other image sensor, such as, for example, CMOS active pixels) matrix sensor is used to collect the relaxation values from the burned area, and the damaged skin tissue (and Surrounding healthy skin) (independent or image recording unit (since its basic function is the same as a video camera,
White light source 1 connected to "camera" below)
Illuminated continuously or pulsating (to save battery power). This provides intense illumination at the wavelength intervals described above and does not distort the measurement results because they are "outshine" (lighter) than multiple ambient light sources.
Due to the strong scattering of light from the (burned) skin, the relaxation intensity is independent of the angle of illumination. Avoiding direct reflection (illuminating the wound at an angle perpendicular to the burned skin surface) ensures that the relaxation intensity is much greater than the reflection intensity. For that part, the camera is automatically "color calibrated" to the illumination used by the internal black and external white reference (standard) colors. With the use of (adjustable) global illumination and a slightly smaller, but comparable field of view camera lens, extensive and even whole body recording is possible in addition to local areas.

The mobile camera can be replaced at any time and uses a rechargeable battery that can be operated for at least 20 minutes with maximum driving and lighting (100 W) and experience is sufficient for recording. Powered. The electronic configuration of the camera is configured to be able to create a video mode that sends a color image to the viewfinder of the camera, as in a professional video camera. At the same time, menus for controlling and monitoring the camera are superimposed in the viewfinder (if needed). As the camera monitors itself, during recording pauses, the light is switched off after a certain period of time (to conserve battery) or the battery level of the charge is controlled by LEDs in the viewfinder and on the camera Is displayed.

When the doctor recognizes in his view that in his opinion it is the best recorded scene, this scene is frozen in the viewfinder by pressing a button (switch setting "freeze") and the observer is , Whether the image is usable, ie should be stored (switch setting “safe”), or its record should be discarded, ie return to video mode (unswitch)
Make a prompt decision on what to do. Since mobile cameras are equipped with a PC chip and flash memory, about 10 (in 4 × 1.5 Mbit b / w images) or more records can be selectively stored in the camera. Records already stored are no longer lost in principle.

After the end of the recording session or when the memory is full (indicated automatically in the viewfinder), the data is transferred to a PC (laptop) hard drive, which may be located remotely. Transferred from the camera (second stage in FIG. 1). This transfer takes place under serial data exchange and can only be performed if the patient identification number has been previously entered.

In the burn record, each image is assigned (at least) to a narrow spectral channel λ ± lambda.
Consists of a 4b / w image. Each "skin pixel" (several mm
² ) is calibrated for total relaxation intensity via the pixels of the CCD sensor by optical imaging and wavelength resolution,
One of the independent variables given, hence are assigned to the four relaxation intensity I _xy 1 to I _xy 4 tensioning a three-dimensional color space 52.

The problem lies in assigning these measured values to "normal" skin and burn rate, since there is no normal skin and the skin (layer) model is It is greatly simplified in comparison because there is a lack of spectroscopic in vivo measurement data for different burn depths for different (individual!) Skin types, different body regions, and different initial processing methods. Therefore, 3
A coordinate axis in a dimensional color space cannot be considered absolute.

The present invention is based on the finding that the spectral data of healthy and (non-uniformly) burned skin forms cluster structures in the color space 52 described above, and therefore a number of skin types and alleviations. Through the use of a clinically proven database describing their burn rates in terms of spectroscopy and fuzzy cluster algorithms (eg, a modification of the “fuzzy c-means” algorithm of Bezdek, 1981) Only one assignment of the measurement to the upper relevant burn depth is possible (without additional reference measurements). Whereby the (at least) can be assigned a measurement clusters each skin pixel having three resulting relaxation strength r _xy 1 to r _xy 3. The obtained relaxation strength is obtained by the following equation, as shown in FIG.

(Equation 5) Here, r _i denotes the captured digitized relaxation values of the recorded n different relaxation values using narrow spectral bands for individual pixels of the skin tissue.

As schematically shown in FIG.
The spectral relaxation intensities collected for pixels from the burned skin are identified and grouped as clusters in color space 52 with a particular intensity ratio. After filtering and pre-classifying the measurement data that should be assigned to the environment and unburned skin (ie dark or overexposed areas), a software algorithm determines the blurred areas (blurred), the distinctive surface shape of the cluster (Pattern recognition) and adapt the coordinates of the "measured" (measured) color space to those in the database. This involves, in a first step, determining the existing different states (clusters) in the parameter space 52 and assigning image pixels to these clusters, and in a second step classifying these state clusters 53 through pattern comparisons 2 Corresponds to a stepwise procedure, whereby the pixels are similarly categorized by the work going back. FIG. 5 shows the qualitative dependence for cluster formation in the upper part of the model and the intensity relationship in the lower part suitable for the degree of human skin burn which is empirically found by appropriate ratio formation. .

FIG. 6 shows the entire basic procedure of the evaluation algorithm. This starts with the recording of the measured data in the four spectral channels and transforms the spectral relaxation values into (n-1) parameter space (by forming ratios, here at four wavelengths: three ratios). , And cluster assignment and cluster assignment up to the classification of the cluster into skin burn degree by comparison with the patterns already classified in the pattern database.

In addition to the spectroscopic data, the algorithm includes, for example, burn type, burn process, time after burn,
Also take into account patient data, such as the type of initial treatment.

During the training process of this method, a representative distribution of burn rates in color space is determined as a basis for cluster classification and the assignment algorithm is trained (axis scaling). This learning phase is also used for statistical confirmation of the classification by comparing the evaluation with the results of histomorphometric studies.

This training is performed using a number of burn cases in microstructure studies on specimens from the relevant tissue sections and the histological assignment of biopsies to different burn degrees.
To do this, for the first three days after the burn,
Locally treated sites are measured and assessed at set time intervals. In addition, a biopsy is taken from the burn tissue during the treatment phase and the burn result is determined histologically.

This evaluation method is further distinguished by its suitability. It does not have a set segment limit, as usual, and thus a strict classification of individual pixels,
Continuously adapted. This is a two-step procedure that first determines the different states (clusters) present in the parameter space, assigns image pixels to these clusters, and classifies these state clusters through pattern comparisons in a second step. , And so on by going back and working on the pixels as well. This avoids unreliable classification due to skin color and the high degree of variation of the fixed segment limit.

The present invention has advantages, inter alia, in clinical diagnosis. Therefore, the degree of burn 2a (2, superficial),
2b (2, deep), 3, provide healthy diagnosis and objective diagnosis support means for reliably discriminating scenes in images that are not related to the patient. With pictorial true and false color representations, objective aids are provided for surgical planning and related procedures. The apparently severe burn depth is determined immediately after wounding. Objective detection allows burn tracking for the first 3-4 days, after which the wounds begin to change and a new algorithm must be employed. In particular, the practitioner has a burn degree of 2a to 2
A diagnostic system is also provided, which also allows to further extend the limits of the traditional wound treatment of the b range, with the above-mentioned positive effects relating thereto. The physician is thus provided with a reliable and reliable method of diagnosis and reproducibility that not only takes into account the effects of the various therapies but also allows the use of auxiliary databases. Reliable objective diagnostic results automatically account for differences between various body parts through the detection of scenes that also show various human skin types and unburned skin in close proximity to the burn in the image To be obtained. Separate calibrations for healthy or fictional reference skins are therefore unnecessary.

The physician is therefore equipped with a mobile image recording system which is easy to operate and makes the most of mobility and is therefore also patient-friendly. There is basically also the opportunity to perform whole-body recordings in addition to local burn wounds. Possible objective determination of the need for whole body recordings, and thereby the ratio of skin area burned to different depths to whole body area, sends severe burn victims to the hospital and plans the required surgery and procedures This is a very important aid in initiating the intensive medical treatment needed (in stages). This opens up the opportunity to perform only the skin grafts that are absolutely necessary in the early stages and thus provide a better opportunity for recovery and a basis for shorter treatment times for skin wounds, especially burns.

An improvement of the method according to the invention, and the associated device, performs the analysis in real time because of the PC chip contained in the camera as well as the independent (electrically) unconnected PC. To perform image evaluation by using an additional memory module to provide the physician simultaneously examining the result display or a different display in the finder with the pseudo color representation described above, in addition to the true color image. is there. This is for imaging purposes and for online remote consulting
It is also possible to have online transfer of these data in real time. For this purpose, data can be displayed to nearby individuals on a display additionally attached to the camera, or through a cable connection or endotracheal tube to any desired receiving module inside or outside the operating room Via a data link, it may then be transferred simultaneously worldwide.

By enhancing the image processing software with the area calculation module, the numerical surface occupancy of the burn area can also be stored accurately (by comparative scale in the image or by extrapolation from body dimensions). Becomes

Furthermore, the method of the present invention also allows, for example, the use of an endoscope for examining "burns" in the human body that occur inside the lungs when inhaling hot air. Vitality and "face" measurements of cosmetic surgery implants are also within the possible uses of the present invention. It is also true for measurements and treatments of carcinogenic skin changes (in addition to individual carcinomas,
If the distribution is dense, extensive parallel analysis
It will also be possible to indicate to a physician examining only those carcinomas that may or may not be positive [missing].

A further use consists in measuring the condition and subsequently treating open wounds. The method according to the invention (in particular,
With the aid of the invention described here, for example, with the aid of laser ablation, both on its surface and in its depth, the future potential of debridement of burn wounds and automation of the removal of “necrotic” tissue) It is also possible to observe the effect of a skin engineering procedure (eg, laser treatment) by taking a number of surface dimensions that have been used and treated.

[Brief description of the drawings]

FIG. 1 shows a diagram of the essential functional components according to the invention.

FIG. 2 shows an image recording unit according to the invention with a wavelength multiplexer for four wavelengths.

FIG. 3 illustrates the concept of color separation of a four-channel image record according to the present invention.

FIG. 4 shows the principle of image data analysis according to the invention applied to the degree of skin burn.

FIG. 5 represents the skin relaxation value as a function of the skin burn degree in the spectral channels selected according to the invention.

FIG. 6 illustrates software functions according to the present invention for adaptive classification of multispectral sensitive image data.

[Explanation of symbols]

1 light source, 2 skin tissue / skin surface, 3 cameras, 31
Lens, 32 IR block filter, 33 wavelength multiplexer / prism block, 331-333 (3
Surface) prism, 334 (4 surfaces) connecting prism, 34
Color splitter layer, 341-343 Color splitter layer / reflection filter layer, 35 spectral bandpass filter, 351-354 Narrow spectral filter (blue, green, red, NIR), 36 CCD sensor, 361-3
64 CCD sensor block, 4 spectral images, 5 evaluation units / computer, 51 spectral relaxation intensity, 52 clusters in parameter / color space, 53 state clusters, 54 database, 6 display units,

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (Reference) H04N 7/18 G06F 15/64 B (72) Inventor Bernd Lustermann Germany Day 07747 Jena Cheekstrasse 34 (72) Inventor Werner Reilland Germany Day 07751 Cospeda Stadt Jena Falken Wake 12 (72) Inventor Werner Eisenweis Germany Day 23636 Gross Grenau Mullenkamp 12

Claims (21)

[Claims] 1. Simultaneous image recording of the surface of skin tissue is performed on a plurality of different spectral channels, and the obtained image data is generated from the different spectral images, and various wounded areas of the skin are displayed as images. Generating data for diagnosing the extent of a wound on a patient's skin tissue, wherein substantially reduced light from the surface of the skin tissue reaches the spectral channel for image recording. The skin tissue is extensively illuminated with white light such that the relaxed light is at least 4
Resolved into two separate optical paths and the wavelength range defined by the spectrum of the spectral channel is extracted stepwise using a continuous spectrally selective cutoff filter placed on the optical surface of the prism; After decomposition, narrow-spectrum filtering is performed on each spectral channel before the photoelectric image conversion, this bandpass filtering taking into account the characteristic changes in their relaxation ability when various skin layers are wounded; The analysis is performed in the relaxation parameter space, and the centers and radii of the clusters are determined and trained by training with a wound degree relaxation value for a known skin tissue pattern, and these trainings for the degree of skin damage are stored. Assigning a relaxation value of the unknown skin tissue to the recorded cluster; the recorded skin surface At least one representation of said surface having varying degrees of skin wounds in the area is displayed based on the cluster allocation method. 2. A method for spatially extracting spectral components stepwise, wherein a cut-off filter effective for reflection is used for said spectral decomposition of said relaxed light within a wavelength range. The method according to claim 1, wherein 3. A low pass filter having a stepwise increasing low pass cutoff wavelength is employed as a continuous reflection filter for said spectral decomposition of said relaxed light. the method of. 4. A high pass filter having a progressively decreasing high pass cutoff wavelength is employed as a continuous cutoff filter for said spectral decomposition of said relaxed light. The described method. 5. The ratio for generating the obtained relaxation values is calculated by the equation: ## EQU1 ## before cluster analysis for each assigned pixel from simultaneously recorded spectral images. Where r _i denotes a digitized relaxation value of n different relaxation values recorded using a narrow spectral band for individual pixels of the imaged skin surface. The method of claim 1, wherein 6. The number of relaxation values recorded using a narrow spectral band, n, is selected to be equal to 4;
2. The method according to claim 1, wherein parts of the green, red and near infrared (NIR) spectral range are used. 7. 450 nm, 550 nm, 650 nm, and 800 nm are adopted as central wavelengths of narrow-pass filtering for skin burn degree classification, and said bandwidth is selected within a range of 5 to 20 nm. 7. The method of claim 6, wherein: 8. The method of claim 1, wherein the pseudo-color representation is selected for varying degrees of sharper representation of the skin wound. 9. In addition to said representations of varying degrees of skin wounds, a true color image is composed of red, green, and blue spectral channels recorded using a narrow band,
9. The method according to claim 8, wherein the two expressions can be switched. 10. An image recording unit for simultaneous recording of a plurality of different images of the surface of the skin tissue by means of a beam splitter, a suitable spectral filter and an image sensor, obtained from the different images of the spectrum. Generating data for diagnosing the degree of wounding of a patient's skin tissue, comprising an evaluation unit including means for generating image data and an output unit for displaying various damaged areas of the skin tissue. A device for extensive collection of relaxed white light from the skin tissue is provided in the image recording unit for sending the collected light to the beam splitter; The splitter is a sequentially arranged prism wavelength multiplexer, which is located on the skin surface. Thus allowing said light to be decomposed into at least four different images,
In each case, the prisms have a spectrally selective cut-off filter layer for spatial extraction of the light beam in a limited wavelength range on their surface remote from the lens, and these selectively Means for performing data analysis in a parameter space generated from different relaxation values of the spectrum of the image sensor, wherein each of the extracted light beams is narrow-pass filtered before impinging on the associated image sensor; Wherein said data analysis comprises a comparison of current data with data already stored in the same parameter space from a degree of skin wound of a known skin tissue, said current data representing a known degree of skin wound. Assigned, device. 11. Each cutoff filter layer is a reflective filter having a steep cutoff that acts as a lowpass filter, the cutoff of the lowpass filter being shifted stepwise from prism to prism to larger wavelengths. 11. The device according to claim 10, wherein: 12. A reflection filter having a steep cutoff in which each cutoff filter layer acts as a high-pass filter, wherein the cutoff of each high-pass filter is stepwise smaller at wavelengths from prism to prism. 11. The device of claim 10, wherein 13. The apparatus of claim 11, wherein all of the cutoff filters are arranged perpendicular to a common plane that is a uniform plane of spatial beam decomposition of the spectrum.
Or the apparatus according to 12. 14. The evaluation unit according to claim 1, wherein By having a means for forming a ratio of the individual pixels of the simultaneously recorded images with different spectra for generating the resultant relaxation values, where r _i, the individual image forming skin surface 11. The arrangement of claim 10, wherein the arrangement shows digitized relaxation values of n different relaxation values measured using a narrow spectral band per pixel. 15. Means for performing a cluster analysis in the parameter space of different relaxation values of the spectrum is provided in the evaluation unit, wherein the center and radius of the cluster are reduced to a degree of a wound relative to a known skin tissue pattern. Values are determined and stored by training on the values, and the relaxation values of the unknown skin tissue can be assigned to these trained clusters of skin wound grade, so that they can be classified with respect to those wounds; 11. The arrangement of claim 10, further comprising at least one means for representing varying degrees of surface of a skin wound in an area of the skin surface currently recorded based on the cluster assignment and classification. Constitution. 16. Four different spectral relaxation values recorded using narrow bands from portions of the blue, green, red and near infrared (NIR) spectral ranges are used to classify the degree of skin burn. 16. The method according to claim 15, wherein
The configuration described. 17. 450 nm, 550 nm, 650 nm, and 800 nm are provided as center wavelengths of narrow-pass filtering for classification of skin burn degree, and the full width at half maximum of the band-pass filter is in the range of 5 to 20 nanometers. 17. The device of claim 16, wherein the device is provided. 18. The wavelength multiplexer comprises three wedge prisms for laterally coupling the first to third image sensors, and adjacent to these prisms for linear coupling of a fourth image sensor. Having a four-sided connecting prism, with respect to the light coming from the lens in the case of the wedge prism, the reflective cut-off filter layer is arranged in each case on the rear side; The image sensor is arranged on the front side for internal reflection of the light extracted by the respective cut-off filter of the prism in the side direction, on which the image sensor is orthogonal to the direction of the internal reflection of the image sensor. The prisms are dimensioned such that the optical path length in each prism is equal. 17. The arrangement according to claim 16, wherein: 19. The lens of claim 19, wherein the lens has a short object length and a large image length for imaging the skin surface via the wavelength multiplexer with different spectral illumination on the image sensor. Item 19. The arrangement according to Item 10 or 18. 20. The image recording unit comprising the lens, the wavelength multiplexer, the cut-off filter layer, the band-pass filter, the image sensor and the image memory is arranged in a compact video camera, and the data transfer to a personal computer is performed. 11. The arrangement of claim 10, provided for cluster analysis and classification of the skin wound only. 21. An intense white light source is securely connected to the video camera, and illumination is aligned so that light reflected directly from the skin surface does not enter the camera lens. The configuration described.