WO2019168381A1 - Apparatus for automatic classification of skin disease, and method for automatic classification of skin disease - Google Patents

Abstract

Disclosed are an apparatus and a method for automatic classification of a skin disease and a recording medium, the apparatus and method being capable of automatically classifying a skin disease, such as melanoma, with a high level of accuracy on the basis of the pixel features of an image captured of the skin disease, and gray level co-occurrence matrix (GLCM) features and gray level run length matrix (GLRLM) features extracted from a GLCM and a GLRLM converted from the image. The method for automatic classification of a skin disease, according to an embodiment of the present invention, comprises the steps of: producing a histogram of the brightness values of pixels in an image captured of a skin legion, and extracting pixel features by means of statistical analysis of the histogram; converting the pixel values of the image to a gray level co-occurrence matrix (GLCM) representing the adjacency between the pixel values; extracting GLCM features on the basis of the pixel values of the gray level co-occurrence matrix; converting the pixel values of the image to a gray level run length matrix (GLRLM) representing the run length of the pixel values; extracting GLRLM features on the basis of the pixel values of the gray level run length matrix; and classifying the skin disease by applying the pixel features, the GLCM features, and the GLRLM features to a machine learning model.

Description

Skin disease automatic classification device and skin disease automatic classification method

The present invention relates to an apparatus and method for automatically classifying skin diseases, and more particularly, from a pixel characteristic of an image photographing a skin disease, a gray degree coexistence matrix (GLCM) and a gray degree continuous length matrix (GLRLM) converted from an image. The present invention relates to an automatic skin disease classification device and an automatic skin disease classification method that can automatically classify skin diseases such as melanoma based on the extracted GLCM and GLRLM features.

Recently, the life span of human beings is gradually extended due to the change of living environment and the development of medical technology. Therefore, early response to diseases caused by lowering immunity in aging society has become important. In particular, senile skin disease is difficult to detect early, and when the treatment is missed progress to a malignant disease may be difficult to treat. Melanoma is one of the most common diseases of the senile skin disease, and initially has a modality similar to nevus. Because of this, melanoma is difficult to detect early in the onset and is often mistaken for nevus. Melanoma is a disease in which malignant tumors spread from the dermal layer of the skin to the muscles. Unlike nevus, the symptoms appear in the affected areas such as vascular malformation, melanin pigmentation and malformation. The most malignant malignant melanoma has no noticeable symptoms such as itching or pain, and it is characterized by ordinary black spots.

Abbasi et al. Developed the ABCD (Asymmetry, Border, Color, Diameter) method, which allows doctors to visually distinguish melanoma based on clinical experience. However, this method is a method of visually discriminating the shape of a lesion, and since subjective judgment of a specialist is involved, the same image may be diagnosed differently for each specialist. Bowling et al. Used a method that reduces the errors that can be caused by observing the ABCD method, which was previously observed with the naked eye, in more detail using dermoscopy. However, the conventional skin mirror is a method of attaching a microscope optical lens to a general camera to enlarge the disease area, but there is a problem of using a conventional method of reading by the naked eye. Hahm et al. Attempted automatic classification of melanoma and nevus by using SVM (support vector machine) in dermoscopic images, but it was difficult to accurately detect the skin condition, hair, and light sources. .

According to the present invention, skins such as melanoma and the like are based on pixel features of an image of a skin disease image, a GLCM feature extracted from a gray degree coexistence matrix (GLCM) and a gray degree continuous length matrix (GLRLM) converted from an image, and a GLRLM feature. An apparatus and method for automatically classifying skin diseases capable of automatically classifying diseases with high accuracy, and a recording medium.

The problem to be solved by the present invention is not limited to the above-mentioned problem. Other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, there is provided a method for automatically classifying skin diseases, comprising: calculating a histogram of brightness values of pixels in an image of a skin lesion, extracting pixel characteristics by statistically analyzing the histogram; Converting pixel values of the image into a gray-level co-occurrence matrix (GLCM) representing adjacency between the pixel values; Extracting a GLCM feature based on pixel values of the gray degree co-occurrence matrix; Converting pixel values of the image into a gray level run length matrix (GLRLM) representing a continuous length of the pixel values; Extracting a GLRLM feature based on pixel values of the gray continuous length matrix; And classifying the skin disease by applying the pixel feature, the GLCM feature, and the GLRLM feature to a machine learning model.

The pixel characteristic may include an average, standard deviation, skew, kottosis, entropy, and root mean square of pixel values of the image.

The GLCM feature is characterized by auto-correlation, contrast, correlation, dissimilarity, energy, and homogeneity of pixel values of the grayscale co-occurrence matrix. It may include.

The GLRLM features include short run emphasis (SRE), long run emphasis (LRE), gray level non-uniformity (GLNU), run percentage (RPN), and RLNU (run) extracted based on pixel values of the gray-scale continuous length matrix. Length Non-Uniformity) and High Gray Level Run Emphasis (HGLRE).

The machine learning model may include a support vector machine (SVM) model.

In accordance with an aspect of the present invention, a method for automatically classifying skin diseases may include: obtaining a dermatoscope image by photographing the skin lesion using dermoscopy; Converting the dermal image into a binary image and detecting an object of a smaller size than the skin lesion in the binary image; Removing noise including hair and skin keratin based on the size difference between the skin lesion and the object; And extracting a region of interest including only skin lesions by converting pixel values of normal skin to 0 in the dermal image.

In the method for automatically classifying skin diseases according to an exemplary embodiment of the present invention, the machine learning model may be acquired by learning the pixel feature, the GLCM feature, and the GLRLM feature based on learning image data including melanoma images and nevus images. The method may further include generating. The step of classifying the skin disease may include classifying the skin lesion into any one of melanoma and nevus based on the learned machine learning model.

According to another aspect of the present invention, a computer-readable recording medium having a program recorded thereon for executing the method for automatically classifying skin diseases is provided.

According to another aspect of the present invention, in the automatic skin disease classification apparatus including at least one processor, the at least one processor, calculates a histogram of the brightness values of the pixels in the image of the skin lesion, the histogram Statistical analysis to extract pixel features; Converting pixel values of the image into a Gray-Level Co-occurrence Matrix (GLCM) representing adjacency between the pixel values; Extract a GLCM feature based on pixel values of the gray degree co-occurrence matrix; Converting pixel values of the image into a gray level run length matrix (GLRLM) representing a continuous length of the pixel values; Extract a GLRLM feature based on pixel values of the gray continuous length matrix; And apply the pixel feature, the GLCM feature, and the GLRLM feature to a machine learning model to classify skin diseases.

The at least one processor is further configured to convert a dermoscopic image of the skin lesion by a dermoscopy into a binary image; Detecting an object of a smaller size than the skin lesion in the binary image; Removing noise including hair and dead skin cells based on the size difference between the skin lesion and the object; The pixel value of the normal skin may be converted to 0 in the dermoscopic image to extract an ROI including only skin lesions.

The at least one processor generates the machine learning model by learning the pixel feature, the GLCM feature and the GLRLM feature using training image data including melanoma images and nevus images; The skin lesion may be classified into one of melanoma and nevus based on the machine learning model trained using the training image data.

According to an embodiment of the present invention, based on a pixel feature of an image photographing a skin disease, a GLCM feature and a GLRLM feature extracted from a gray degree coexistence matrix (GLCM) and a gray degree continuous length matrix (GLRLM) converted from an image Provided are an apparatus and method for automatically classifying skin diseases and a recording medium capable of automatically classifying skin diseases such as melanoma with high accuracy.

The effects of the present invention are not limited to the effects described above. Effects that are not mentioned will be clearly understood by those skilled in the art from the present specification and the accompanying drawings.

1 is a schematic flowchart of a method for automatically classifying skin diseases according to an exemplary embodiment of the present invention.

2 is a block diagram of an automatic skin disease classification apparatus according to an embodiment of the present invention.

3 is a flowchart of a method for automatically classifying skin diseases according to an exemplary embodiment of the present invention.

4 is a detailed flowchart of step S30 of FIG. 3.

5 is an exemplary diagram of size distribution diagrams of objects extracted from a dermoscopic image.

6 is an exemplary view of a process of preprocessing a dermoscopic image according to an embodiment of the present invention.

7 is a block diagram of a skin disease classification unit constituting an automatic skin disease classification apparatus according to an embodiment of the present invention.

8 is an exemplary diagram of a gray degree co-occurrence matrix (GLCM) converted from an image according to an embodiment of the present invention.

9 is an exemplary diagram of a gray scale continuous length matrix (GLRLM) converted from an image according to an embodiment of the present invention.

10 is an exemplary diagram of melanoma images (a) and nevus images (b) used in machine learning according to an embodiment of the present invention.

11 is a diagram showing skin disease classification accuracy performance of the automatic skin disease classification method according to an embodiment of the present invention.

Other advantages and features of the present invention, and methods of achieving them will become apparent with reference to the following embodiments in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, and the present invention is only defined by the scope of the claims. If not defined, all terms used herein (including technical or scientific terms) have the same meaning as commonly accepted by universal techniques in the prior art to which this invention belongs. General descriptions of known configurations may be omitted so as not to obscure the subject matter of the present invention. In the drawings of the present invention, the same reference numerals are used for the same or corresponding configurations. In order to help the understanding of the present invention, some of the components in the drawings may be somewhat exaggerated or reduced.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise", "have" or "include" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification. Or any other feature or number, step, operation, component, part, or combination thereof.

As used throughout the present specification, '~ part' is a unit for processing at least one function or operation, and may mean, for example, a hardware component such as software, FPGA, or ASIC. However, '~' is not meant to be limited to software or hardware. '~ Portion' may be configured to be in an addressable storage medium or may be configured to play one or more processors.

As an example, '~' means components such as software components, object-oriented software components, class components, and task components, and processes, functions, properties, procedures, and subs. Routines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided by the component and the '~' may be performed separately by the plurality of components and the '~', or may be integrated with other additional components.

In the automatic skin disease classification method according to an embodiment of the present invention, features for skin diseases are extracted by image processing, and based on the extracted features, an automatic classification of skin diseases such as melanoma is performed through an algorithm of a machine learning technique. can do. 1 is a schematic flowchart of a method for automatically classifying skin diseases according to an exemplary embodiment of the present invention. Referring to FIG. 1, first, skin lesions are photographed by dermoscopy to obtain a dermoscopy image (S1), and after pretreatment of the dermis image (S2), a region of interest (ROI) is obtained. Extract (S3). Then, image features related to skin diseases such as melanoma are extracted from the region of interest of the image (S4 to S5).

Image features include pixel features extracted by first-order statistical analysis using information of the pixel itself, and gray-level co-occurrence matrix (GLCM) and gray level run length matrix (GLRLM) converted from an image into a matrix form. It may include GLCM features and GLRLM features extracted from. The extracted features (pixel features, GLCM features, and GLRLM features) may be applied to machine learning discrimination algorithms such as a Support Vector Machine and used to classify skin diseases. According to an embodiment of the present invention, in particular, it is possible to accurately classify nevus and melanoma correctly.

2 is a block diagram of an automatic skin disease classification apparatus according to an embodiment of the present invention. 2, the automatic skin disease classification apparatus 100 according to an embodiment of the present invention, the control unit 110, the learning unit 120, the image acquisition unit 130, the image preprocessor 140, skin disease classification The unit 150 and the storage unit 160 may be included. The controller 110 includes at least one processor and controls the learner 120, the image acquirer 130, the image preprocessor 140, the skin disease classifier 150, and the storage 160. Execute the function (program) for automatic disease classification.

3 is a flowchart of a method for automatically classifying skin diseases according to an exemplary embodiment of the present invention. Referring to FIGS. 2 and 3, the learner 120 may learn features of skin disease using learning image data including melanoma images and nevus images (S10). In an embodiment, the learner 120 may extract features of melanoma using an image of a recognized skin disease database and generate a machine learning model for classifying melanoma and nevus. The learner 120 may calculate a histogram from pixel values of an image of a skin disease database, and extract pixel features by statistically analyzing the histogram. In addition, the learner 120 may convert the images of the skin disease database into GLCM and GLRLM, respectively, and then extract GLCM features and GLRLM features from GLCM and GLRLM. The learner 120 learns the extracted pixel features, GLCM features, and GLRLM features to generate a machine learning model. The machine learning model generated by the learner 120 may be stored in the storage 160 to be used for classification of skin diseases.

The image acquirer 130 may acquire an image by photographing a skin lesion part (S20). In an embodiment, the image acquirer 130 may include dermoscopy for acquiring a dermoscopic image by photographing a skin lesion suspected of melanoma. The image acquired by the image acquirer 130 may be stored in the storage 160.

In the dermoscopic image, the focus of the image and the location of the disease may be different within the image for each specialist. In addition, due to the skin condition and hair of the patient, various noises except the disease are generated in the image. This noise can cause errors in extracting features from the image. In order to prevent this problem, the image preprocessing unit 140 converts the dermal image to a binary image, removes noise such as hair and keratin (S30), and a region of interest including only skin lesions in the dermal image. Can be extracted. The image preprocessed by the image preprocessor 140 may be stored in the storage 160.

4 is a detailed flowchart of step S30 of FIG. 3. 2 and 4, the image preprocessor 140 converts the dermoscopic image into a binary image (S32). In an embodiment, the Otsu technique may extract a set of objects having similar values from the histogram of pixel values of the dermal image, thereby creating a threshold value for maximizing variance between divided regions. The total variance may be expressed as the sum of the variance in the class and the variance between the classes, and may be expressed as in Equations (1) to (3) below.

[Formulas (1) to (3)]

In Equations (1) to (3), σ _ω ² is the variance in the class, σ _c ² is the variance between the classes, ω _i is the weight of the probability that the pixel is included in the class i, and μ is the average value of the class. Binary images may include noise, which is unnecessary information that reduces the accuracy of skin disease classification. The noise is mainly caused by skin conditions such as hair or keratin of the patient. This noise causes errors in extracting the features of the skin lesions and causes problems of deterioration of accuracy.

To compensate for this problem, the image preprocessor 140 removes noise based on the binary image obtained in the preprocessing process. Since the noise information is characterized in that the size information is relatively smaller than the skin disease in the image, the image preprocessing unit 140 removes the noise by using the size of the noise, and the region of interest based on the image from which the noise is removed Can be extracted.

5 is an exemplary diagram of size distribution diagrams of objects extracted from a dermoscopic image. Typically, the size of hair, keratin, etc. is smaller than the size of the skin component (e.g., melanoma), so that small-sized objects whose size difference from the skin lesion exceeds the set value are obtained. By removing from, it is possible to remove noise such as hair and dead skin cells. Skin lesions contain the largest number of pixels in the image, and the noise has a smaller component than the lesion.

The image preprocessor 140 may compare the sizes of the components (objects) based on a binary image obtained in the preprocessing process to remove noise such as hair, normal skin, and keratin other than the skin lesion. That is, the image preprocessor 140 may detect objects (eg, hair, keratin, etc.) having a smaller size than the skin lesion in the binary image (S34). The image preprocessor 140 removes noise including hair and skin keratin based on the size difference between the skin lesion and the object (S36).

6 is an exemplary view of a process of preprocessing a dermoscopic image according to an embodiment of the present invention. 6A shows a binary image converted from a dermal image, (B) a noise-removed binary image, (c) a skin lesion boundary, and (d) shows a region of interest extraction. In Figure 6 (b), it can be confirmed that the noise caused by hair or keratin, etc. is effectively removed.

When the noise is removed from the binary image, the image preprocessor 140 extracts information on the skin lesion area by using the image from which the noise is removed, and proceeds with image reconstruction to minimize information on normal skin based on the extracted information. do. In an embodiment, the image preprocessor 140 may extract the ROI including only the skin lesion by converting the pixel value of the normal skin to 0 in the dermal image (S38).

In an embodiment, the image preprocessor 140 may extract a boundary line for the skin lesion area of the image and calculate a size of the skin lesion area based on the extracted boundary line. The image preprocessor 140 removes the information on the normal skin from the lesion area based on the calculated size of the skin lesion area, and removes the skin lesion area to prevent the feature from being extracted from the area except the skin lesion area. The region of interest may be extracted by converting the pixel value of the region (normal skin) to 0, as shown in FIG. Accordingly, it is possible to minimize the error of features caused by normal skin and noise.

In the related art, a specialist diagnoses a disease by the ABCD method by visually confirming the dermal image based on clinical experience. Such a subjective diagnosis has a problem that may cause an error. To solve this problem, an embodiment of the present invention utilizes all of the pixel information in the dermal image, generates two transformation matrices based on the similarity of pixel clusters, and classifies skin diseases by extracting features from the transformation matrices. do.

7 is a block diagram of a skin disease classification unit constituting an automatic skin disease classification apparatus according to an embodiment of the present invention. 2, 3, and 7, the skin disease classifying unit 150 may use two transformation matrices GLCM and GLRLM generated from brightness values, histograms, and image information of pixels of the dermal image. In the region of interest, various features related to skin disease (pixel features, GLCM features, and GLRLM features) are extracted (S40 ˜ S90), and the extracted features are machine learning models learned by the learning unit 120 (eg, For example, it may be applied to a support vector machine model) to classify skin diseases such as melanoma (S100). In an embodiment, the skin disease classifier 150 may include a histogram calculator 151, a pixel feature extractor 152, a GLCM converter 153, a GLCM feature extractor 154, a GLRLM converter 155, and a GLRLM. The feature extractor 156 and the classifier 157 may be included.

The histogram calculator 151 calculates a histogram of brightness values of pixels in the ROI of the image of the skin lesion (S40). The pixel feature extractor 152 extracts pixel features by first performing statistical analysis on a histogram of brightness values of pixels (S50). Pixel features using only pixel information are features related to histograms using pixel-specific information and frequency of pixel information, and may be used as important features representing overall information of an image.

In an embodiment, the pixel feature extractor 152 includes a pixel including an average, standard deviation, skew, kutosis, entropy, and root mean square of pixel values of an image. Features can be extracted. Equations (4) to (7) are equations of pixel features used for feature extraction for skin lesions.

[Formulas (4) to (7)]

In Equations (4) to (7), X is pixel information (pixel value) in the region of interest, P is a histogram for pixels in the region of interest, N is the number of pixels, Ent is the entropy of the histogram, and Kur is the keratos of the pixel values. , RMS is the root mean square of the pixel values, and STD is the standard deviation of the pixel values.

Entropy in Equation (4) is a measure of disorder, which is a feature of the frequency of histograms of pixels in an image. Kurtosis of Equation (5) represents a measure of probability distribution indicating how much the distribution of pixel values in an image is distributed at a specific value. The root mean square and standard deviation of Eqs. (6) and (7) can be used as a statistical measure of the magnitude of change in pixel values and as a pixel feature related to the scatter of values in the matrix.

In order to extract a feature related to pixel continuity, the GLCM converter 153 converts the pixel values in the ROI of the image into a gray-level co-occurrence matrix (GLCM) representing the adjacency between the pixel values. The conversion is made (S60). 8 is an exemplary diagram of a gray degree co-occurrence matrix (GLCM) converted from an image according to an embodiment of the present invention.

GLCM represents the frequency of pixel values of adjacent pixels, and is a matrix having an N × N size (N is the total number of pixel values). The GLCM conversion unit 153 generates an image of (X, Y) pixels whenever an adjacent pixel group corresponding to the X th (X is an integer less than or equal to N) pixel value and the Y th (Y is an integer less than or equal to N) pixel value is found. You can create GLCM by increasing the value by 1. In the example of FIG. 8, since the pixels having the pixel value '1' have a continuous frequency of 1, the component of the first row and the first column of the GLCM is 1, and the pixel having the pixel value '2' (the second row of the GLCM) and the pixel value Since pixels having '1' (the first column of GLCM) have a continuous frequency of 2, the components of the second row and the first column of GLCM become two. By using the converted GLCM, it is possible to analyze the continuity of one pixel and an adjacent pixel in the image in the ROI.

The GLCM feature extractor 154 extracts the GLCM features based on pixel values of the gray degree co-generation matrix GLCM (S70). In an embodiment, the GLCM feature extractor 154 may include auto-correlation, contrast, correlation and analogy of pixel values of a gray-level co-occurrence matrix (GLCM). GLCM features can be extracted including dissimilarity, energy and homogeneity. In an embodiment, the GLCM feature extractor 154 may extract GLCM features from GLCM according to Equations (8) to (10) below.

[Formulas (8) to (10)]

In Equations (8) to (10), Cont is the contrast of pixel values of GLCM, Corr is the correlation of GLCM, Homo is the homogeneity of GLCM, i and j are the positions of the matrix, and P (i, j) is the GLCM Pixel value, N _g is the pixel value of the image before conversion to GLCM, μ _x is the mean value of p _x , P _x (i) is the probability for the row in GLCM, μ _y is the mean value of p _y , and P _y (i) Is the probability of heat in GLCM, σ _x is the standard deviation of p _x , and σ _y is the standard deviation of p _y .

Contrast of Equation (8) is used as a feature of the measure of the contrast of the pixels in the image. The correlation of Equation 9 is used as a feature of the measure of how similar the pixel values in the image are to each other. The homogeneity of equation (10) is used as a feature of the measure of how similar pixel values are distributed in the pixels of an image.

The GLRLM converter 155 converts the pixel values of the image into a gray level run length matrix (GLRLM) representing a continuous length of the pixel values (S80). 9 is an exemplary diagram of a gray scale continuous length matrix (GLRLM) converted from an image according to an embodiment of the present invention. The GLRLM is a matrix of how long pixels having a specific brightness value are continuously maintained at the same value (grouping). The GLRLM may be generated by accumulating the frequency of successive pixel values in an image in the ROI. The rows of GLRLM are pixel values, and the columns are the number of successive pixel values. The GLRLM converter 153 may generate the GLRLM by incrementing the value of the (x, y) pixel by 1 each time the x-th (x is an integer less than or equal to N) pixel appears consecutively y times. In the example of FIG. 9, since the frequency of the pixel having the pixel value '1' (one row of GLRLMs) appears twice in succession (two columns of GLRLMs), the one-row, two-column component value is one, and the pixel value is one. Since the frequency of the pixel having '4' (four rows of GLRLMs) appears three times consecutively (three columns of GLRLMs), the four-row, three-column component value is one.

The GLRLM feature extractor 156 extracts GLRLM features based on the pixel values of the gray continuous length matrix (S90). In an exemplary embodiment, the GLRLM feature extractor 156 may include short run emphasis (SRE), long run emphasis (LRE), gray level non-uniformity (GLNU), and RP based on pixel values of a gray scale continuous length matrix (GLRLM). GLRLM features can be extracted including Run Percentage, Run Length Non-Uniformity (RLNU), and High Gray Level Run Emphasis (HGLRE).

The classifier 157 classifies skin diseases by applying pixel features, GLCM features, and GLRLM features to a machine learning model (S100). In an embodiment, the machine learning model can include a support vector machine (SVM) model. According to an embodiment of the present invention, by extracting the characteristics of the objective skin diseases that are difficult to determine with the naked eye and using them for classification of skin diseases, it is possible to classify skin diseases such as melanoma and nevus with high accuracy, and exclude subjective judgment. And can objectively classify skin diseases.

In order to evaluate the skin disease classification performance and effectiveness of the automatic skin disease classification method according to an embodiment of the present invention, learning and testing 43 and 76 images of authorized melanoma data of DerMIS and Dermquest, and MATLAB R2017a ( The experiment was performed in MathWorks) environment and the skin classification accuracy was compared with the conventional method using the same learning data and test data. For SVM learning, after preprocessing the training data and extracting the region of interest, the pixel, GLCM, and GLRLM transformations were performed using the image of the region of interest, and feature values were extracted from the transformed matrices. 10 is an exemplary diagram of melanoma images (a) and nevus images (b) used in machine learning according to an embodiment of the present invention.

Table 1 shows the feature values (Pixel features, GLCM features, GLRLM features) for melanoma and Nevus used in the training data. From Table 1, it can be seen that differences that were not visible to the naked eye in pixel information due to the characteristics of melanoma and nevus appear in feature values extracted from the transformation matrices GLCM and GLRLM. Skin characteristic classification was performed through the SVM classifier using any test image using these feature values. FIG. 11 is a diagram illustrating skin disease classification accuracy performance of an automatic skin disease classification method according to an exemplary embodiment of the present invention. In FIG. 11, Alpha represents classification accuracy when melanoma and nevus are classified using only ABCD method, Beta represents classification accuracy using ABCD method and pixel-based feature, and Theta represents ABCD method, pixel feature, and GLCM feature. Is the classification accuracy when the ABCD method, pixel features, GLCM features, and GLRLM features are applied.

According to an embodiment of the present invention, melanoma may be applied by applying micromatrix information (pixel features) that are difficult to visually identify and transformation matrixes based on similarity of adjacent pixels and GLCM / GLRLM features extracted from transformation matrices. It can be seen that the classification accuracy is improved. The method for automatically classifying skin diseases according to an embodiment of the present invention may be used for quantitative and objective analysis and diagnosis in determining skin diseases, and may be particularly useful for an early detection system for skin diseases such as melanoma. In addition, the automatic skin disease classification method according to an embodiment of the present invention can be utilized to effectively provide information about the disease before the invasive diagnosis such as biopsy for other skin diseases other than melanoma.

The method for automatically classifying skin diseases according to an exemplary embodiment of the present invention may be implemented by, for example, a computer executable program, and may be implemented in a general-purpose digital computer operating the program using a computer readable recording medium. . The computer-readable recording medium may be volatile memory such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), Nonvolatile memory, such as electrically erasable and programmable ROM (EEPROM), flash memory device, phase-change RAM (PRAM), magnetic RAM (MRAM), resistive RAM (RRAM), ferroelectric RAM (FRAM), floppy disk, hard disk, or Optical reading media may be, for example, but not limited to, a storage medium in the form of CD-ROM, DVD, and the like.

The above embodiments are presented to aid the understanding of the present invention, and do not limit the scope of the present invention, from which it should be understood that various modifications are within the scope of the present invention. The technical protection scope of the present invention should be defined by the technical spirit of the claims, and the technical protection scope of the present invention is not limited to the literary description of the claims per se, but the invention is in the range of substantially equal technical values. It should be understood that it extends to.

Claims (15)

Calculating a histogram of brightness values of pixels in an image of a skin lesion, and extracting pixel features by statistically analyzing the histogram; Converting pixel values of the image into a gray-level co-occurrence matrix (GLCM) representing adjacency between the pixel values; Extracting a GLCM feature based on pixel values of the gray degree co-occurrence matrix; Converting pixel values of the image into a gray level run length matrix (GLRLM) representing a continuous length of the pixel values; Extracting a GLRLM feature based on pixel values of the gray continuous length matrix; And And classifying the skin disease by applying the pixel feature, the GLCM feature, and the GLRLM feature to a machine learning model. The method of claim 1, Wherein the pixel characteristic comprises an average, standard deviation, skew, kottosis, entropy, and root mean square of pixel values of the image. The method of claim 1, The GLCM feature is characterized by auto-correlation, contrast, correlation, dissimilarity, energy, and homogeneity of pixel values of the grayscale co-occurrence matrix. Skin disease automatic classification method comprising a. The method of claim 1, The GLRLM features include short run emphasis (SRE), long run emphasis (LRE), gray level non-uniformity (GLNU), run percentage (RPN), and RLNU (run) extracted based on pixel values of the gray-scale continuous length matrix. Automatic classification of skin diseases including Length Non-Uniformity (HGLRE) and High Gray Level Run Emphasis (HGLRE). The method of claim 1, The machine learning model includes a support vector machine (SVM) model for automatically classifying skin diseases. The method of claim 1, Capturing the skin lesion using dermoscopy to obtain a dermoscopic image; Converting the dermal image into a binary image and detecting an object of a smaller size than the skin lesion in the binary image; Removing noise including hair and skin keratin based on the size difference between the skin lesion and the object; And And extracting a region of interest including only skin lesions by converting pixel values of normal skin to 0 in the dermal image. The method of claim 1, Generating the machine learning model by learning the pixel feature, the GLCM feature, and the GLRLM feature based on training image data including melanoma images and nevus images, The classifying the skin disease may include classifying the skin lesion into any one of melanoma and nevus based on the learned machine learning model. A computer-readable recording medium having recorded thereon a program for executing the method for automatically classifying skin diseases. An apparatus for automatically classifying skin diseases including at least one processor, The at least one processor, Calculating a histogram of brightness values of pixels in an image of a skin lesion, and extracting pixel characteristics by statistically analyzing the histogram; Converting pixel values of the image into a Gray-Level Co-occurrence Matrix (GLCM) representing adjacency between the pixel values; Extract a GLCM feature based on pixel values of the gray degree co-occurrence matrix; Converting pixel values of the image into a gray level run length matrix (GLRLM) representing a continuous length of the pixel values; Extract a GLRLM feature based on pixel values of the gray continuous length matrix; And And apply the pixel feature, the GLCM feature, and the GLRLM feature to a machine learning model to classify skin diseases. The method of claim 9, And the pixel characteristic comprises an average, standard deviation, skew, kottosis, entropy, and root mean square of pixel values of the image. The method of claim 9, The GLCM feature is characterized by auto-correlation, contrast, correlation, dissimilarity, energy, and homogeneity of pixel values of the grayscale co-occurrence matrix. Skin disease automatic classification device comprising a. The method of claim 9, The GLRLM features include short run emphasis (SRE), long run emphasis (LRE), gray level non-uniformity (GLNU), run percentage (RPN), and RLNU (run) extracted based on pixel values of the gray-scale continuous length matrix. Automatic device for classifying skin diseases including Length Non-Uniformity (HGLRE) and High Gray Level Run Emphasis (HGLRE). The method of claim 9, The machine learning model is a skin disease automatic classification device comprising a support vector machine (SVM) model. The method of claim 9, The at least one processor, Converting the dermoscopic image of the skin lesion to a binary image by dermoscopy; Detecting an object of a smaller size than the skin lesion in the binary image; Removing noise including hair and dead skin cells based on the size difference between the skin lesion and the object; And And automatically extracting an ROI including only skin lesions by converting pixel values of normal skin to 0 in the dermal image. The method of claim 9, The at least one processor, Generate the machine learning model by learning the pixel feature, the GLCM feature, and the GLRLM feature using training image data including melanoma images and nevus images; And And automatically classify the skin lesion into one of melanoma and nevus based on the machine learning model trained using the learning image data.

Images