KR20200139606A - Cervical cancer diagnosis system - Google Patents

Abstract

The present invention relates to an automatic cervical cancer diagnosis system which classifies cervical data according to accurate criteria, and performs machine learning to automatically diagnose cervical cancer. The system includes: a learning data generation unit for generating new classification criterion-specific learning data by classifying unclassified photographed image data for the cervix transmitted from an external device or storage unit according to a combination of multi-level classification criteria in a learning mode; a photographed image preprocessing unit for preprocessing the cervical photographed image; a cervical cancer diagnosis unit that includes a machine learning model for cervical cancer to learn the characteristics of the learning data generated based on the classification criteria in the learning mode, wherein the machine learning model generates diagnosis information on an occurrence of the cervical cancer with respect to the preprocessed cervical photographed image in a diagnostic mode; a screen display control unit for displaying and outputting a user interface screen for displaying the diagnosis information and inputting evaluation information through reading; a re-learning data generation unit for requesting re-learning of the machine learning model by extracting information necessary for re-learning among the evaluation information input through the user interface screen; and a diagnosis and evaluation information storage unit for storing diagnosis information regarding the occurrence of cervical cancer and evaluation information inputted through the user interface screen.

Description

Automatic cervical cancer diagnosis system {CERVICAL CANCER DIAGNOSIS SYSTEM}

The present invention relates to an automatic diagnosis system for cervical cancer, and in particular, automatic diagnosis of cervical cancer for automatically diagnosing cervical cancer based on the machine learning by classifying cervical data required for automatic diagnosis of cervical cancer according to an accurate criterion. It's about the system.

Cervical cancer ranks first among Korean women's most feared cancers, because it can affect pregnancy and childbirth due to hysterectomy, and there is a fear of loss as a woman.

According to statistics in 2013, the number of cervical cancer patients in Korea was 26,207, ranking fourth among female cancers (data from the Ministry of Health and Welfare). In addition, cervical cancer is one of the seven major cancers recommended for screening in Korea, and as it was included in the national cancer screening project in 1999, the rate of early diagnosis is increasing. In recent years, cervical epithelial cancer (precancerous stage), called cervix'phase 0'cancer, is also on the rise, and women with sexual experience are advised to undergo an annual examination.

Looking at the current status of the market for the screening, first of all, the rate of cervical epithelial cancer in young women is increasing, and since 2016, screening targets have been lowered from 30 to 20 years old. In particular, unlike other cancers, health insurance benefits are applied to 300% of the screening costs related to cervical cytology. However, as the false negative rate (i.e., false diagnosis rate) of screenings reaches a maximum of 55%, it is recommended to perform a cervical enlargement examination as a supplement. As of 2013, the global cervical cancer screening market was about 6.86 trillion won. In terms of scale, of these, cervical magnification scans account for 30%, reaching about 2 trillion won.

FIG. 1 is a conceptual diagram schematically showing a method of a cervical cytology and an enlarged cervical examination, which were conventionally performed to diagnose cervical cancer. Referring to the lower part of FIG. 1, a predetermined amount from outside the vagina of a female subject When a photographed image of the cervix is acquired through an imaging device (for example, a cervix magnifying glass shown in FIG. 1), it is possible to reduce the false diagnosis rate for cervical cancer by analyzing the image and using the result.

However, when using a conventional cervical magnifying glass, the medical staff checks the onset of cervical cancer with respect to the image of the cervix in the light of education and experience. This method is often repetitive and ambiguous, so even experienced doctors take time. It will take a long time and the accuracy can also decrease.

In order to overcome these shortcomings, cervical cancer that acquires a photographed image of the cervix and generates and provides analysis information on whether the subject has cervical cancer based on a machine learning model for cervical cancer from the acquired cervical image. Devices for determining whether an outbreak has occurred are also introduced.

The main factor for evaluating the performance of these judgment devices is that the images to be used for learning must be accurately classified and organized to proceed with learning. If such data classification is not done accurately and clearly, the accuracy of analysis on cervical cancer incidence will inevitably decrease. Usually, cervical cancer colposcopy images appear in various forms different from general medical images depending on the shooting environment and the photographer. Therefore, in devices for determining whether cervical cancer has occurred, it is necessary to classify the images to be used for learning according to more clear and strict criteria to proceed with learning.

In addition, even if the learning data is classified according to clear and strict criteria and machine learning is performed, a misdiagnosis may occur in the diagnosis (analysis) result. In this case, an evaluation by a specialist can be made to prevent cervical cancer based on machine learning. There is an urgent need to introduce a system capable of enhancing the reliability of analysis devices for automatic diagnosis or assisting diagnosis performance.

Korean Registered Patent Publication No. 10-0850347 Republic of Korea Patent Publication No. 10-2016-0047720

Accordingly, the present invention is an invention invented in accordance with the above-described necessity, and the main object of the present invention is to classify cervical data for learning necessary for automatic diagnosis of cervical cancer according to an accurate criterion, machine learning, and cervical cancer based on the machine learning. It is to provide an automatic diagnosis system for cervical cancer that automatically diagnoses

Furthermore, another object of the present invention is to induce an accurate diagnosis of the cervix by preventing excessive learning about only certain types of cervical images or not learning about certain types of images. It is to provide an automatic diagnosis system for cervical cancer.

In addition, another object of the present invention is to provide convenience of reading by providing various information automatically analyzed based on a machine learning model to a reader (expert), and input information evaluated by reading intention can be used for real-time relearning. It is to provide an automatic diagnosis system for cervical cancer that allows the performance of the machine learning model to be continuously improved by configuring the user interface screen so that it can be used.

Cervical cancer automatic diagnosis system according to an embodiment of the present invention for achieving the above object,

A learning data generator for classifying unclassified photographed image data for the cervix transmitted from an external device or a storage unit in the learning mode according to a combination of multi-level classification criteria and generating learning data for each classification criteria;

A photographed image preprocessor for preprocessing the cervix photographed image;

In the learning mode, a machine learning model for cervical cancer that learns characteristics of the learning data generated by the classification criteria is included, and the machine learning model determines whether the onset of cervical cancer is detected in the preprocessed cervical image in the diagnosis mode. A cervical cancer diagnosis unit for generating diagnostic information about the;

A screen display control unit for displaying and outputting a user interface screen for displaying the diagnosis information and inputting evaluation information by reading intention;

A relearning data generation unit for requesting relearning of the machine learning model by extracting information necessary for relearning from among the evaluation information input through the user interface screen;

And a diagnosis and evaluation information storage unit for storing diagnosis information on whether the cervical cancer has occurred and evaluation information input through the user interface screen.

In the above-described system configuration, the learning data generation unit further generates additional learning data by applying mirroring or cropping to the learning data for each classification criteria in order to adjust the numerical balance of the learning data for each classification criteria. And

Furthermore, the learning data generation unit,

The first-stage classification standard based on color, the second-stage classification standard based on the size of the cervix in the captured image data, and a three-step classification standard based on the combination of colors and shapes in the cervix image data. Another feature is that classification is performed using a combination of at least two classification criteria among the four classification criteria, which use exposure and focus as classification criteria.

In some cases, the learning data generation unit,

According to the first-stage classification criteria based on color, the unclassified photographed image data for the cervix is first classified, and the size of the cervix in the first classified unclassified photographed image data is used as the classification criteria. According to the second classification, the unclassified photographed image data for the cervix is classified in a third order according to a three-stage classification criterion based on a combination of colors and shapes in the second classified unclassified photographed image data.

Meanwhile, in the automatic diagnosis system for cervical cancer described above, the first-stage classification criterion includes color values for identifying each of an acetic acid reaction image, a Lugol solution reaction image, a green filter image, and a general image as a classification reference value,

The three-stage classification criterion includes a combination of color values and shapes for identifying any one or more of blood, mucus, loops, colposcopy, treatment traces, and surgical instruments in the cervical image data as the classification criterion value. It is characterized.

Furthermore, the cervical cancer diagnosis unit includes classification information regarding negative, atypical, benign, and malignant cervical cancer for the cervical image of the subject based on the machine learning model, accuracy information of the classification information, and negative information. It is characterized by generating diagnostic information including positive determination information and morphological finding information,

The screen display control unit includes a display area of all or part of the cervix photographed image of the subject, a history information display area of another image previously captured for the same subject, an exposure area of the subject information, an input area of negative and positive determination information, and a shape. The user interface screen displays a screen divided into an academic finding information input area, a technical defect information input area for the quality of a captured image, the diagnostic information output area derived based on the machine learning model, and a reading finding input area. It is characterized by being.

According to the above-described technical problem solving means, the automatic cervical cancer diagnosis system according to an embodiment of the present invention classifies unclassified photographed image data for the cervix according to a combination of multi-level classification criteria to train an artificial intelligence network such as a machine learning model. Then, more accurate lesion diagnosis can be made because image data having more diverse characteristics can be learned than simply learning by using the classified data using classification criteria having only one characteristic (color, shape, brightness, etc.). There is an advantage to be able to.

In addition, the present invention further generates additional learning data to adjust the numerical balance of the learning data for each classification criterion and uses it for learning, so that excessive learning is performed only on certain types of cervical (cancer) image data. On the contrary, by preventing the phenomenon that learning does not proceed with respect to a specific type (or type) of an image, it provides the effect of accurately diagnosing the presence or absence of a lesion on the cervix.

Furthermore, the present invention automatically diagnoses whether or not there is a lesion for cervical cancer based on a machine learning model, and displays the result to the outside through a user interface screen, and induces the reader to input evaluation information on the user interface screen. By extracting some of the evaluation information and using it for re-learning of the machine learning model, there is an advantage that the performance of the machine learning model can be continuously upgraded.

1 is a conceptual diagram schematically showing a method of a cervical cell examination and an enlarged cervical examination, which were conventionally performed to diagnose cervical cancer.
2 is an exemplary configuration diagram of an automatic diagnosis system for cervical cancer according to an embodiment of the present invention.
3 is a view for explaining a method of classifying cervical learning data according to an embodiment of the present invention.
4 is a view for explaining in more detail multi-level classification criteria for generating cervical learning data according to an embodiment of the present invention.
5 is a view for explaining an automatic diagnosis process for cervical cancer according to an embodiment of the present invention.
6A to 6E are exemplary views of a user interface (UI) screen provided in an automatic diagnosis process for cervical cancer according to an embodiment of the present invention.
7 is an exemplary view of malignant atypical vessels.

DETAILED DESCRIPTION OF THE INVENTION The detailed description of the present invention to be described below refers to the accompanying drawings, which illustrate specific embodiments in which the present invention may be practiced in order to clarify the objects, technical solutions and advantages of the present invention. These embodiments are described in detail sufficient to enable a person skilled in the art to practice the present invention.

And throughout the detailed description and claims of the present invention,'learning' is a term referring to performing machine learning according to a procedure, and that it is not intended to refer to a mental action such as human educational activity. Will be able to understand. Also throughout the detailed description and claims of the present invention, the word'comprise' and variations thereof are not intended to exclude other technical features, additions, components or steps. Other objects, advantages, and features of the present invention to those skilled in the art will appear, partly from the present disclosure, and partly from the practice of the present invention. The examples and drawings below are provided by way of example and are not intended to limit the invention. Moreover, the present invention covers all possible combinations of the embodiments indicated herein. It is to be understood that the various embodiments of the present invention are different from each other, but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the present invention in relation to one embodiment. In addition, it is to be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the present invention. Accordingly, the detailed description to be described below is not intended to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all scopes equivalent to those claimed by the claims. Like reference numerals in the drawings refer to the same or similar functions over several aspects.

Unless otherwise indicated in this specification or clearly contradicting the context, items referred to in the singular encompass the plural unless otherwise required by that context. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function may obscure the subject matter of the present invention, a detailed description thereof will be omitted.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to allow those skilled in the art to easily implement the present invention.

For reference, a convolutional neural network (CNN) model, which is one of the machine learning models used in the present invention, can be briefly described in the form of stacking artificial neural networks in multiple layers. In other words, it is expressed as a deep neural network (deep neural network) in the sense of a deep-structured network, and automatically learns the features of each image by learning a large amount of data in a structure consisting of a multi-layered network. Through this, the network is trained in a way to minimize the error of the objective function. At first glance, this can be expressed as a connection between neurons in the human brain, and accordingly, it has established itself as a representative of artificial intelligence. In particular, CNN is a model suitable for classification of two-dimensional images such as images, and reduces the size of the convolution layer and feature map that creates a feature map using a plurality of filters for each region of the image. By repeating a pooling layer (sub-sampling layer) that allows you to extract features that are invariant with changes in position or rotation, from low-level features such as points, lines, and planes to complex and meaningful high-level features. It is possible to extract features of various levels, and finally, if the features extracted through the fully-connected layer are used as input values of the existing model, there is an advantage of constructing a classification model with higher accuracy. .

2 is a diagram illustrating the configuration of an automatic diagnosis system for cervical cancer according to an embodiment of the present invention. The automatic diagnosis system 200 for cervical cancer according to an embodiment of the present invention is a single computer that is connected to the imaging device 100, the storage medium 300, the communication unit 400, and the display unit 500 as shown. It may be a system or a server computer system, and may be implemented as a collection of executable code data (in the form of an application program) installed in the memory of the computer system and executable.

Referring to FIG. 2, the cervical cancer automatic diagnosis system 200 according to an exemplary embodiment of the present invention may directly or indirectly communicate with a computing device (not shown) located at a remote location through the communication unit 400. Of course, the reading intention may directly input (using a keyboard or mouse) evaluation information to be described later through the user interface screen displayed on the display unit 500.

The imaging apparatus 100 acquires a cervix image of the subject in the diagnosis mode and transmits it to the automatic cervical cancer diagnosis system 200 to be described later, and the storage medium 300 is the cervix to be learned by a machine learning model to be described later. It stores unclassified photographed image data for.

Hereinafter, the configuration of the automatic cervical cancer diagnosis system 200 will be further described with reference to FIG. 2, the automatic cervical cancer diagnosis system 200 according to an embodiment of the present invention,

In the learning mode, a learning data generation unit 210 that classifies unclassified photographed image data for the cervix transmitted from an external device or a storage unit (storage medium, 300) according to a combination of multi-level classification criteria and generates learning data according to new classification criteria. Wow,

A photographed image preprocessing unit 220 for pre-processing (image quality improvement, blurring, noise processing) of the cervix photographed image, and

In the learning mode, a machine learning model for cervical cancer that learns characteristics of the learning data generated by the classification criteria is included, and the machine learning model determines whether the onset of cervical cancer is detected in the preprocessed cervical image in the diagnosis mode. Cervical cancer diagnosis unit 260 for generating diagnostic information,

A screen display control unit 250 for displaying and outputting a user interface screen for displaying the diagnosis information and inputting evaluation information by reading intention;

A relearning data generation unit 230 for requesting relearning of the machine learning model by extracting information necessary for relearning from among the evaluation information input through the user interface screen,

And a diagnosis and evaluation information storage unit 240 for storing diagnosis information on whether the cervical cancer has occurred and evaluation information input through the user interface screen.

For reference, the multi-level classification criteria, diagnostic information and evaluation information for the subject may be stored in the storage medium 300.

The learning data generation unit 210 further learns by applying a vertical or horizontal mirroring or cropping technique to the learning data for each classification criteria in order to adjust the numerical balance of the learning data for each classification criteria generated in the learning mode. By generating more data, excessive learning is not performed only on certain types of cervical (cancer) image data, or, conversely, a phenomenon in which normal learning is not performed on certain types (or types) of images is prevented from occurring.

Meanwhile, in classifying unclassified photographed image data for the cervix in the learning mode, the learning data generation unit 210 uses a first-stage classification criterion based on color as a classification criterion and a cervix size in the captured image data as a classification criterion. Classification using a combination of at least two of the second-stage classification criteria, the three-stage classification criteria based on the combination of color and shape in the cervical image data, and the four-stage classification criteria based on exposure and focus It is characterized by being.

In a specific implementation method, the learning data generation unit 210,

According to the first-stage classification criteria, unclassified photographed image data for the cervix is first classified according to the first-stage classification criteria based on color, and the size of the cervix in the first classified unclassified photographed image data is used as the classification criteria. The unclassified photographed image data for the cervix may be classified in a third order according to a three-step classification criterion based on a second classification, and a combination of colors and shapes in the second classified unclassified photographed image data as a classification standard.

In addition, the learning data generation unit 210 may classify the unclassified photographed image data for the cervix classified in the third order according to the four-step classification criteria based on exposure and focus as classification criteria. The four-stage classification criterion is applied prior to the first-stage classification criterion and may be used as learning data for filtering photographed image data that cannot be diagnosed (the lesion is not identified). For example, if the exposure is under/overexposed, the histogram is extremely skewed to one side, so it can be classified using this, and if the focus is out of focus, the edge (edge) cannot be detected or the color contrast is ambiguous. It can be classified (quaternary classification) using characteristics.

Additionally, the learning data generation unit 210 first classifies the unclassified photographed image data for the cervix according to a first-stage classification criterion based on color, and the first-stage classification criteria are acetic acid reaction images and lugol. A color value for identifying at least one of a solution reaction image, a green filter image, and a general image is included as a classification reference value.

In addition, the learning data generation unit 210 may include the size of the cervix in the first classified unclassified photographed image data, for example, the size of the cervix in the image 150%, 100%, 80%, 50%, a magnifying glass and other parts. Secondary classification is based on the size.

Furthermore, the learning data generation unit 210 unclassifies the cervix according to a three-stage classification criterion based on a combination of color and shape in the cervix image data from the unclassified photographed image data for the cervix classified as a secondary classification. The photographed image data is classified in a third order, but the three-stage classification criteria is a combination of color values and shapes to identify at least one of blood, mucus, loops, colposcopy, treatment traces, and surgical instruments within the cervical image data. Foreign substances affecting the cervix are classified by including as the classification standard value.

For example, blood mainly appears in the form of red flowing down from the center of the cervix, mucus mainly appears in the form of pale yellow flowing down from the center of the cervix, and the loop is mainly located in the center of the cervix and is generally a boomerang. The shape of the wire is clearly visible. Colposcopy and other surgical instruments appear in a different color (silver, blue,...) than the pink cervix, so using the combination of color and shape of each foreign body affects the cervix as illustrated above. Foreign substances can be classified.

In the illustrated first to fourth classification processes, each classification may be performed using CNN, a deep learning technique. In the 1st, 2nd, and 4th classification, since the features to be extracted are clear, high-accuracy classification is possible with the composition of a small layer, and in the 3rd classification, since there are many features to be extracted, it is a deep layer. It can be configured to increase accuracy.

Hereinafter, a method of classifying learning data of the automatic cervical cancer diagnosis system 200 according to an embodiment of the present invention will be described in more detail with reference to FIGS. 3 and 4.

First, FIG. 3 is a diagram illustrating a method of classifying cervical learning data according to an embodiment of the present invention, and FIG. 4 is a detailed description of a multi-level classification criterion for generating cervical learning data according to an embodiment of the present invention. It shows a drawing for explanation.

Referring to FIG. 3, first, prior to learning about unclassified photographed image data for the cervix, the storage medium 300 includes unclassified photographed image data for the cervix transmitted from the imaging device 100 or an external device. It is assumed that they are stored (step S100) or are stored.

Under this assumption, if there is an instruction to execute the learning mode from the administrator, the learning data generation unit 210 classifies each unclassified photographed image data for the cervix based on a neural network algorithm such as CNN, but the combination of a plurality of multi-level classification criteria It is classified according to and first stored (S200 step).

For example, the learning data generation unit 210 first classifies the unclassified photographed image data for the cervix first according to a first-stage classification criterion based on color as a classification criterion.

For the first classification, the learning data generation unit 210 includes color values for identifying each of an acetic acid reaction image, a Lugol solution reaction image, a green filter image, and a general image as classification reference values, thereby classifying the above four images. I can.

Specifically, the acetic acid reaction image can be distinguished from the pink cervix and vagina because white spots appear on the cervix. Since the Lugol solution reaction image has a brown or dark orange color, and the green filter image has a strong green color throughout the image, the color values representing the characteristics of each of these images are used as the classification reference value to obtain unclassified image data for the cervix. Can be classified.

When the primary classification is completed, the learning data generation unit 210 performs secondary classification according to a second-level classification criterion based on the size of the cervix in the first classified photographed image data.

The cervix is a circle the size of a 500 won coin and is usually located in the center of the image. Therefore, based on the size of the cervix in the image (150%, 100%, 80%, etc.), only the cervix is enlarged, the entire cervix is shown, the cervix is only 80% of the image, and the cervix is 50% of the image. It can be classified as a second-order image, such as an image showing only degree, and an image containing cervix + magnifying glass + external part.

Thereafter, the learning data generation unit 210 thirdly classifies foreign substances affecting the cervix according to a three-stage classification criterion based on a combination of colors and shapes in the secondly classified cervix image data.

As mentioned above, blood mainly appears in the form of red flowing down from the center of the cervix, mucus mainly appears in the form of pale yellow flowing down from the center of the cervix, and the loop is mainly located in the center of the cervix. The boomerang-shaped wire clearly appears. Colposcopy and other surgical instruments appear in a different color (silver, blue,...) than the pink cervix, so using the combination of color and shape of each foreign body affects the cervix as illustrated above. Foreign substances can be classified.

As described above, the photographed image data according to classification criteria classified according to the combination of the multi-level classification criteria may be temporarily stored in the storage medium 300 or in the memory of the system 200.

When the classification of the unclassified photographed image data is completed, the learning data generation unit 210 generates the photographed image data for the cervix classified by the classification criteria as learning data for each classification criteria and stores it in the storage medium 300 (step S300). do. In this learning data generation process, the learning data generation unit 210 may further generate additional learning data to adjust the numerical balance of the learning data for each classification criterion, and the additional learning data is generated based on the learning data for each classification criterion. It is desirable to do.

As a method to create additional learning data, vertical mirroring that changes left and right using mirroring and horizontal mirroring that changes vertically can be used, and cropping is used to crop up/down/left/right to a smaller size than the original. You can ping to generate additional training data. In addition, when mirroring and cropping are used together, it is possible to create up to 16 times more additional training data.

For reference, in the above example, training data was created by applying a combination of the first and second classification criteria, and the first and second level classification criteria, the first and third level classification criteria combination, and the second stage. And a combination of three-stage classification criteria can be applied to generate training data, and as shown in FIG. 4, unclassified photographed image data for the cervix are classified by simply applying the first to third-stage classification criteria, It could also be used as training data.

As described above, when the unclassified photographed image data for the cervix is classified according to the combination of the multi-level classification criteria and training data is generated, the machine learning model 262 for the cervix learns the characteristics of the training data generated for each classification. Verification is executed (step S400). If unclassified photographed image data for the cervix is additionally secured, the above-described learning mode is performed again.

As described above, since the automatic diagnosis system 200 for cervical cancer according to an embodiment of the present invention generates training data by classifying unclassified image data for the cervix according to a combination of multiple classification criteria, the number of classifications of the training data As the number of classifications of the training data increases and the number of classifications of the training data increases, the image data having various characteristics is learned, so that the presence or absence of a lesion can be more accurately diagnosed.

In addition, the present invention further generates additional learning data for adjusting the numerical balance of learning data for each classification criterion to be used for learning, so that excessive learning is performed only on certain types of cervical (cancer) image data. On the contrary, by preventing the phenomenon that learning does not proceed with respect to a specific type (or type) of an image, it provides the effect of accurately diagnosing the presence or absence of a lesion on the cervix.

Hereinafter, the automatic diagnosis process of cervical cancer of the cervical cancer automatic diagnosis system 200 in which the machine learning model 262 is trained using the learning data generated by classifying unclassified photographed image data for the cervix according to the combination of the above-described multi-level classification criteria. Let's explain.

5 is a diagram illustrating a process for automatically diagnosing cervical cancer according to an embodiment of the present invention, and FIGS. 6A to 6E are a user interface (UI) provided in the automatic diagnosis process for cervical cancer according to an embodiment of the present invention; user interface) screen.

Referring to FIG. 5, first, in the diagnosis mode, the automatic cervical cancer diagnosis system 200 acquires a photographed image of the cervix of a subject (step S410). The photographed image of the cervix of the subject may be obtained from the imaging apparatus 100 or may be obtained from the outside through the communication unit 400 or a storage medium. The captured image 514 of the cervix obtained as described above is displayed on the user interface screen as shown in FIG. 6A by the screen display controller 250. A user interface screen may be configured so that information on a subject, which is a patient, together with the captured image 514 may be displayed through the subject information input area 510 as illustrated in FIG. 6A.

In order to generate more accurate diagnosis information before displaying the acquired image of the cervix of the subject on the user interface screen, it is preferable that the image pre-processing unit 220 performs pre-processing on the image of the cervix. . Here, the pre-processing may include at least one of RGB-HSV conversion for robustness against illumination and noise of the captured image, image quality improvement through histogram smoothing, etc., blurring, and noise processing.

Referring back to FIG. 5, when pre-processing is completed by acquiring a photographed image of the cervix of a subject, the machine learning model 262 of the cervical cancer diagnosis unit 260 shows the onset of cervical cancer with respect to the pre-processed cervix image. Diagnosis information regarding whether or not is generated (step S420).

The machine learning model 262 includes a plurality of pre-input training information, that is, (i) learning data by classification criteria according to a combination of multi-level classification criteria, and (ii) lesions of cervical cancer exist in the plurality of training data images. Diagnosis information regarding the onset of cervical cancer is generated by using information including data indicating whether or not the lesion exists, and (iii) data indicating whether the lesion is present in any part of the image. For reference, the machine learning model 262 may be a convolutional neural network (CNN) model, and may be a model combining a CNN and a support vector machine (SVM).

In order to increase the accuracy and reliability of the machine learning model 262, a large amount of training data is required, and the greater the number of training data is, the more accurate and reliable the machine learning model can be.

Meanwhile, the diagnostic information may include classification information on whether the cervical cancer is negative, atypical, positive, or malignant. In addition, the classification information may include probability information indicating how accurate the classification is.

More specifically, the diagnostic information may include negative or positive determination information, which is information about whether the risk is high or low (low cancer risk vs. high cancer risk) if it is negative or if it is positive, and Acetowhite Epithelium, Mosaic, Morphological findings information such as erosion or ulceration, irregular surface contour, punctation, atypical vessels, discolaration, etc. may be further included.

The diagnostic information may be listed and provided to correspond to a plurality of captured images 514 by the screen display controller 250 as shown in FIG. 6B. The user interface screen shown in FIG. 6B indicates whether the onset of cervical cancer is suspected according to the subject information 520, the input time point information 522, the classification information and probability information calculated by the machine learning model 262. Information (marked'suspicious'; 524) is displayed. Furthermore, on the user interface screen shown in FIG. 6B,'Evaluation' buttons 526 corresponding to a specific captured image are shown so that subsequent steps following step S420 can be continuously performed.

Through the'Evaluation' button 526, the reader may read whether the onset of cervical cancer, which is automatically diagnosed based on the machine learning model, and input evaluation information. That is, when there is an operation of the'Evaluation' button 526, the screen display controller 250 supports multiple screens to read whether cervical cancer has occurred.

For example, as shown in Fig. 6C, by providing all or part of the captured image on the user interface screen, the reader can use the tool window to display various necessary displays while determining whether there is a medically unusual area. Yes (eg, rectangle, arrow, text input, etc.).

In addition, the screen display control unit 250 may display diagnostic information corresponding to a reading command input through the user interface screen, and support inputting evaluation information by reading intention (step S430).

The evaluation information refers to whether the provided diagnostic information is accurate, that is, information on whether the outbreak included in the diagnosis information is correct or incorrect, and whether the classification information included in the diagnosis information is correct or incorrect, and if so, what classification is correct. May include information about.

In addition, the evaluation information may include information on quality of the captured image, for example, information on technical defects of the captured image. For example, such a technical flaw is that it is difficult to accurately determine the captured image due to excessive mucus or blood in the captured image, or it is difficult to determine whether cervical cancer has occurred due to the angle of the captured image or the location of the captured part. There may be an acetic acid reaction, but the acetic acid reaction is insufficient, out of focus, overexposure, or underexposure.

6D shows a user interface screen displayed by the screen display controller 250, in which all or part of the captured image 540 is displayed, and history information of other images previously captured for the same subject ( An area where 541) is displayed, a subject information exposure area 542 as illustrated in FIG. 6A, a negative-positive determination information input area 543 into which negative-positive determination information can be input, and morphological findings information by reading intention Derived by the morphological finding information input area 544 into which is input, the technical defect information input area 545 into which information about the technical defect of the photographed image 540 can be input, and the machine learning model 262 An artificial intelligence diagnostic information output area 546 representing the diagnosed information and a user opinion input area 547 through which the reader can input the findings based on the captured image are displayed on the user interface screen, whereby automatic diagnosis information It becomes easier for the reader to evaluate whether the onset of cervical cancer corresponding to.

On the other hand, if the reader inputs evaluation information for the diagnosis information analyzed based on the machine learning model 262 and the captured image of the cervix of the subject through the user interface screen shown in FIG. 6, the diagnosis and evaluation information storage unit 240 stores diagnostic information regarding the onset of cervical cancer and evaluation information of the read intention input through the user interface screen in the storage medium 300 (step S440), or the display unit 500 at the request of an administrator or read Print it out to

On the other hand, when evaluation information is input through the user interface screen, the relearning data generation unit 230 includes information necessary for relearning among the evaluation information input through the user interface screen, for example, negative and positive determination information, and technical defect information. It is possible to request relearning of the machine learning model 262 by extracting, etc.

In addition, the above-described evaluation information may be processed and provided in the form of a medical result report. For example, a user interface provided for this is well shown in FIG. 6E, and the medical result report 550 includes whether or not cervical cancer has occurred, classification Information may be included. The medical result report may be provided through another device located far from the place where the captured image was initially acquired (ie, the place where the captured image was taken) or the place where the captured image was read.

In the above embodiment, the description was made on the assumption that the reader is located in the cervical cancer automatic diagnosis system 200 and the evaluation information is input. However, the photographed image and diagnostic information are transmitted through the communication unit 400 without any modification. It can be sent to so that the evaluation information can be entered.

According to the embodiment of the present invention described above, since the onset of cervical cancer can be automatically diagnosed based on the machine learning model 262 that has been learned in advance, the medical staff can take the cervical image acquired through the cervical magnifying glass. Compared to the previous method, which directly checks the condition of the cervix based on education and experience, it has the advantage of being able to diagnose the onset of cervical cancer more quickly and accurately.

In addition, the present invention has the advantage of performing more accurate reading because the evaluation information can be used again as data for relearning the machine learning model.

Meanwhile, in the above-described embodiment, the first-stage classification standard based on color, the second-stage classification standard based on the size of the cervix in the captured image data, and a combination of colors and shapes in the cervix image data are used as the classification criteria. The three-stage classification criteria and the four-stage classification criteria based on exposure and focus are exemplified, but the learning data generation unit 210 classifies the malignant atypical vessels pattern as shown in FIG. It is also possible to generate training data by classifying unclassified photographed image data for the cervix according to the five-stage classification criteria alone or a combination of the five-stage classification criteria and the other classification criteria described above by further adding the five-stage classification criteria.

When learning about the malignant atypical blood vessel pattern shown in FIG. 7 is performed, the cervical cancer diagnosis unit 260 may generate a diagnosis for the onset of cervical cancer indicating the malignant atypical blood vessel pattern based on the learned machine learning model. I will be able to.

Based on the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be achieved through a combination of software and hardware, or can be achieved only with hardware. The objects of the technical solution of the present invention or parts contributing to the prior art may be implemented in the form of program instructions that can be executed through various computer components and recorded in a machine-readable recording medium. The machine-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the machine-readable recording medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in the computer software field.

Examples of machine-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magnetic-optical media such as floptical disks. media), and a hardware device specially configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of the program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa. The hardware device may include a processor such as a CPU or GPU coupled with a memory such as ROM/RAM for storing program instructions and configured to execute instructions stored in the memory, and can exchange signals with an external device. It may include a communication unit. In addition, the hardware device may include a keyboard, a mouse, and other external input devices for receiving commands written by developers.

In the above, the present invention has been described by specific matters such as specific elements and limited embodiments and drawings, but this is provided only to help a more general understanding of the present invention, and the present invention is not limited to the above embodiments. , Anyone with ordinary knowledge in the technical field to which the present invention pertains can make various modifications and variations from these descriptions. Therefore, the spirit of the present invention is limited to the above-described embodiments and should not be defined, and all modifications equivalently or equivalently to the claims as well as the claims to be described later fall within the scope of the spirit of the present invention. I would say.

Claims (9)

A learning data generation unit that classifies unclassified photographed image data of the cervix transmitted from an external device or a storage unit in the learning mode according to one or a combination of multi-level classification criteria and generates learning data for each new classification criteria;
A photographed image preprocessor for preprocessing the cervix photographed image;
In the learning mode, a machine learning model for cervical cancer that learns characteristics of the learning data generated by the classification criteria is included, and the machine learning model determines whether the onset of cervical cancer is detected in the preprocessed cervical image in the diagnosis mode. A cervical cancer diagnosis unit for generating diagnostic information about the;
A screen display control unit for displaying and outputting a user interface screen for displaying the diagnostic information and inputting evaluation information by reading intention;
A relearning data generation unit for requesting relearning of the machine learning model by extracting information necessary for relearning from evaluation information input through the user interface screen;
And a diagnosis and evaluation information storage unit for storing diagnosis information on whether the cervical cancer has occurred and evaluation information input through the user interface screen. The automatic cervical cancer according to claim 1, wherein the learning data generation unit further generates additional learning data by applying mirroring or cropping to the learning data for each classification criteria in order to adjust the numerical balance of the learning data for each classification criteria. Diagnostic system. The method of claim 1, wherein the learning data generation unit,
The first-stage classification standard based on color, the second-stage classification standard based on the size of the cervix in the captured image data, and a three-step classification standard based on the combination of colors and shapes in the cervix image data. , Cervical cancer automatic diagnosis system, characterized in that the unclassified photographed image data is classified using a combination of at least two classification criteria among four classification criteria based on exposure and focus as classification criteria. The method of claim 1, wherein the learning data generation unit,
According to the first-stage classification criteria based on color, the unclassified photographed image data for the cervix is first classified, and the size of the cervix in the first classified unclassified photographed image data is used as the classification criteria. The uterus, characterized in that the second classification is performed according to the second classification, and the third classification of the unclassified photographed image data for the cervix according to a three-stage classification criterion based on a combination of colors and shapes in the second classified unclassified photographed image data as a classification standard. Automatic cervical cancer diagnosis system. The method according to claim 3 or 4, wherein the first-stage classification criterion includes color values for identifying each of an acetic acid reaction image, a Lugol solution reaction image, a green filter image, and a general image as a classification reference value,
The three-step classification criterion is characterized in that it includes a combination of color values and shapes for identifying any one or more of blood, mucus, loops, colposcopy, treatment traces, and surgical instruments in the cervical image data as the classification criterion value. Automatic cervical cancer diagnosis system. The method according to claim 1, wherein the cervical cancer diagnosis unit,
Based on the machine learning model, classification information on the negative, atypical, benign, and malignant cervical cancer of the subject's cervical image, accuracy information of the classification information, negative and positive determination information and morphological findings Automatic cervical cancer diagnosis system, characterized in that generating diagnostic information including information. The method according to claim 1, wherein the screen display control unit,
Display area of all or part of the cervix image of the subject, display area of history information of other images previously captured for the same subject, exposure area of subject information, input area of negative and positive determination information, input area of morphological findings information , A screen divided into an input area for technical defect information regarding the quality of a captured image, an output area for outputting diagnosis information derived based on the machine learning model, and an input area for reading findings is displayed as the user interface screen. Automatic cervical cancer diagnosis system. The automatic diagnosis system for cervical cancer according to claim 1, wherein the multi-level classification criterion includes malignant atypical blood vessel patterns as one classification criterion. The method of claim 1, wherein the learning data generation unit,
The first-stage classification standard based on color, the second-stage classification standard based on the size of the cervix in the captured image data, and a three-step classification standard based on the combination of colors and shapes in the cervix image data. , Unclassified photographing using a combination of at least two classification criteria of each of the four-stage classification criteria based on exposure and focus as the classification criteria, and the five-stage classification criteria based on malignant atypical vascular pattern classification criteria, or among the above 1-5 classification criteria An automatic diagnosis system for cervical cancer characterized by classifying image data.

Images