KR20230125412A - Classification method and analysis apparatus for uric acid stone using radiomics parameters of computed tomography - Google Patents

Abstract

A method for classifying urinary stones using radiomix parameters includes the steps of receiving a CT image of a subject by an analysis device, segmenting a urinary stone area in the CT image by the analysis device, and The method includes extracting radiomix parameters and classifying uric acid stone components of urinary stones based on values calculated by inputting the radiomix parameters to a learning model learned in advance by the analysis device.

Description

Uric acid stone classification method and analysis device for urinary stones using radiomix parameters of computed tomography images

The technique described below relates to a technique for classifying uric acid stone components of urinary stones using radiomix parameters of computed tomography images.

Urolithiasis requires active treatment because it can cause complications such as severe pain, urinary tract infection, and deterioration of renal function. Urethral stones are composed of calcium oxalate stone, calcium phosphate stone, uric acid stone, cystine stone, and magnesium-ammonium-phosphate stone. As for the treatment of urinary stones, interventional treatment such as extracorporeal shock wave lithotripsy or surgery to physically remove stones is mainly performed.

Recently, among oral medications used for the purpose of preventing stones, it has been reported that potassium citrate preparations have the effect of dissolving uric acid stones. In addition, in a study on the analysis of urinary stone components in Koreans using a large-scale domestic database, the prevalence of uric acid stones was reported to be about 20% or more, and when oral medications are used to dissolve uric acid stones, invasive treatment can be avoided and medical expenses can be reduced for patients with urolithiasis. Cases that can be reduced account for a significant proportion.

Therefore, if it is possible to accurately predict the presence of uric acid stones in urinary stone patients using imaging diagnostic devices, it is possible to avoid invasive treatment such as extracorporeal shock wave lithotripsy or surgical treatment in many patients and treat urinary stones, resulting in huge medical cost savings. can be expected to be expected.

European Patent Publication No. 2696191

Computed tomography (CT) is the gold standard for diagnosing urinary stones. The technology described below attempts to classify or predict the components of uric acid stones among urinary stones based on radiomics parameters extracted from CT images.

A method for classifying urinary stones using radiomix parameters includes the steps of receiving a CT image of a subject by an analysis device, segmenting a urinary stone area in the CT image by the analysis device, and The method includes extracting radiomix parameters and classifying uric acid stone components of urinary stones based on values calculated by inputting the radiomix parameters to a learning model learned in advance by the analysis device.

An analysis device that classifies uric acid stone components of urinary stones using radiomix parameters includes an input device that receives a CT image of a subject, a storage device that stores a learning model that classifies uric acid stone components based on the radiomix parameters of the CT image, and An arithmetic unit for extracting radiomix parameters from the segmented urolithiasis region in the CT image, inputting the extracted radiomix parameters to the learning model, and classifying uric acid stone components of the urolithiasis based on output values. .

The technology described below can predict the components of urinary stones in patients with urolithiasis and determine patients for whom dissolution of uric acid stones using oral drugs is effective. Therefore, the technique described below can avoid invasive treatment for urolithiasis patients and reduce medical expenses.

1 is an example of a uric acid stone classification system for urinary stones using radiomix.
2 is an example of a process of classifying uric acid stones of urinary stones using radiomix.
3 is an example showing important radiomix parameters in the CatBoost model for uric acid stone component classification.
4 is an example of an analysis device for classifying uric acid stone components of urinary stones.

Since the technology to be described below can have various changes and various embodiments, specific embodiments will be illustrated in the drawings and described in detail. However, this is not intended to limit the technology described below to specific embodiments, and it should be understood to include all modifications, equivalents, or substitutes included in the spirit and scope of the technology described below.

Terms such as first, second, A, B, etc. may be used to describe various elements, but the elements are not limited by the above terms, and are merely used to distinguish one element from another. used only as For example, without departing from the scope of the technology described below, a first element may be referred to as a second element, and similarly, the second element may be referred to as a first element. The terms and/or include any combination of a plurality of related recited items or any of a plurality of related recited items.

In terms used in this specification, singular expressions should be understood to include plural expressions unless clearly interpreted differently in context, and terms such as “comprising” refer to the described features, numbers, steps, operations, and components. , parts or combinations thereof, but it should be understood that it does not exclude the possibility of the presence or addition of one or more other features or numbers, step-action components, parts or combinations thereof.

Prior to a detailed description of the drawings, it is to be clarified that the classification of components in the present specification is merely a classification for each main function in charge of each component. That is, two or more components to be described below may be combined into one component, or one component may be divided into two or more for each more subdivided function. In addition, each component to be described below may additionally perform some or all of the functions of other components in addition to its main function, and some of the main functions of each component may be performed by other components. Of course, it may be dedicated and performed by .

In addition, in performing a method or method of operation, each process constituting the method may occur in a different order from the specified order unless a specific order is clearly described in context. That is, each process may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the reverse order.

The technology to be described below uses a CT image among medical images. Of course, if it is a medical image capable of radiomix analysis of urinary stones, such as X-ray or ultrasound, uric acid stone components can be analyzed using the corresponding image. However, since CT images are usually used to detect urinary stones, the following description will focus on CT images.

Radiomix corresponds to texture analysis of medical images. Texture analysis refers to analyzing brightness values of pixels of a region of interest using a mathematical method. For example, texture analysis is based on (i) a method of calculating feature values based on the distribution (histogram) of brightness values within a region of interest, (ii) GLCM (Gray A method of calculating feature values using a level co-occurrence matrix) or an intensity size zone matrix in which matrices are formed according to the size with the same brightness value, (iii) a method for calculating feature values for three or more pixels There are methods such as calculating feature values using matrices that can investigate relationships.

Hereinafter, a device that classifies uric acid components of urinary stones using CT images is referred to as an analysis device. The analysis device is a device that processes certain images and data. For example, the analysis device may be implemented as a device such as a PC, smart device, or server.

1 is an example of a uric acid stone classification system 100 using radiomix. The uric acid stone classification system 100 for urolithiasis may include a medical image scanning device 110 , a clinical information database 120 , and analysis devices 150 and 180 .

The medical image scanning device 110 may be imaging equipment installed in a medical institution. The medical image scanning device 110 generates a CT image of a subject. The subject is a patient whose components of urolithiasis are to be analyzed using CT images. CT images of urolithiasis may be at least one of images from various viewpoints (sagittal plane, coronal plane, axial plane).

An electronic medical record (EMR) 120 may store a CT image of a patient generated by the medical image scanning device 110 .

1 shows an analysis server 150 and an analysis PC 180 as an example of an analysis device. The analysis device may be implemented in various forms. For example, the analysis device may be implemented as a portable mobile device.

The analysis server 150 receives a CT image of a subject from the medical image scanning device 110 or the EMR 120 . The analysis server 150 may extract radiomix parameters from the subject's CT image. The analysis server 150 may predict or classify the uric acid component of the urinary stone by inputting the extracted pre-learned learning model. The analysis server 150 delivers the analyzed result to the user 10 . The user 10 may check the analysis result or image of the analysis server 150 through the user terminal. A user terminal refers to a device such as a PC, smart device, or portable terminal.

The analysis PC 180 receives a CT image of a subject from the medical image scanning device 110 or the EMR 120 . The analysis PC 180 may receive a CT image of a subject through a wired or wireless network. Alternatively, the analysis PC 180 may receive an input from a medium (USB, SD card, CD, etc.) storing the subject's CT image. The analysis PC 180 may extract radiomix parameters from the subject's CT image. The analysis PC 180 may predict or classify the uric acid component of the urinary stone by inputting the extracted pre-learned learning model. The user 20 may check the analysis result through the analysis PC 180 .

2 is an example of a process 200 for predicting uric acid stones in urinary stones using radiomix.

The analysis device first receives a CT image of the subject (210). In this case, the CT image may be at least one of images of various viewpoints.

The analysis device may also receive settings for a specific region of interest (ROI) where the urolithiasis region is located (215). The urolithiasis area refers to an area where urolithiasis is located. The ROI can be set by the user using an interface device while viewing the CT image.

The analysis device may segment the urinary stone region in the subject's CT image (220). For example, (1) the analyzer can use the fact that the degree of X-ray attenuation of stones is different through the results of previous studies. That is, the analysis device may segment a point having a predetermined attenuation threshold or higher in the CT image as a urinary stone region. (2) Alternatively, the analysis device may segment the urinary stone region using a previously learned segmentation model. The segmentation model is a deep learning model that classifies the area based on the characteristics of the urolithiasis area.

The analysis device may extract radiomix parameters based on the segmented urolithiasis area (230). Radiomix parameters can be composed of various types of parameters. The radiomix parameters actually extracted by the researcher will be described later. The analysis device may post-process the radio mix parameters into a form to be input to the learning model. For example, the analyzer may perform one-hot vector coding on the radio mix parameters and convert them into a form such as a first order matrix or a second order matrix.

As described above, the analysis device may classify components of urinary stones to be analyzed by inputting the subject's radiomix parameters (or post-processed data) to the pre-learned learning model (240). The learning model calculates information or a probability value on whether the component of urinary stones is uric acid stone based on the input information. Since the learning model is a model that classifies whether or not it is a uric acid stone component, it can be named a classification model.

Meanwhile, the analyzer may select valid parameters from among all radio mix parameters extracted in step 230 (235). An effective parameter means a parameter whose degree of contribution to urate stone classification among all radiomix parameters is greater than the reference value. In this case, the analyzer can classify whether the component of the urinary stone is uric acid stone by inputting only valid parameters to the classification model.

Hereinafter, the researcher explains the process of building a classification model based on the data of real patients.

The researcher used the data of patients who underwent urolith component analysis after surgical treatment for urolithiasis at Kyungpook National University Hospital from January 2017 to December 2020. A total of 696 patients, excluding cases where the size of the urinary stone was small and not suitable for radiomix analysis, when the result of component analysis after urinary stone surgery was not accurately known, and when CT was not taken before urinary stone surgery, were analyzed. We analyzed 888 urinary stones extracted from CT.

CT dicom files of the researcher's urolithiasis patients were extracted.

The researcher segmented the urolithiasis area in the CT image. The researcher used the attenuation threshold to detect urolithiasis area. The researcher segmented an area larger than 130 HU (Hounsfield Unit) and defined it as a urolithiasis area.

The researcher performed texture analysis on the segmented area using python's pyradiomics module.

The researcher obtained a total of 960 lyomics parameters through texture analysis. The 960 radio mix parameters are as follows.

(1) 14 shape features

(2) 198 primary brightness statistical characteristics: including average, slope, kurtosis, energy and entropy for the gray level of the entire area. Entropy represents the irregularity of the distribution of brightness values.

(3) 242 GLCM (Gray Level Co-occurrence Matrix) features: GLCM corresponds to a histogram for a second gray level. GLCM represents the spatial relationship between pairs of pixels or voxels in different directions.

(4) High-dimensional features: High-dimensional features include 176 Gray level Size Zone Matrix (GLSZM), 154 Gray level Dependence Matrix (GLDM) and 176 Gray-Level Run Length Matrix (GLRLM). GLRLM represents the spatial distribution of contiguous pixels having the same gray level. GLSZM represents information about the number of groups of pixels (or voxels) that have the same gray level and are connected to each other. GLDM represents information about the number of connected pixels (or voxels) having the same gray level at a constant distance from a central pixel (or voxel).

The researcher standardized the extracted radiomix parameters using the z-score method.

The researcher built a learning model to predict the presence of uric acid in urinary stones by using the finally calculated radiomix parameters as input values. The learning model is trained using input data (radiomix parameters) for each specific patient and label values for the corresponding data (results of component analysis of urolithiasis for the patient).

The researcher built various learning models individually to verify their performance. Among the learning models, the performance of CatBoost, a gradient boost technique, was good. Therefore, the model was verified centering on CatBoost. Of course, uric acid stone prediction of urinary stones can be implemented with other learning models (deep learning, support vector machine, random forest, etc.).

The researcher divided the entire data into 10 data sets, used 9 of them as training data, and verified the remaining 1 data set by 10-fold cross validation.

Of the total 888 cases of urinary stones used by the researcher, 215 cases of urinary stones, which are a component of uric acid stone, showed a distribution of 24.2%. Table 1 below shows the results of 10-cross validation, and when explained based on the average value, the classification model achieved accuracy of 91.89%, AUC (Area Under the Curve) of 0.9602, recall of 79.55%, and precision of 86.32%. seemed

3 is an example showing important radiomix parameters in the CatBoost model for predicting uric acid stone components. 3 is a result of extracting parameters that greatly affect classification in the CatBoost application. Therefore, developers can select specific radio mix parameters effective for classification through analysis tools or repeated experiments. In this case, a classification model for predicting uric acid stones in urinary stones may be constructed based on selected effective radiomix parameters. That is, the classification model may be learned using specific parameters rather than all radiomix parameters.

4 is an example of an analyzer 300 for classifying uric acid stone components of urinary stones. The analysis device 300 corresponds to the above-described analysis devices (150 and 180 in FIG. 1). The analysis device 300 may be physically implemented in various forms. For example, the analysis device 300 may have a form of a computer device such as a PC, a network server, and a chipset dedicated to data processing.

The analysis device 300 may include a storage device 310, a memory 320, an arithmetic device 330, an interface device 340, a communication device 350, and an output device 360.

The storage device 310 may store a CT image of the subject.

The storage device 310 may store a program for segmenting a urolithiasis region in a CT image.

The storage device 310 may store a learning model (classification model) for classifying uric acid stone components using radiomix parameters. The stored classification model is a pretrained model.

The memory 320 may store data and information generated in the course of the analysis device 300 classifying uric acid components of urinary stones.

The interface device 340 is a device that receives certain commands and data from the outside. The interface device 340 may receive a CT image of the subject from a physically connected input device or an external storage device. Alternatively, the interface device 340 may receive a set value of a region of interest in the CT image.

The communication device 350 refers to a component that receives and transmits certain information through a wired or wireless network. The communication device 350 may receive a CT image of the subject from an external object. The communication device 350 may receive a setting value of a region of interest in the CT image. The communication device 350 may transmit an analysis result of the subject to an external object.

The communication device 350 or interface device 340 is a device that receives certain data or commands from the outside. The communication device 350 or the interface device 340 may be referred to as an input device to receive certain data.

The arithmetic device 330 may pre-process (normalize) the input CT image constantly.

The arithmetic device 330 may segment the urolithiasis region in the CT image. The arithmetic device 330 may segment a region having an attenuation threshold or higher in the CT image into a urolithiasis region. Alternatively, the arithmetic device 330 may segment the urinary stone region in the CT image using a previously learned segmentation model.

The arithmetic device 330 may extract radiomix parameters for the urolithiasis area. Radio mix parameters are as described above. Furthermore, the calculator 330 may select valid parameters from among the radio mix parameters. Valid parameters are information set in advance.

The calculation device 330 may classify whether the current urinary stone is a component of uric acid stone by inputting all radiomix parameters or valid parameters into the classification model. In some cases, the classification model may output a value proportional to the ratio of the uric acid component to the total urinary stones. In this case, if the uric acid stone component is more than a certain ratio based on the value output by the classification model, the medical staff may perform a method of dissolving uric acid stone using an oral drug.

The arithmetic device 330 may be a device such as a processor, an AP, or a chip in which a program is embedded that processes data and performs certain arithmetic operations.

The output device 360 is a device that outputs certain information. The output device 360 may output interfaces and analysis results necessary for data processing. The output device 360 may output a result of uric acid component classification of urinary stones for the subject. The classification result may be any one of information about the presence or absence of uric acid stone components among urinary stones, whether or not there are uric acid stone components that are more than enough to implement the uric acid stone dissolving method, or the ratio of uric acid stones to all urinary stones. Classification results may yield different values depending on which learning data is used to build the learning model in the learning process.

In addition, the radiomix parameter extraction method or the uric acid component prediction or classification method of urinary stones as described above may be implemented as a program (or application) including an executable algorithm that can be executed on a computer. The program may be stored and provided in a temporary or non-transitory computer readable medium.

A non-transitory readable medium is not a medium that stores data for a short moment, such as a register, cache, or memory, but a medium that stores data semi-permanently and can be read by a device. Specifically, the various applications or programs described above are CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM (read-only memory), PROM (programmable read only memory), EPROM (Erasable PROM, EPROM) Alternatively, it may be stored and provided in a non-transitory readable medium such as EEPROM (Electrically EPROM) or flash memory.

Temporary readable media include static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), and enhanced SDRAM (Enhanced SDRAM). SDRAM, ESDRAM), Synchronous DRAM (Synclink DRAM, SLDRAM) and Direct Rambus RAM (DRRAM).

This embodiment and the drawings accompanying this specification clearly represent only a part of the technical idea included in the foregoing technology, and those skilled in the art can easily understand it within the scope of the technical idea included in the specification and drawings of the above technology. It will be obvious that all variations and specific examples that can be inferred are included in the scope of the above-described technology.

Claims (9)

receiving a CT image of the subject by an analysis device;
Segmenting, by the analysis device, a urolithiasis region in the CT image;
extracting, by the analyzer, radiomix parameters from the urolithiasis region; and
A method of classifying uric acid stones of urinary stones using radiomix parameters, comprising the step of classifying, by the analysis device, the uric acid stone components of urinary stones based on values calculated by inputting the radio mix parameters into a pre-learned learning model. According to claim 1,
The method of classifying uric acid stones in urinary stones using a radiomix parameter in which the analysis device segments a region above an attenuation threshold in the CT image into the urolithiasis region. According to claim 1,
The radio mix parameters include shape features, primary brightness statistical characteristics, GLCM (Gray Level Co-occurrence Matrix), GLSZM (Gray level Size Zone Matrix), GLDM (Gray level Dependence Matrix) and GLRLM (Gray Level Co-occurrence Matrix). A method for classifying uric acid stones in urinary stones using radiomix parameters including -Level Run Length Matrix). According to claim 1,
The learning model is a method of classifying uric acid stones of urinary stones using a radiomix parameter that outputs whether or not there is a uric acid stone component higher than the level at which the urinary stones of the subject can perform the dissolution usage. an input device that receives a CT image of the subject;
a storage device for storing a learning model for classifying uric acid stone components based on radiomix parameters of a CT image; and
An arithmetic device that extracts radiomix parameters from the segmented urolithiasis area in the CT image, inputs the extracted radiomix parameters to the learning model, and classifies the uric acid component of the urolithiasis based on the value output An analysis device that classifies the components of uric acid stones in urinary stones using radiomix parameters. According to claim 5,
wherein the arithmetic device classifies the uric acid component of the urinary stone using a radiomix parameter for segmenting a region above an attenuation threshold in the CT image into the urinary stone region. According to claim 5,
wherein the calculator classifies the uric acid component of the urinary stone using the radio mix parameter inputting only some valid parameters among the radio mix parameters extracted from the urinary stone area to the learning model. According to claim 5,
The radio mix parameters include shape features, primary brightness statistical characteristics, GLCM (Gray Level Co-occurrence Matrix), GLSZM (Gray level Size Zone Matrix), GLDM (Gray level Dependence Matrix) and GLRLM (Gray Level Co-occurrence Matrix). -Level Run Length Matrix), an analysis device that classifies uric acid stone components of urinary stones using radiomix parameters. According to claim 5,
The learning model is an analysis device for classifying the uric acid stone component of the urinary stone using a radiomix parameter that outputs whether or not the uric acid stone component exists to the extent that the urinary stone of the subject can perform the dissolution usage.

Images