WO2020101428A1 - Lesion area detection device, lesion area detection method, and computer program - Google Patents

Abstract

The present specification discloses a lesion area detection method comprising: a step in which a lesion area detection device receives a diffusion weighted image (DWI) and an apparent diffusion coefficient (ADC) image of a subject; a step in which the lesion area detection device normalizes the DWI; in which the lesion area detection device normalizes the ADC image; a step in which the lesion area detection device extracts lesion areas from the normalized DWI and the normalized ADC image by using a lesion area detection model; and a step in which the lesion area detection device outputs one or more lesion areas.

Description

Lesion area detection device, lesion area detection method and computer program

The present disclosure relates to a lesion area detection apparatus, a lesion area detection method, and a computer program.

Magnetic resonance imaging (MRI) imaging device is a device that photographs an object using a magnetic field, and it shows the bone, as well as the disc, joints, nerve ligaments, heart, and cerebral blood vessels in three dimensions from the desired angle. It is widely used for diagnosis. Magnetic resonance imaging has the advantage of being able to obtain various contrast ratios by adjusting several parameters, and by using this, clinical diagnosis is performed by obtaining images of multiple contrast ratios for the same site. Korean Patent Publication No. 2009-0075644 discloses a magnetic resonance imaging device that obtains a steady-state image of a patient by changing the spin phases of the patient's fat and moisture to generate a contrast ratio of the magnetic resonance image.

One embodiment of the present invention is to provide a lesion area detection apparatus, method, and computer program for detecting a lesion area using a weighted diffusion image and a diffusion coefficient image of an object.

According to an embodiment of the present invention, a method for detecting a lesion region comprises: a lesion region detecting apparatus receiving a weighted diffusion image (DWI) and a diffusion coefficient image (ADC) of an object; Normalizing the weighted diffusion image by the lesion area detection apparatus; Normalizing the diffusion coefficient image by the lesion area detection apparatus; Extracting lesion regions from the weighted diffusion image and the diffusion coefficient image normalized by the lesion region detection apparatus using a lesion region detection model; And outputting one or more lesion areas extracted by the lesion area detection device.

The lesion area detection model repeatedly receives pre-registered lesion masks corresponding to weighted diffusion images and diffusion coefficient images, and among the weighted diffusion images and input data extracted from the diffusion coefficient images, a lesion mask It can be characterized by selecting the data to be the basis for detecting them and learning based on the selected data.

The lesion area detection model may be characterized in that it is changed to another lesion area detection model when it is determined that the predetermined evaluation criteria are not satisfied in comparison with a previously registered lesion mask.

The lesion area detection model is characterized by receiving the transformed spreading coefficient images and weighted spreading images through a process of standardizing spreading coefficient images using a preset threshold and a standard deviation of spreading coefficients of each pixel. can do.

A data receiving unit for receiving a weighted diffusion image (DWI) and a diffusion coefficient image (ADC) of an object by the lesion area detection apparatus according to an embodiment of the present invention; A pre-processing unit to normalize the weighted diffusion image and the diffusion coefficient image; A lesion region detection unit for extracting lesion regions from the weighted diffusion image and the diffusion coefficient image normalized using a lesion region detection model; And an input / output unit outputting the extracted one or more lesion regions.

A computer program according to an embodiment of the present invention may be stored in a medium to execute any one of the methods of detecting a lesion area according to an embodiment of the present invention using a computer.

In addition to this, another method for implementing the present invention, another system, and a computer-readable recording medium for recording a computer program for executing the method are further provided.

Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

According to an embodiment of the present invention, a lesion region of an object may be automatically detected through a lesion region detection model trained using a weighted diffusion image and a diffusion coefficient image.

1 is a block diagram of a lesion area detection system according to embodiments of the present invention.

2 is a block diagram of an apparatus for detecting a lesion area according to embodiments of the present invention, and FIG. 3 is a view for explaining a structure of a storage unit of the apparatus for detecting a lesion area.

4 is a block diagram of a learning model unit.

5 and 6 are flowcharts of a method for detecting a lesion area according to embodiments of the present invention.

7 is a diagram illustrating a learning process and a testing process in the lesion area detection apparatus.

8 is an exemplary view of lesion masks predicted based on diffusion weighted images and diffusion coefficient images.

9 is an exemplary diagram of a user interface for a diagnostic device that automatically measures the distribution of lesions and the volume of lesions based on the methods mentioned in the specification.

10 is a table showing a value (ICC) corresponding to a matching ratio between an actual lesion region and a lesion region obtained through various methods.

Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the contents described in the accompanying drawings. Also, a method of configuring and using an electronic device according to an embodiment of the present invention will be described in detail with reference to contents described in the accompanying drawings. The same reference numerals or symbols in each drawing denote parts or components that perform substantially the same function.

Terms including ordinal numbers such as first and second may be used to describe various components, but the components are not limited by the terms. The terms are only used to distinguish one component from other components. For example, the first component may be referred to as a second component without departing from the scope of the present invention, and similarly, the second component may be referred to as a first component. The term and / or includes a combination of a plurality of related items or any one of a plurality of related items.

Terms used in the present specification are used to describe examples, and are not intended to limit and / or limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as include or have are intended to designate the existence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, and one or more other features, numbers, or steps. It should be understood that it does not preclude the existence or addition possibility of the operation, components, parts or combinations thereof.

Throughout the specification, when a part is connected to another part, this includes not only the case of being directly connected, but also the case of being electrically connected with another element in between. Also, when a part includes a certain component, this means that other components may be further included rather than excluding other components unless otherwise specified. In addition, terms such as “... unit” and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software. .

In this specification, an image may mean multidimensional data composed of discrete image elements (eg, pixels in a 2D image and voxels in a 3D image). For example, the image may include an X-ray device, a CT device, an MRI device, an ultrasound diagnostic device, and a medical image of an object acquired by another medical imaging device.

In the present specification, the subject may include a person or an animal, or a part of a person or animal. For example, the subject may include blood vessels or devices such as the liver, heart, uterus, brain, breast, and abdomen. In addition, the object may include a phantom. Phantom refers to a material having a volume very close to the density and effective atomic number of an organism, and may include a spherical phantom having properties similar to a body.

In the present specification, the user is a medical professional, and may be a doctor, a nurse, a clinical pathologist, a medical imaging expert, or the like, but may be a technician repairing a medical device, but is not limited thereto.

In the present specification, the diffusion-weighted image is an image that maps the diffusion motion of molecules, particularly water molecules, in a biological tissue of a subject. The diffusion of water molecules within the tissue is not free. Diffusion weighted images reflect water molecules colliding with fibrous tissue or membranes. Therefore, the diffusion pattern of water molecules indicates the normal or abnormal state of the tissue. In particular, the diffusion weighted image may well represent normal and abnormal states of the white matter fiber structure or gray matter of the brain.

The diffusion coefficient image refers to the processed image based on the value processed through a function of temperature as a diffusion coefficient. The diffusion coefficient can be calculated using a diffusion weighted image because the cell walls are present in the body and the temperature is non-uniform. The infarct area reduces the diffusion of water outside the cell due to the expansion of the cell. When the diffusion-reduced area is photographed as a diffusion-weighted image, a signal decrease becomes a small area, and the diffusion-weighted image appears bright. On the other hand, the diffusion-reduced region appears darker than normal in the diffusion coefficient. Water such as cerebrospinal fluid (CSF) is a free diffusion region, where the diffusion coefficient image is bright and the diffusion weighted image is dark.

1 is a block diagram of a lesion area detection system according to embodiments of the present invention.

The lesion area detection system 10 may include an image photographing device 100 and a lesion area detection device 200.

The lesion area detection system 10 transmits an image of the object 1 acquired through the image capturing apparatus 100 to the lesion area detection device 200, and the lesion area detection device 200 models a lesion area detection model in blood vessels The lesion area of the object 1 is detected based on the obtained data.

The imaging device 100 is a device that outputs an image or image of the object 1, and may be an X-ray device, a CT device, an MRI device, an ultrasound diagnostic device, and other medical imaging devices. The imaging apparatus 100 may acquire blood vessel images of all or part of the object 1. The imaging apparatus 100 may photograph an object and output a TOF image, diffusion weighted imaging (DWI), or apparent diffusion coefficient imaging (ADC).

The lesion area detection apparatus 200 extracts input data necessary for detecting a lesion area from the diffusion weighted image and the diffusion coefficient image of the object, and uses the lesion area detection model learned using the input data to detect one or more lesion areas of the object. Can deduce. The lesion region detection apparatus 200 may detect a lesion region distributed in a blood vessel of the object based on the input image. The lesion area detection device 200 may be a computing device including one or more processors and storage media.

2 is a block diagram of the lesion area detection apparatus 200 according to embodiments of the present invention, and FIG. 3 is a view for explaining the structure of the storage unit 250 of the lesion region detection device.

The lesion area detection apparatus 200 may include a control unit 210, a communication unit 220, an input / output unit 240, and a storage unit 250.

The control unit 210 may be implemented with one or more processors, and configured to process instructions of a computer program by performing basic arithmetic, logic, and input / output operations. The command may be provided to the control unit 210 by the storage unit 250 and the communication unit 220. For example, the control unit 210 may be configured to execute a received command according to program code stored in a recording device such as the storage unit 250.

The communication unit 220 may provide a function for communicating with an external device through a network. In one example, the request generated by the control unit 210 of the lesion area detection device 200 according to the program code stored in the recording device such as the storage unit 250 is taken from the external image through the network under the control of the communication unit 220 It can be delivered to the device 100, a database, or another user terminal. For example, a control signal or command received through the communication unit 220 may be transmitted to the control unit 210 or the storage unit 250, and the received image image may be stored by the storage unit 250.

The storage unit 250 is a computer-readable recording medium, and may include a non-destructive mass storage device such as random access memory (RAM), read only memory (ROM), and a disk drive. In addition, an operating system and at least one program code may be stored in the storage unit 250. These software components may be loaded from a computer-readable recording medium separate from the storage unit 250 using a drive mechanism. Such a separate computer-readable recording medium may include a computer-readable recording medium such as a floppy drive, disk, tape, DVD / CD-ROM drive, and memory card. In other embodiments, software components may be loaded into storage 250 via communication unit 220 rather than a computer-readable recording medium. For example, at least one program is a storage unit 250 based on a program (for example, the application described above) installed by files provided by a file distribution system that distributes installation files of developers or applications through a network. Can be loaded on.

The input / output unit 240 may receive a lesion area detection input, a lesion area input, and the like, and display a screen showing a blood vessel image, a lesion area, and the like. For example, the input / output unit 240 may include an operation panel for receiving user input, a display panel for displaying the screen, and the like.

Specifically, the input unit may include devices capable of receiving various types of user input, such as a keyboard, a physical button, a touch screen, a camera, or a microphone. Also, the output unit may include a display panel or a speaker. However, the present invention is not limited thereto, and the input / output unit 240 may include a configuration supporting various input / output.

The lesion region detection apparatus 200 may include an image receiving unit 251, an image preprocessing unit 252, a lesion region detection unit 253, and a learning model unit to detect the lesion region based on the diffusion weighted image and the diffusion coefficient image of the object (254).

The image receiving unit 251 may receive a diffusion weighted image and a diffusion coefficient image of the object.

The image pre-processing unit 252 standardizes the diffusion weighted image and the diffusion coefficient image in order to input the lesion region detection model. The standardization process can be expressed by the following equation.

,

는 각각 뇌조직에서 정상부위의 DWI신호의 평균 및 표준편차이다. DWI(x, y) 는 (x,y) 위치에서 픽셀의 DWI 신호 값이고,

는 (x, y)위치에서 픽셀의 표준화된 DWI 신호 값이 된다. 표준화를 하게 되면, 다른 MRI 기계에서 얻은 DWI신호들을 하나의 신호 범위로 통일할 수 있고, 또한 neural network 모델을 학습하는데 있어서도 신호 범위가 적절하다고 볼 수 있다. 확산 계수 이미지 (ADC) 표준화 과정은 DWI 표준화 과정과 다르게 진행된다. ADC 표준화 과정은 다음의 식으로 나타낼 수 있다.

,

Is the average and standard deviation of the DWI signal in the normal region in each brain tissue. DWI (x, y) is the DWI signal value of the pixel at (x, y),

Is the standardized DWI signal value of the pixel at the (x, y) position. When standardized, DWI signals obtained from other MRI machines can be unified into a single signal range, and the signal range is also suitable for training a neural network model. The diffusion coefficient image (ADC) standardization process is different from the DWI standardization process. The ADC standardization process can be expressed by the following equation.

는 미리 정의된 확산 계수 임계 값(threshold)으로 주어진다.

는 뇌조직에서 정상부위의 확산 계수 값의 표준편차이다. 급성 뇌졸중 영상 진단에서 문턱값으로 주로 사용되는 600 (10^-6 mm²/s)이 될 수 있다.

는

위치에서 픽셀의 확산 계수 값이고

는

위치에서 픽셀의 표준화된 확산 계수 값이 된다.

Is given as a predefined diffusion coefficient threshold.

Is the standard deviation of the value of the diffusion coefficient at the normal site in brain tissue. It can be 600 (10 ^-6 mm ² / s), which is mainly used as a threshold in acute stroke imaging.

The

The pixel's diffusion coefficient at position

The

This is the pixel's normalized spreading factor value at the location.

The image preprocessing unit 252 normalizes the spreading coefficient image using the standard deviation of the spreading coefficient values of each pixel of the spreading coefficient image and the threshold value of the spreading coefficient. The image pre-processing unit 252 may standardize the diffusion weighted image in a manner similar to the diffusion coefficient image.

The lesion region detection unit 252 detects a lesion region of an object photographed using a lesion region detection model trained using a standardized diffusion coefficient image and a diffusion weighted image. The lesion region may be part of the subject's brain.

The lesion region detection apparatus 200 according to an embodiment of the present invention may detect a lesion region of an object by using the lesion region detection model learned through the learning model unit 253.

Specifically, the learning model unit 254 may include a data learning unit 2251 and a data recognition unit 2542. The operation of each component is as follows.

The data learning unit 2251 may learn criteria for detecting a lesion area. The data learning unit 2251 may learn criteria to determine what data to use to determine a predetermined lesion area and how to determine the lesion area using the data. The data learning unit 2251 may be trained to use a diffusion coefficient image and a weighted diffusion image as data to be used for learning. In addition, the data learning unit 2251 may learn to detect a lesion region using a standardized diffusion coefficient image and a weighted diffusion image. The data learning unit 2251 may output a mask image representing the lesion area.

The data recognition unit 2252 may detect the lesion area of the object based on the input data. The data recognition unit 2252 may detect the lesion region of the object from predetermined data using the trained lesion region detection model. The data recognition unit 2252 acquires predetermined data according to a preset criterion by learning, and uses the acquired data as an input value to use the lesion area detection model, thereby detecting the lesion area of the object based on the predetermined data. Can be. Further, the result value output by the lesion region detection model using the obtained data as an input value may be used to update the lesion region detection model.

At least one of the data learning unit 2251 and the data recognition unit 2542 may be manufactured in the form of at least one hardware chip and mounted on an electronic device. For example, at least one of the data learning unit 2251 and the data recognition unit 2252 may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or an existing general-purpose processor (for example, a CPU) Alternatively, it may be manufactured as a part of an application processor or a graphics-only processor (for example, a GPU) and mounted on various electronic devices described above.

In this case, the data learning unit 2251 and the data recognition unit 2542 may be mounted on one electronic device, or may be mounted on separate electronic devices, respectively. For example, one of the data learning unit 2251 and the data recognition unit 2252 may be included in the electronic device, and the other may be included in the server. In addition, the data learning unit 2251 and the data recognition unit 2252 may provide the model information constructed by the data learning unit 2251 to the data recognition unit 2252 through wired or wireless communication. The data input to 2542) may be provided to the data learning unit 2251 as additional learning data.

Meanwhile, at least one of the data learning unit 2251 and the data recognition unit 2252 may be implemented as a software module. When at least one of the data learning unit 2251 and the data recognition unit 2542 is implemented as a software module (or a program module including an instruction), the software module is a computer-readable, non-transitory readable It may be stored in a readable media (non-transitory computer readable media). In addition, in this case, at least one software module may be provided by an operating system (OS) or may be provided by a predetermined application. Alternatively, some of the at least one software module may be provided by an operating system (OS), and the other may be provided by a predetermined application.

As shown in FIG. 4, the data learning unit 2251 includes a data acquisition unit 2541-1, a pre-processing unit 2251-2, a training data selection unit 2541-3, and a model learning unit 2541-4 , May include a model evaluation unit 2251-5.

The data acquisition unit 251-1 may acquire data necessary for detecting a lesion region. The data acquisition unit 251-1 may acquire data necessary for learning to detect a lesion region. The data acquisition unit 2521-1 may acquire images, images, and text related to blood vessels from the lesion area detection device 200 or the image capture device 100. Specifically, the data acquisition unit 2541-1 may acquire a diffusion weighted image and a diffusion coefficient image.

The preprocessing unit 2251-2 may preprocess the acquired data so that the acquired data can be used for learning to detect the lesion area. The pre-processing unit 2251-2 may process the acquired data in a preset format so that the model learning unit 2251-4, which will be described later, can use the acquired data for learning to detect the lesion area. The preprocessing unit 2251-2 may resize the diffusion weighted image and the diffusion coefficient image. The preprocessing unit 2251-2 resizes the resolution of the diffusion weighted image and the diffusion coefficient image while increasing or decreasing the resolution at a predetermined ratio. The pre-processing unit 2251-2 can standardize the diffusion coefficient image and the weighted diffusion image. Specifically, the pre-processing unit 2251-2 can standardize the spreading coefficient image by using the standard deviation of the spreading coefficient values of each pixel of the spreading coefficient image and a preset threshold.

The learning data selection unit 2251-3 may select data necessary for learning from pre-processed data. The selected data may be provided to the model learning unit 2251-4. The learning data selector 2251-3 may select data necessary for learning from pre-processed data according to a preset criterion for detecting a lesion region. In addition, the learning data selection unit 313 may select data according to criteria set by learning by the model learning unit 314 to be described later.

The model learning unit 2251-4 may train a lesion area detection model used for lesion area detection using learning data. In this case, the lesion area detection model may be a model learning using an artificial neural network algorithm. For example, the lesion area detection model may extract features using a low-resolution weighted diffusion image and a diffusion coefficient image with a reduced resolution of the weighted diffusion image and the diffusion coefficient image.

The model learning unit 2541-4 restores regions corresponding to the lesion to a high dimension by using low-dimensional feature information extracted from the weighted diffusion image and the diffusion coefficient image, and the regions corresponding to the lesion are a lesion mask image, that is, a segmentation map ( segmentation map).

The lesion area detection model may be constructed in consideration of the application field of the recognition model, the purpose of learning, or the computer performance of the device. The lesion area detection model may be, for example, a model based on a neural network. For example, a model such as a deep neural network (DNN), a recurrent neural network (RNN), or a bidirectional recurrent deep neural network (BRDNN) may be used as a lesion area detection model, but is not limited thereto.

In addition, the model learning unit 2251-4 may train the lesion area detection model, for example, through supervised learning using learning data as an input value. In addition, the model learning unit 2251-4, for example, learns the type of data necessary for the detection of a lesion region without any guidance, thereby unsupervised learning to discover a criterion for the detection of the lesion region. Through this, the lesion area detection model can be trained. In addition, the model learning unit 314 may train the lesion area detection model, for example, through reinforcement learning using feedback on whether the result of the detection of the lesion area according to learning is correct.

In addition, when the lesion region detection model is trained, the model learning unit 2251-4 may store the trained lesion region detection model. In this case, the model learning unit 2251-4 may store the trained lesion area detection model in the memory of the electronic device including the data recognition unit 2542. Alternatively, the model learning unit 2251-4 may store the trained lesion area detection model in the memory of the electronic device including the data recognition unit 2542 to be described later. Alternatively, the model learning unit 2251-4 may store the trained lesion area detection model in a memory of a server connected to a wired or wireless network with the electronic device.

In this case, the memory in which the learned lesion area detection model is stored may store, for example, instructions or data related to at least one other component of the electronic device. Also, the memory may store software and / or programs. The program may include, for example, a kernel, middleware, application programming interface (API), and / or application program (or "application").

The model evaluation unit 2251-5 inputs evaluation data to the lesion area detection model, and, when the recognition result output from the evaluation data does not satisfy a predetermined criterion, causes the model learning unit 2251-4 to learn again. Can be. In this case, the evaluation data may be preset data for evaluating the lesion area detection model.

For example, the model evaluator 2251-5, when the number or ratio of the evaluation data for which the lesion area detection result is not accurate among the recognition results of the trained lesion area detection model for the evaluation data exceeds a preset threshold. It can be evaluated that the predetermined criteria are not satisfied.

On the other hand, when there are a plurality of trained lesion area detection models, the model evaluator 2251-5 evaluates whether or not the predetermined criteria are satisfied for each trained lesion area detection model, and finalizes the model that satisfies the predetermined criteria. It can be determined as a lesion area detection model. In this case, when there are a plurality of models satisfying a predetermined criterion, the model evaluator 2251-5 may determine, as a final lesion area detection model, any one or a predetermined number of models preset in order of highest evaluation score.

On the other hand, the data acquisition unit 2541 in the data learning unit 2251, the pre-processing unit 2251-2, the learning data selection unit 2541-3, the model learning unit 2541-4 and the model evaluation unit 2541 At least one of -5) may be manufactured in the form of at least one hardware chip and mounted on an electronic device. For example, at least one of the data acquisition unit 2541-1, the pre-processing unit 2251-2, the training data selection unit 2251-3, the model learning unit 2251-4 and the model evaluation unit 2551-5 One may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or as part of an existing general-purpose processor (eg, CPU or application processor) or graphics-only processor (eg, GPU), as described above. It may be mounted on various electronic devices.

In addition, the data acquisition unit (2541-1), the pre-processing unit (2541-2), the training data selection unit (2541-3), the model learning unit (2541-4) and the model evaluation unit (2541-5) is one electronic It may be mounted on a device, or may be mounted on separate electronic devices, respectively. In addition, at least one of the data acquisition unit (2541-1), the pre-processing unit (2541-2), the training data selection unit (2541-3), the model learning unit (2541-4) and the model evaluation unit (2541-5) It can be implemented as a software module. At least one of the data acquisition unit 251-1, the pre-processing unit 2551-2, the training data selection unit 2551-3, the model learning unit 2541-4, and the model evaluation unit 2551-5 is a software module When implemented as (or a program module including an instruction), the software module may be stored in a computer-readable non-transitory computer readable media.

Next, the structure of the data recognition unit 2542 will be described.

The data recognition unit 2252 according to an embodiment of the present invention includes a data acquisition unit 2542-1, a preprocessing unit 2542-2, a recognition data selection unit 2542-3, and a recognition result providing unit 2542-4 And a model update unit 2542-5.

The data acquisition unit 2252-1 may acquire data necessary for detecting the lesion area, and the pre-processing unit 2252-2 may preprocess the acquired data so that the acquired data for detecting the lesion area can be used. have. The pre-processing unit 2542-2 may process the acquired data in a preset format so that the recognition result providing unit 2542-4, which will be described later, can use the acquired data to detect the lesion area.

The preprocessing unit 2251-2 may resize the diffusion weighted image and the diffusion coefficient image. The preprocessing unit 2251-2 resizes the resolution of the diffusion weighted image and the diffusion coefficient image while increasing or decreasing the resolution at a predetermined ratio. The pre-processing unit 2251-2 can standardize the diffusion coefficient image and the weighted diffusion image. Specifically, the pre-processing unit 2251-2 can standardize the spreading coefficient image by using the standard deviation of the spreading coefficient values of each pixel of the spreading coefficient image and a preset threshold.

The recognition data selection unit 2542-3 may select data necessary for detecting a lesion area from among pre-processed data. The selected data may be provided to the recognition result providing unit 2542-4. The recognition data selection unit 2542-3 may select some or all of the pre-processed data according to a preset criterion for detecting a lesion area. In addition, the recognition data selection unit 2542-3 may select data according to a preset criterion by training by the model learning unit 2251-4.

The recognition result providing unit 2542-4 may apply the selected data to the lesion region detection model to determine the lesion region detection. The recognition result providing unit 2542-4 may provide a recognition result according to the purpose of recognizing data. The recognition result providing unit 2542-4 can apply the selected data to the lesion area detection model by using the data selected by the recognition data selection unit 2542-3 as an input value. In addition, the recognition result is determined by the lesion region detection model, and may include a segmentation map representing only the region corresponding to the lesion of the object.

The model update unit 2252-5 may cause the lesion area detection model to be updated based on the evaluation of the recognition result provided by the recognition result providing unit 2542-4. For example, the model updating unit 2542-5 provides the model learning unit 2542-4 with the recognition result provided by the recognition result providing unit 2542-4, so that the model learning unit 2542-4 The lesion area detection model can be updated.

On the other hand, the data acquisition unit 2542-1, the pre-processing unit 2252-2, the recognition data selection unit 2542-3, the recognition result providing unit 2542-4 and the model update unit in the data recognition unit 2542 ( At least one of 2542-5) may be manufactured in the form of at least one hardware chip and mounted on an electronic device. For example, among the data acquisition unit 2542-1, the pre-processing unit 2542-2, the recognition data selection unit 2542-3, the recognition result providing unit 2542-4, and the model update unit 2542-5 At least one may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), or it may be manufactured as part of an existing general-purpose processor (for example, a CPU or application processor) or a graphics-only processor (for example, GPU). It may be mounted on various electronic devices.

4 and 5 are flowcharts of a method for detecting a lesion area according to embodiments of the present invention.

As shown in FIG. 4, the method for detecting a lesion area receives an image captured from the image capturing apparatus 100.

In S120, the lesion area detection apparatus 200 receives a weighted diffusion image and a diffusion coefficient image of the object.

In S130, the lesion area detection apparatus 200 standardizes the weighted diffusion image and the diffusion coefficient image, respectively.

In S140, the lesion region detection apparatus 200 extracts lesion regions from a standardized weighted diffusion image and a diffusion coefficient image using a lesion region detection model.

In S150, the lesion area detecting apparatus 200 outputs the extracted lesion areas. The lesion area detecting apparatus 200 may output a lesion mask image representing a region corresponding to the lesion as 1, that is, a segmentation map.

As shown in FIG. 5, the lesion area detection apparatus 200 may train a lesion area detection model.

In S210, the lesion area detection apparatus 200 receives a weighted diffusion image and a diffusion coefficient image of the object.

In S220, the lesion area detection apparatus 200 preprocesses the weighted diffusion image and the diffusion coefficient image. The lesion area detection apparatus 200 may resize the diffusion weighted image and the diffusion coefficient image. The lesion area detection apparatus 200 resizes the resolution of the diffusion weighted image and the diffusion coefficient image while increasing or decreasing the resolution at a predetermined ratio. The lesion region detection apparatus 200 may standardize the diffusion coefficient image and the weighted diffusion image. Specifically, the lesion area detection apparatus 200 may standardize the diffusion coefficient image using a standard deviation and a preset threshold value of the diffusion coefficient values of each pixel of the diffusion coefficient image.

In S230, the lesion region detection apparatus 200 trains a lesion region detection model using a pre-processed weighted diffusion image and a diffusion coefficient image.

In S240, the lesion area detecting apparatus 200 outputs a lesion mask image specific to the lesion area in consideration of the location of the lesion and the degree of the lesion output through the trained lesion area detection model.

In S250, the lesion area detection apparatus 200 compares a preset lesion mask and an output lesion mask image to evaluate a lesion area detection model.

In S260, the lesion area detection apparatus 200 stores a lesion area detection model that satisfies a predetermined evaluation criterion in the storage unit.

Through this, the lesion region detection apparatus 200 according to the embodiments of the present invention may receive the lesion region detection model through the learning model apparatus 300 and infer the vascular age of the subject using the lesion region detection model. .

As shown in FIG. 7, the learning model unit 254 may learn through a weighted diffusion image and a diffusion coefficient image and a lesion mask set by an expert. The learning model unit 254 forms a lesion area detection model based on the parameters obtained as a result of the learning.

The learning model unit 254 may generate a supervised learning of U-net model through a lesion mask selected by experts for a weighted diffusion image (DWI) and a diffusion coefficient image (ADC).

The lesion mask is output by inputting a weighted diffusion image (DWI) and a diffusion coefficient image (ADC) of an actual patient in the learned lesion area detection model. In this case, the trained lesion area detection model may be trained through various algorithms of machine learning such as U-net, ED-CNN, and SegNet.

The batch size to be learned may be determined according to processor performance, graphics processing power, instruction processing power, etc. of the computing device including the learning model unit 254.

FIG. 8 is an example showing, in each case, a map of lesions predicted by an actual machine learning model and a map of lesions designated by a stroke specialist when weighted diffusion images and diffusion coefficient images are given as inputs. It can be seen that the maps of the two lesions qualitatively match each other.

9 is an exemplary diagram of a user interface for a diagnostic device that automatically measures the distribution of lesions and the volume of lesions based on the methods mentioned in the specification.

10 is a table showing a value (ICC) corresponding to a matching ratio between an actual lesion region and a lesion region obtained through various methods.

According to the table of FIG. 10, as a result of comparing a manual segmentation method by a diagnostic doctor with a lesion area detection model according to an embodiment of the present invention, ICC according to a lesion area detection model (RAPID-U-net (DWI + ADC)) The value was found to be 0.99. According to the corresponding table, it was found that the lesion area extracted by considering only the weighted diffusion image (DWI) had a correlation value with the actual lesion area of only 0.86. That is, as a result of the experiment, it was found that the lesion area detection model of the present invention that detects the lesion area in consideration of both DWI and ADC exhibits excellent performance. The lesion area detection model may be generated by optimizing parameters such as the number of layers, the convolution filter, the number of epochs, and the batch size by the learning model unit 254.

The device described above may be implemented with hardware components, software components, and / or combinations of hardware components and software components. For example, the devices and components described in the embodiments include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors (micro signal processors), microcomputers, field programmable gate arrays (FPGAs). , A programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions, may be implemented using one or more general purpose computers or special purpose computers. The processing device may perform an operating system (OS) and one or more software applications running on the operating system. In addition, the processing device may access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of understanding, a processing device may be described as one being used, but a person having ordinary skill in the art, the processing device may include a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that may include. For example, the processing device may include a plurality of processors or a processor and a controller. In addition, other processing configurations, such as parallel processors, are possible.

The software may include a computer program, code, instruction, or a combination of one or more of these, and configure the processing device to operate as desired, or process independently or collectively You can command the device. Software and / or data may be interpreted by a processing device, or to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodied in the transmitted signal wave. The software may be distributed over networked computer systems, and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, or the like alone or in combination. The program instructions recorded in the medium may be specially designed and configured for the embodiments or may be known and usable by those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs, DVDs, and magnetic media such as floptical disks. -Hardware devices specifically configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, etc., as well as machine language codes produced by a compiler. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described by a limited embodiment and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques are performed in a different order than the described method, and / or the components of the described system, structure, device, circuit, etc. are combined or combined in a different form from the described method, or other components Alternatively, even if substituted or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (9)

A lesion area detection apparatus receiving a weighted diffusion image (DWI) and a diffusion coefficient image (ADC) of the object; Normalizing the weighted diffusion image by the lesion area detection apparatus; Normalizing the diffusion coefficient image by the lesion area detection apparatus; Extracting lesion regions from the weighted diffusion image and the diffusion coefficient image normalized by the lesion region detection apparatus using a lesion region detection model; And And outputting one or more lesion areas extracted by the lesion area detection device. According to claim 1, The lesion area detection model Weighted diffusion images and diffusion coefficient images corresponding to pre-registered lesion masks are repeatedly input, and among the weighted diffusion images and input data extracted from the diffusion coefficient images, a reference is made to detecting lesion masks. A method of detecting a lesion area, characterized by selecting data and learning based on the selected data. According to claim 1, The lesion area detection model When it is determined that a predetermined evaluation criterion is not satisfied in comparison with a previously registered lesion mask, a method of detecting a lesion region is characterized in that it is changed to another lesion region detection model. According to claim 1, The lesion area detection model A method for detecting a lesion area, characterized in that the diffusion coefficient images and weighted diffusion images received through the process of standardizing the diffusion coefficient images using a preset threshold value and a standard deviation of the diffusion coefficients of each pixel are input and learned. . A data receiving unit that receives a weighted diffusion image (DWI) and a diffusion coefficient image (ADC) of the object; A pre-processing unit to normalize the weighted diffusion image and the diffusion coefficient image; A lesion region detection unit for extracting lesion regions from the weighted diffusion image and the diffusion coefficient image normalized using a lesion region detection model; And And an input / output unit for outputting the extracted one or more lesion regions. The method of claim 5, The lesion area detection model Weighted diffusion images and diffusion coefficient images corresponding to pre-registered lesion masks are repeatedly input, and among the weighted diffusion images and input data extracted from the diffusion coefficient images, a reference is made to detecting lesion masks. A device for detecting a lesion area, characterized in that data are selected and learned based on the selected data. The method of claim 5, The lesion area detection model When it is determined that the predetermined evaluation criteria are not satisfied in comparison with the previously registered lesion mask, the lesion area detection device is characterized in that it is changed to another lesion area detection model. The method of claim 5, The lesion area detection model A device for detecting a lesion area, characterized in that the received diffusion coefficient images and weighted diffusion images are learned through the process of standardizing the diffusion coefficient images using a preset threshold and a standard deviation of the diffusion coefficient of each pixel. . A computer program stored in a computer-readable storage medium for executing the method of claim 1 using a computer.

Images