KR102229939B1 - Apparatus for classifying pacemaker using artificial neural network, method thereof and computer recordable medium storing program to perform the method - Google Patents

Abstract

The present invention relates to an apparatus for classifying a pacemaker using an artificial neural network, a method therefor, and a computer-readable recording medium in which a program performing the method is recorded. Includes a plurality of layers constituting the next layer through a plurality of operations to which a weight determining intensity is applied, and when a radiographic image including a pacemaker is input, whether the pacemaker is a pacemaker capable of MRI imaging by performing the plurality of operations After learning of the artificial neural network is completed, the artificial neural network that outputs the probability of the probabilities as an output value according to the calculation, and after the learning of the artificial neural network is completed, the radiographic image of the subject to be examined is input to the artificial neural network to derive the output value, and the examination A device for classifying a pacemaker, characterized in that it includes a classification module that determines whether the pacemaker worn by the subject is a pacemaker capable of MRI imaging, a method for the same, and a computer-readable recording medium in which a program for performing the method is recorded. to provide.

Description

Apparatus for classifying pacemaker using artificial neural network, method thereof and computer recordable medium storing program to perform the method }

The present invention relates to a pacemaker classification technology, and more particularly, an apparatus for classifying a pacemaker capable of MRI imaging from an X-RAY image using an artificial neural network, a method for the same, and a program for performing the method are recorded. It relates to a computer-readable recording medium.

Artificial neural networks are statistical learning algorithms inspired by biological neural networks (especially the brain in the central nervous system of animals) in machine learning and cognitive science. The artificial neural network refers to the overall model with problem-solving ability by changing the strength of synaptic bonding through learning by artificial neurons (nodes) that form a network through synaptic bonding. In a narrow sense, it may refer to a multilayer perceptron using the error backpropagation method, but this is an incorrect usage, and artificial neural networks are not limited thereto. Artificial neural networks include teacher learning, which is optimized for problems by input of teacher signals, and non-history learning, which does not require teacher signals. Teacher learning is used when there is a clear answer, and comparative history learning is used for data clustering. Artificial neural networks rely on many inputs and are generally used to guess and approximate a function wrapped in a veil. In general, it is expressed as an interconnection of neuronal systems that calculate values from inputs and has adaptability, allowing machine learning such as pattern recognition to be performed. For example, a neural network for handwriting recognition is defined as a set of input neurons, which are activated by pixels in the input image. After the transformation and weighting of the function is applied, the activation of that neuron is transmitted to other neurons. This process repeats until the last output neuron is activated, which is determined by which character has been read. Like other machine learning-learning from data-neural networks are commonly used to solve a wide range of problems such as computer vision or speech recognition that are difficult to solve with rule-based programming.

Korean Patent Publication No. 2018-0040287 published on April 20, 2018 (Name: Medical image reading and diagnosis integrated system through machine learning)

The pacemaker worn by the patient is likely to stop due to the high magnetic field during MRI, which is an imaging device that uses a high magnetic field. Therefore, patients wearing a pacemaker could not perform MRI scans. However, recently, an MR conditional pacemaker, a pacemaker that can operate in high magnetic fields, has been developed, and more and more patients have undergone MR conditional pacemaker. Therefore, it is required to confirm what kind of pacemaker the patient is wearing. Accordingly, an object of the present invention is to provide an apparatus capable of discriminating whether a pacemaker worn by a patient is a pacemaker capable of MRI imaging, a method therefor, and a computer-readable recording medium in which a program for performing the method is recorded.

In order to achieve the above-described object, the apparatus for classifying a pacemaker according to a preferred embodiment of the present invention selects the next layer through a plurality of operations in which a weight for determining the strength of the connection between the layers is applied to the output of any one layer. An artificial neural network that includes a plurality of layers constituting, and when a radiographic image including a pacemaker is input, performs the plurality of calculations and outputs a probability of whether the pacemaker is a pacemaker capable of MRI imaging as an output value according to the calculation, After the learning of the artificial neural network is completed, a radiographic image of the subject to be examined is input to the artificial neural network to derive the output value, and according to the derived output value, a division to determine whether the pacemaker worn by the subject to be examined is a pacemaker capable of MRI imaging Includes modules.

Before completing the learning of the artificial neural network, a radiographic image including a pacemaker, which is learning data, is input to the artificial neural network so that the difference between the target value corresponding to the learning data and the output value of the artificial neural network according to the learning data is minimized. It further comprises a learning module for modifying the weight of the artificial neural network through a diffusion algorithm.

The artificial neural network includes at least one convolutional layer including a plurality of kernels, wherein the kernel extracts features of an external shape of a pacemaker from the radiographic image.

The artificial neural network includes at least one convolutional layer including a plurality of kernels, wherein the kernel extracts at least one feature of a pacemaker texture, a slope, an interval, a thickness, and a shadow from the radiographic image. .

Before the radiographic image is input to the artificial neural network, a preprocessing module further includes a preprocessing module for removing a portion representing bones other than a pacemaker from the radiographic image.

And a pre-processing module for emphasizing the pacemaker by adjusting a value of the histogram before the radiographic image is input to the artificial neural network.

An apparatus for discriminating a pacemaker according to a preferred embodiment of the present invention for achieving the above-described object includes an image classifying unit that receives a radiographic image of a subject to be examined and determines whether the subject is wearing a pacemaker capable of MRI imaging; It is determined that the test subject is wearing a pacemaker capable of MRI imaging in both the image classification unit and the information verification unit, and an information verification unit that receives medical information of the subject to be examined and determines whether the subject is wearing a pacemaker capable of MRI imaging. Then, it includes an imaging determination unit that finally determines that the subject is capable of MRI imaging.

The method for discriminating a pacemaker according to a preferred embodiment of the present invention for achieving the above-described object includes the steps of receiving a radiographic image of a subject to be examined by an artificial neural network, and a plurality of operations in which weights are applied to the radiographic image. Outputting the probability of whether the pacemaker is a pacemaker capable of MRI imaging as an output value according to the calculation, and the classification module determines whether the pacemaker worn by the subject to be examined is a pacemaker capable of MRI imaging according to the derived output value It includes the step of.

Before the outputting, the learning module inputs a radiographic image including a pacemaker, which is training data, to the artificial neural network, and the artificial neural network performs a plurality of operations to which weights are applied to obtain an output value corresponding to the training data. Outputting, and modifying, by the learning module, the weight of the artificial neural network through a despreading algorithm so that a difference between a target value corresponding to the training data and an output value of the artificial neural network according to the training data is minimized. Includes.

The outputting of the output value according to the calculation includes the step of extracting, by the artificial neural network, an external shape of the pacemaker from the radiographic image through at least one convolutional layer including a plurality of kernels.

In the outputting of the output value according to the calculation, the artificial neural network extracts at least one feature of the texture, slope, interval, thickness, and shadow of the pacemaker from the radiographic image through at least one convolution layer including a plurality of kernels. It includes the step of.

Before the step of receiving the input, the step of removing a portion representing a bone other than a pacemaker from the radiographic image is further included.

Before the step of receiving the input, the step of emphasizing the pacemaker by adjusting the value of the histogram in the radiographic image is further included.

A method for discriminating a pacemaker according to a preferred embodiment of the present invention for achieving the above-described object includes the steps of determining whether the subject is wearing a pacemaker capable of MRI by receiving a radiographic image of a subject to be examined by an image classifier; , The information verification unit receives the medical information of the subject to be examined and determines whether the subject is wearing a pacemaker capable of MRI imaging, and the imaging determination unit allows both the image classifier and the information verification unit to perform MRI imaging. If it is determined that the pacemaker is worn, finally determining that the subject to be examined is capable of MRI imaging.

In a computer-readable recording medium on which a program for performing a method for distinguishing a pacemaker according to a preferred embodiment of the present invention for achieving the above object is recorded, the output of any one layer is the strength of the connection between the layers. It includes a plurality of layers constituting the next layer through a plurality of operations to which a weight for determining the value is applied, and outputs a probability of whether the pacemaker is a pacemaker capable of MRI imaging as an output value according to the calculation by performing the plurality of operations. The step of learning an artificial neural network, the step of deriving the output value by inputting the radiographic image of the subject to be examined by the classification module to the artificial neural network, and the pacemaker worn by the test subject according to the derived output value by the classification module. It provides a computer-readable recording medium in which a program for performing a method for discriminating a pacemaker, comprising the step of determining whether or not a pacemaker capable of MRI imaging is recorded.

The training of the artificial neural network includes inputting, by the learning module, a radiographic image including a pacemaker, which is training data, to the artificial neural network, and when the artificial neural network outputs an output value through a plurality of calculations corresponding to the training data. And modifying the weight of the artificial neural network through a despreading algorithm so that a difference between the target value corresponding to the training data and the output value is minimized.

The deriving of the output value includes, by the artificial neural network, extracting features of the external shape of the pacemaker from the radiographic image through at least one convolutional layer including a plurality of kernels; Way.

The step of deriving the output value includes the step of extracting at least one feature of the texture, slope, interval, thickness, and shadow of the pacemaker from the radiographic image through at least one convolutional layer including a plurality of kernels by the artificial neural network. Includes.

Before the step of deriving the output value, the step of removing a portion representing bones other than the pacemaker from the radiographic image is further included.

Before the step of deriving the output value, the step of emphasizing the pacemaker by adjusting the value of the histogram in the radiographic image is further included.

Using an artificial neural network, it is possible to know whether the pacemaker worn by the test subject is capable of MRI. Therefore, in the case of a patient wearing a pacemaker capable of MRI according to the type of pacemaker, a more detailed diagnosis is possible through MRI.

1 is a block diagram illustrating an apparatus for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention.
2 is a block diagram illustrating a configuration of an image classifying unit according to an embodiment of the present invention.
3 is a diagram for explaining the configuration of an artificial neural network of an image segment according to an embodiment of the present invention.
4 is a flowchart illustrating a method for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention.
5 is a flowchart illustrating a method of learning an artificial neural network according to an embodiment of the present invention.
6 is a view for explaining a method of learning an artificial neural network according to an embodiment of the present invention.
7 is a flowchart illustrating a method for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention.

Prior to the detailed description of the present invention, terms or words used in the present specification and claims to be described below should not be construed as being limited to their usual or dictionary meanings, and the inventors will use their own invention in the best way. In order to explain, based on the principle that it can be appropriately defined as a concept of a term, it should be interpreted as a meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are only the most preferred embodiments of the present invention, and do not represent all the technical ideas of the present invention, and thus various equivalents that can replace them at the time of application It should be understood that there may be water and variations.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this case, it should be noted that the same components in the accompanying drawings are indicated by the same reference numerals as possible. In addition, detailed descriptions of known functions and configurations that may obscure the subject matter of the present invention will be omitted. For the same reason, some components in the accompanying drawings are exaggerated, omitted, or schematically illustrated, and the size of each component does not entirely reflect the actual size.

First, a configuration of an apparatus for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention will be described. 1 is a block diagram illustrating an apparatus for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention. Referring to FIG. 1, the heart rate device 10 includes an image classification unit 100, an information verification unit 200, and a photographing determination unit 300.

The image classifying unit 100 is for receiving a radiographic image of a subject to be examined and determining whether the subject is wearing a pacemaker capable of MRI imaging. To this end, the image classifying unit 100 includes an artificial neural network. The image classifier 100 may determine whether a pacemaker of a radiographic image of a subject to be examined is a pacemaker capable of MRI imaging through an artificial neural network.

The information verification unit 200 receives medical information of a subject to be examined and determines whether the subject is wearing a pacemaker capable of MRI imaging. Here, the medical information includes personal information, clinical information, and reason for photographing. Accordingly, the information verification unit 200 may determine whether the subject to be examined is wearing a pacemaker capable of MRI imaging through personal information, clinical information, photographing reason, and the like.

When both the image classifying unit 100 and the information verifying unit 200 determine that the test subject is wearing a pacemaker capable of MRI imaging, the imaging determination unit 300 finally determines that the subject is capable of MRI imaging. As described above, by making a final determination only when it is determined that both the image classifying unit 100 and the information verifying unit 200 are wearing a pacemaker capable of MRI imaging, the reliability of the corresponding information can be improved.

Then, in more detail, the image classifying unit 100 according to an embodiment of the present invention will be described. 2 is a block diagram illustrating a configuration of an image classifying unit according to an embodiment of the present invention. In addition, FIG. 3 is a view for explaining the configuration of an artificial neural network of an image segment according to an embodiment of the present invention.

Referring to FIG. 2, the image classification unit 100 includes an artificial neural network 110, a preprocessing module 120, a learning module 130, and a classification module 140.

The artificial neural network 110 includes a plurality of layers. The plurality of layers configures the next layer through a plurality of operations in which the output of any one layer is weighted. Here, the weight determines the strength of the inter-layer connection. When a radiographic image including a pacemaker is input, the artificial neural network 110 performs a plurality of calculations to which weights are applied, and outputs a result of the calculation. The output value according to the result of this operation represents the probability of whether the pacemaker of the radiographic image is a pacemaker capable of MRI imaging.

The preprocessing module 120 is for performing preprocessing to facilitate learning or discrimination of the data before inputting data to the artificial neural network 110. For example, the preprocessing module 120 may perform preprocessing of removing portions representing bones other than the pacemaker (PM) from the radiographic image. According to another embodiment, the pre-processing module 120 may perform pre-processing of emphasizing the pacemaker (PM) by adjusting the value of the histogram of the radiographic image. This preprocessing module 120 may be selectively omitted.

The learning module 130 is for learning so that the artificial neural network 110 can output a probability of whether the pacemaker of the radiographic image is a pacemaker capable of MRI imaging by using a plurality of learning data. The learning module 130 uses, as learning data, a radiographic image that already knows whether the pacemaker of the radiographic image is a pacemaker capable of MRI imaging. In addition, the learning module 130 learns the artificial neural network based on whether or not it is a pacemaker capable of MRI imaging of a known radiographic image as a target value. The learning module 130 inputs the radiographic image, which is the training data, into the artificial neural network 110, and artificially through a back-propagation algorithm so that the difference between the output value of the artificial neural network 110 and the target value is minimized. The weight of the neural network 110 is corrected.

When the learning of the artificial neural network 110 is completed, the classification module 140 inputs a radiographic image of the subject to be examined, which is real data, to the artificial neural network 110 to derive an output value, and a pacemaker worn by the subject to be examined according to the derived output value. It is determined whether is a pacemaker capable of MRI imaging.

3, the artificial neural network 110 according to an embodiment of the present invention includes an input layer (IL), a convolution layer (CL), a pooling layer (PL), and a fully connected layer. It includes a (fully-connected layer: FL) and an output layer (ML).

The input layer (IL) is made up of a matrix. Each element of the input layer (IL) matrix corresponds to each pixel of the radiographic image. 3, the artificial neural network 110 includes a first convolutional layer (CL1), a first pooling layer (PL1), a second convolutional layer (CL2), and a second pooling layer (PL2). Although illustrated as being composed of two pairs, the present invention is not limited thereto, and each of the convolutional layer CL and the pooling layer PL may exist as one or two or more pairs. In addition, according to an embodiment, as shown in FIG. 3, when the convolutional layer CL and the pooling layer PL exist in two or more pairs, the convolutional layer CL and the pooling layer PL are alternately Although shown to be arranged as, the present invention is not limited thereto.

Each of the convolutional layers (CL: CL1, CL2) and the pooling layers (PL: PL1, PL2) consists of a plurality of feature maps, and each of these feature maps is a matrix having a predetermined size. The value of each element of the matrix constituting the feature map is calculated by applying a convolution operation or a pooling operation (pooling or subsampling) using the kernel (W1, W2, W3, W4, W5) to the previous layer. Here, the kernel (W1, W2, W3, W4, W5) is a matrix of a predetermined size, and the value of each element of the matrix constituting the kernel (W1, W2, W3, W4, W5) becomes the weight (w). .

The fully-connected layer (FL) includes a plurality of nodes (or sigmoid: f1, f2, f3, ..., fn), and the weight (w) is applied to the operation of the fully-connected layer (FL). ) Is input to the output nodes (o1, o2).

The output layer OL may be composed of two output nodes (or sigmoid: o1, o2). Each of the first and second output nodes o1 and o2 corresponds to a probability of whether or not a pacemaker capable of performing MRI imaging. That is, the output value, which is the output of each of the output nodes o1 and o2, is a probability value indicating whether the pacemaker of the radiographic image is a pacemaker capable of MRI imaging. For example, the first output node o1 corresponds to the probability that the pacemaker of the radiographic image is a pacemaker capable of MRI imaging, and the first output value, which is the output of the first output node o1, is a pacemaker in which the pacemaker of the radiographic image can perform MRI imaging. Represents the probability of being For example, if the first output value, which is the output of the first output node o1, is 0.91, it indicates that the probability that the pacemaker of the radiographic image is a pacemaker capable of MRI imaging is 91%. As another example, the second output node o2 corresponds to the probability that the pacemaker of the radiographic image is a pacemaker in which MRI is not possible, and the second output value, which is the output of the second output node o2, is the pacemaker of the radiographic image. It represents the probability of an impossible pacemaker. For example, if the first output value, which is the output of the second output node o2, is 0.09, the probability that the pacemaker of the radiographic image is a pacemaker in which MRI imaging is impossible is 9%.

Each of the plurality of layers (IL, CL, PL, FL, OL) includes a plurality of operations. A weight (w) is applied to each of a plurality of operations of a plurality of layers (IL, CL, PL, FL, OL), and the calculation result to which the weight (w) is applied is transferred to the next layer. That is, the operation result of the previous layer becomes the input of the next layer.

In more detail, the operation of each layer and its weight (w) will be described with the example shown in FIG. 3.

As described above, the elements of the matrix of the input layer IL are pixel units. For each element of the matrix, a pixel value (eg, RGB value) of each pixel of the radiographic image is input as binary data to an element of the matrix of the input layer IL. Then, a convolution operation using each of the plurality of kernels W1 is performed on the input layer IL matrix, and the result of the operation is input to a plurality of feature maps of the first convolutional layer CL1. Here, each of the plurality of kernels K1 may use a matrix having a predetermined size in which an element of the matrix is a weight w. In addition, each of the plurality of feature maps of the first convolutional layer CL1 is a matrix having a predetermined size.

Next, a pooling operation (subsampling) using a plurality of kernels W2 is performed on a plurality of feature maps of the first convolutional layer CL1. The plurality of kernels W2 is also a matrix of a predetermined size, each of which is composed of a weight w. The result of the subsampling operation is input to a plurality of feature maps of the first pooling layer PL1. Each of the plurality of feature maps of the first pooling layer PL1 is also a matrix having a predetermined size.

Subsequently, a convolution operation (convolution) is performed using a kernel W3, which is a matrix of a predetermined size in which each element of the matrix is a weight w for a plurality of feature maps of the first polling layer PL1. A second convolutional layer CL2 made of a map is formed. Next, a second pooling consisting of a plurality of feature maps is performed by performing a subsampling operation using a kernel W4, which is a matrix consisting of a plurality of weights w, for a plurality of feature maps of the second convolutional layer CL2. Configure the layer (PL2). Each of the second pooling layers PL2 is also a matrix having a predetermined size.

Then, a convolution operation (convolution) using a plurality of kernels W5 is performed on a plurality of feature maps of the second polling layer PL2. The plurality of kernels W5 is also a matrix of a predetermined size, whose elements are weights w. A fully connected layer FL is generated according to a result of a convolution operation using a plurality of kernels W5. In other words, a result of a convolution operation using a plurality of kernels K5 is input to a plurality of nodes f1 to fn.

Each of the plurality of nodes (f1 to fn) of the fully connected layer (FL) performs a predetermined operation using a transfer function, etc. for an input from the second polling layer (PL2), and applies a weight (w) to the operation. Input to each node of the output layer (OL). Accordingly, the nodes o1 and o2 of the output layer OL perform a predetermined operation on a value input from the fully connected layer FL, and output an output value as a result. As described above, each of the output nodes o1 and o2 corresponds to whether the pacemaker is a pacemaker capable of MRI imaging, and the output value of each of these output nodes o1 and o2 is a probability value corresponding to whether the pacemaker is a pacemaker capable of MRI imaging. to be.

As described above, each of the plurality of layers of the artificial neural network 110 consists of a plurality of operations, and a weight w is applied to a result of any one operation of any one layer and input to a subsequent layer. For example, a first convolutional layer consisting of a plurality of feature maps (CL1) is performed by performing a convolution operation on a matrix of the input layer IL using a plurality of kernels W1, which is a matrix in which each element is a weight (w). ). Here, it is assumed that the number of kernels K1 is 4 and the size of the 4 kernels K1 is 256 × 256. Then, there are 256 × 256 × 4 weights (w), and since the number of input matrices is one, the calculation of 1 × 256 × 256 × 4 is performed. As another example, each of the plurality of nodes f1 to fn of the fully connected layer FL performs a predetermined operation using a transfer function or the like for an input from the second polling layer PL2, and the operation result is a weight (w ) Is applied and input to each of the plurality of nodes o1 to om of the output layer OL. Here, assuming that the fully connected layer (FL) consists of 256 nodes (f1 to f256) (n=256), and the output layer (OL) consists of two nodes (o1, o2), nodes (f1 to f256) and the weights (w) connecting the nodes (o1 to o2) are 256 × 2, and 256 × 2 operations are performed.

Meanwhile, some of the plurality of kernels of at least one convolutional layer among the plurality of convolutional layers according to an embodiment of the present invention perform a convolution operation on a radiographic image input to the artificial neural network 110 to obtain an external appearance of a pacemaker. Extract. In addition, some of the plurality of kernels of at least one convolutional layer among the plurality of convolutional layers according to an embodiment of the present invention perform a convolution operation on the radiographic image input to the artificial neural network 110 to determine the texture and slope of the pacemaker. , At least one feature of spacing, thickness, and shading may be extracted. According to an embodiment of the present invention, the weight (w) of the artificial neural network 100 is modified to extract at least one of the above-described features of the external shape of the pacemaker, or at least one of the texture, slope, interval, thickness, and shadow of the pacemaker. To carry out learning to do.

When a plurality of images, which are training data, are input to the artificial neural network 110, the artificial neural network 110 will output the calculation result through a plurality of operations to which the weight w as described above is applied.

In addition, the output values of each of the output nodes o1 and o2 of the output layer OL represent a probability value corresponding to the corresponding node of the input learning data. The learning data is data that has already been known whether the target value, that is, the pacemaker of the radiographic image is a pacemaker capable of MRI imaging or a pacemaker that cannot be MRI imaging. Therefore, when the learning module 130 inputs the radiographic image, which is the training data, to the artificial neural network 110, the weight w of the artificial neural network 110 is adjusted so that the difference between the output value of the artificial neural network 110 and the target value is minimized. It needs to be calibrated.

Since the output value is a probability value, the learning module 130 needs to correct the weight w of the artificial neural network 110 so that the output value corresponding to the target value among the output values of the output nodes o1 and o2 has the highest probability value. However, the artificial neural network 110 that has not been sufficiently learned has a difference between the output value and the target value. Therefore, the learning module 130 calculates the weight (w) of the artificial neural network 110 through a back-propagation algorithm so that the difference between the target value and the output value is minimized whenever a plurality of images, which are training data, are input. Perform learning to modify. In this way, in the deep learning according to an embodiment of the present invention, a target value corresponding to the input learning data is determined, and the weight applied to the calculation of the artificial neural network 110 so that the difference between the output value and the target value is minimized. Correct (w).

The learning module 130 repeatedly performs the above-described deep learning procedure using a plurality of different learning data until it is determined that the artificial neural network 110 is sufficiently learned. Here, the learning module 130 determines that the artificial neural network 110 has sufficiently learned if the output value does not change even when some learning data (multiple images) is input while the difference between the target value and the output value is less than a predetermined value. I can.

Next, a method for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention will be described. 4 is a flowchart illustrating a method for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention.

Referring to FIG. 4, the image classifying unit 100 receives a radiographic image using the artificial neural network 110 in step S110 and determines whether a subject to be examined wears a pacemaker capable of MRI imaging.

In addition, the information verification unit 200 receives medical information of the subject to be examined in step S120, and determines whether the subject is wearing a pacemaker capable of MRI imaging. The information verification unit 200 may include an artificial neural network for classifying text input.

When the imaging determination unit 300 determines that both the image classifying unit 100 and the information verification unit 200 have worn a pacemaker capable of MRI imaging in step S130, the subject is finally determined that the subject is capable of MRI imaging. do.

Before performing the above-described step S110, the artificial neural network 110 must be trained. This learning method will be described. 5 is a flowchart illustrating a method of learning an artificial neural network according to an embodiment of the present invention. 6 is a view for explaining a method of learning an artificial neural network according to an embodiment of the present invention.

5, the learning module 130 receives learning data in step S210. The learning data is a radiographic image including a pacemaker that is known whether MRI is possible. Accordingly, a radiographic image is input, and whether MRI imaging is possible is set as a target value.

Next, the preprocessing module 120 performs preprocessing on the radiographic image, which is the training data, in step S220. According to an embodiment, the pre-processing module 120 may perform pre-processing of removing portions representing bones other than the pacemaker (PM) in the radiographic image of FIG. 6. According to another embodiment, the preprocessing module 120 may perform preprocessing of emphasizing the pacemaker PM by adjusting the value of the histogram of the radiographic image of FIG. 6.

Next, the artificial neural network 110 calculates an output value by performing a plurality of calculations to which a weight w is applied to the radiographic image, which is the training data pre-processed in step S230, and outputs the calculated output value.

The learning module 130 modifies the weight w of the artificial neural network 110 through a back-propagation algorithm so that the difference between the output value of the artificial neural network 110 and the target value is minimized in step S240.

Next, the learning module 130 determines whether or not learning of the artificial neural network 110 is completed in step S250. The learning module 130 determines that the learning of the artificial neural network 110 is completed if the difference between the target value and the output value is less than a predetermined value and there is no change in the output value even when different learning data is input. As a result of the determination in step S250, if it is determined that the learning has not been completed, steps S210 to S250 are repeated, and if it is determined that the learning has been completed, the learning process is terminated in step S260.

Next, a method for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention will be described. 7 is a flowchart illustrating a method for classifying a pacemaker using an artificial neural network according to an embodiment of the present invention. To emphasize, FIG. 7 is a flow chart for explaining in more detail step S110 of FIG. 4 described above.

Referring to FIG. 7, the classification module 140 receives actual data in step S310. The actual data is a radiographic image taken by a subject to be examined wearing a pacemaker, and it is not known whether the pacemaker included in the radiographic image can be MRI.

Next, the preprocessing module 120 performs preprocessing on the radiographic image, which is actual data, in step S320. As described above, the preprocessing module 120 may perform preprocessing of removing portions representing bones other than the pacemaker (PM) from the radiographic image. According to another embodiment, the pre-processing module 120 may perform pre-processing of emphasizing the pacemaker (PM) by adjusting the value of the radiographic image histogram.

Subsequently, the artificial neural network 110 calculates an output value by calculating an output value by performing a plurality of calculations to which a weight w is applied to the radiographic image, which is the training data preprocessed in step S330, and outputs the calculated output value.

In step S340, the classification module 140 determines whether the pacemaker worn by the test subject is a pacemaker capable of MRI imaging according to the output value of the artificial neural network 110. As an example, referring to FIG. 3, it is assumed that the output of the artificial neural network 110 has a first output value corresponding to a pacemaker capable of MRI imaging is 0.21, and a second output value corresponding to a pacemaker incapable of MRI imaging is 0.79. In this case, the probability that the pacemaker of the radiographic image is a pacemaker capable of MRI imaging is 21%, and the probability that the pacemaker of the radiographic image is not capable of MRI is 79%. Accordingly, the classification module 140 determines that the subject to be examined is wearing a pacemaker in which MRI imaging is impossible.

As another example, referring to FIG. 3, it is assumed that the output of the artificial neural network 110 has a first output value corresponding to a pacemaker capable of MRI imaging is 0.88, and a second output value corresponding to a pacemaker incapable of MRI imaging is 0.12. In this case, the probability that the pacemaker of the radiographic image is a pacemaker capable of MRI imaging is 88%, and the probability that the pacemaker of the radiographic image is not capable of MRI is 12%. Accordingly, the classification module 140 determines that the subject to be examined is wearing a pacemaker capable of MRI imaging.

Meanwhile, the method according to the embodiment of the present invention described above may be implemented in the form of a program readable through various computer means and recorded on a computer-readable recording medium. Here, the recording medium may include a program command, a data file, a data structure, or the like alone or in combination. The program instructions recorded on the recording medium may be specially designed and constructed for the present invention, or may be known and usable to those skilled in computer software. For example, the recording medium includes magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic-optical media such as floptical disks ( magneto-optical media), and a hardware device specially configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions may include not only a machine language such as produced by a compiler, but also a high-level language that can be executed by a computer using an interpreter or the like. Such a hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

Although the present invention has been described using several preferred embodiments, these embodiments are illustrative and not limiting. As such, those of ordinary skill in the art to which the present invention pertains will understand that various changes and modifications can be made according to the equivalence theory without departing from the spirit of the present invention and the scope of the rights presented in the appended claims.

100: image division
110: artificial neural network
120: preprocessing module
130: learning module
140: classification module
200: Information Verification Department
300: shooting decision unit

Claims (16)

In the device for classifying a pacemaker,
The output of any one layer includes a plurality of layers constituting the next layer through a plurality of operations to which a weight determining the strength of the connection between layers is applied, and when a radiographic image including a pacemaker is input, the plurality of operations An artificial neural network module configured to output a probability of whether the pacemaker is a pacemaker capable of MRI imaging as an output value according to the calculation; And
After learning of the artificial neural network module is completed, a radiographic image of the subject to be examined is input to the artificial neural network module to derive the output value, and according to the derived output value, it is determined whether the pacemaker worn by the subject to be examined is a pacemaker capable of MRI imaging. A device for distinguishing a pacemaker, comprising: a classification module. The method of claim 1,
Before the learning of the artificial neural network module is completed, the difference between the target value corresponding to the learning data and the output value of the artificial neural network module according to the learning data by inputting a radiographic image including a pacemaker as learning data into the artificial neural network module is minimal. And a learning module for correcting the weight of the artificial neural network module through a despreading algorithm to be. The method of claim 1,
The artificial neural network module
It includes at least one convolutional layer including a plurality of kernels,
Wherein the kernel extracts features of an external shape of a pacemaker from the radiographic image. The method of claim 1,
The artificial neural network module
It includes at least one convolutional layer including a plurality of kernels,
The kernel extracts at least one of a texture, a slope, a gap, a thickness, and a shadow of the pacemaker from the radiographic image. The method of claim 1,
And a pre-processing module for removing a portion representing bones other than a pacemaker from the radiographic image before the radiographic image is input to the artificial neural network module. The method of claim 1,
And a pre-processing module for emphasizing the pacemaker by adjusting a value of the histogram before the radiographic image is input to the artificial neural network module. In the device for classifying a pacemaker,
Calculate the probability of whether the pacemaker is a pacemaker capable of MRI imaging by receiving a radiographic image including a pacemaker of the test subject and performing a plurality of calculations in which weights are applied to the radiographic image through an artificial neural network, Accordingly, an image classifying unit for determining whether the subject to be examined is wearing a pacemaker capable of MRI imaging;
An information verification unit that receives medical information of a subject to be examined and determines whether the subject is wearing a pacemaker capable of MRI imaging; And
And an imaging decision unit configured to determine that the subject is capable of MRI imaging when it is determined that both the image classifying unit and the information verifying unit wear a pacemaker capable of MRI imaging, and finally determining that the subject is capable of MRI imaging. Device for. In the method for classifying a pacemaker,
Receiving, by the artificial neural network, a radiographic image of a subject to be examined;
Performing a plurality of calculations to which weights are applied to the radiographic image and outputting a probability of whether the pacemaker is a pacemaker capable of MRI imaging as an output value according to the calculation; And
And determining, by the classification module, whether the pacemaker worn by the subject to be examined is a pacemaker capable of MRI imaging according to the derived output value. The method of claim 8,
Before the outputting step,
Inputting, by a learning module, a radiographic image including a pacemaker, which is training data, into the artificial neural network;
Outputting, by the artificial neural network, an output value corresponding to the training data by performing a plurality of calculations to which weights are applied;
And modifying, by the learning module, a weight of the artificial neural network through a despreading algorithm so that a difference between a target value corresponding to the training data and an output value of the artificial neural network according to the training data is minimized. How to distinguish between heart rate and pacemakers. The method of claim 8,
The step of outputting as an output value according to the operation is
The artificial neural network
And extracting features of the external shape of the pacemaker from the radiographic image through at least one convolution layer including a plurality of kernels. The method of claim 8,
The step of outputting as an output value according to the operation is
And extracting, by the artificial neural network, at least one of the texture, slope, interval, thickness, and shadow of the pacemaker from the radiographic image through at least one convolution layer including a plurality of kernels. How to identify pacemakers. The method of claim 8,
Before the step of receiving the input,
And removing a portion representing bones other than the pacemaker from the radiographic image. The method of claim 8,
Before the step of receiving the input,
And enhancing the pacemaker by adjusting the value of the histogram in the radiographic image. In the method for classifying a pacemaker,
Determining whether the subject is wearing a pacemaker capable of MRI imaging by receiving a radiographic image of the subject to be examined by the image classifying unit;
An information verification unit receiving medical information of the subject to be examined and determining whether the subject is wearing a pacemaker capable of MRI imaging; And
And finally determining that the subject is capable of MRI imaging when the imaging determination unit determines that both the image classifying unit and the information verification unit have worn a pacemaker capable of MRI imaging, and finally determining that the subject is capable of MRI imaging. The way to differentiate. In a computer-readable recording medium in which a program for performing a method for distinguishing a pacemaker is recorded,
The learning module includes a plurality of layers constituting the next layer through a plurality of operations in which the output of any one layer is applied with a weight that determines the strength of the connection between the layers, and the pacemaker performs the MRI by performing the plurality of operations. Learning an artificial neural network that outputs a probability of whether a cardiac pacemaker can be photographed as an output value according to the calculation;
Deriving the output value by inputting, by a classification module, a radiographic image of a subject to be examined into the artificial neural network; And
The identification module, the step of determining whether the pacemaker worn by the subject to be examined is a pacemaker capable of MRI imaging according to the derived output value; a computer-reading program in which a program for performing a method for distinguishing a pacemaker comprising: Recordable media available. The method of claim 15,
Learning the artificial neural network
The learning module
Inputting a radiographic image including a pacemaker, which is training data, into the artificial neural network; And
When the artificial neural network outputs an output value through a plurality of operations corresponding to the training data, the weight of the artificial neural network is modified through an inverse diffusion algorithm so that the difference between the target value corresponding to the training data and the output value is minimized. A computer-readable recording medium on which a program for performing a method for discriminating a pacemaker, comprising a step; is recorded thereon.

Images