JP2020062218A - Learning device, estimation device, learning method, estimation method and program - Google Patents

Abstract

To provide a learning device, an estimation device, a learning method, an estimation method and a program which can improve the prediction accuracy on whether or not a bile duct cannulation is difficult.SOLUTION: A learning device includes: an acquisition unit which acquires a plurality of pieces of information in which an image of a Vater's papilla of a duodenum of a bile duct is associated with information indicating a cannulation method being the method of inserting a catheter into the bile duct; a learning unit which machine-learns the information indicating the cannulation method as teacher data on the basis of the image of the Vater's papilla of the duodenum of the bile duct; and a storage unit which stores the result of the machine-learning of the learning unit in association with the information indicating the cannulation method.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a learning device, an estimation device, a learning method, an estimation method, and a program.

Endoscopic retrograde cholangiopancreatography (ERCP) is a procedure in which a special endoscope is inserted from the mouth and advanced to the duodenum, and a thin tube is inserted from the outlet (papillary) of the bile duct / pancreatic duct to give an image. It is an inspection / treatment in which a drug is injected for inspection, stones are removed, and a stent is inserted.
ERCP-related procedures begin with cannulation of inserting a catheter into the bile duct. Cannulation can be said to be the first step that requires the most attention, because stumbling with cannula leads to prolongation of procedure time and occurrence of complications.
Although normal bile duct cannulation is performed by basic methods such as an imaging method and a wire guide method, sometimes there are cases in which bile duct cannulation is difficult. For cases where bile duct cannulation is difficult, a rescue method such as the pancreatic duct guide wire method or precut method, which has a higher degree of difficulty, is used after the basic method (basic method).

It is generally said that the rescue method for cases where cannulation is difficult is more likely to cause an accident such as post-ERCP pancreatitis than the basic method. The reasons for this are considered to be (1) and (2) below.
(1) The rescue technique itself causes an accident.
(2) When cannulation is difficult and the rescue method is reached, since the basic method has already been performed for a certain period of time, the total procedure time becomes long and an accident occurs.
If it is possible to predict whether cannulation is difficult or not, it is reported that by giving up the basic method in a short time and performing the rescue method early, it is possible to prevent the occurrence of accidents.
Factors that make bile duct cannulation difficult include the shape of the duodenal fatta papilla at the bile duct opening and the position and orientation of the duodenal fatta papilla on the endoscopic screen. Therefore, it is possible to predict whether biliary canalization will be difficult by evaluating the shape, position, and orientation of the papilla. Heretofore, based on the prediction of bile duct cannulation, rescue methods have been performed early to prevent the occurrence of accidents due to a long procedure time.
As a technique for determining the state of a subject, a technique for determining the physiological state of the subject's mind or body is known (see, for example, Patent Document 1). This technology detects "specific operation information" corresponding to a specific operation of the inspection device, acquires "face change information" indicating a time-series change of the subject's face data, and obtains face change information by singular value decomposition, ... is decomposed into a plurality of components 1, 2, 3, ... By principal component analysis or independent component analysis, and from the plurality of components 1, 2, 3 ,. , And the uncomfortable state of the subject is determined based on the determination component.

JP, 2018-94032, A

Predicting cases in which bile duct cannulation is difficult by evaluating the shape of the duodenal fatta papilla at the bile duct opening and the position and orientation of the duodenal fatta papilla on the endoscopic screen depends on the experience value of the operator. Is extremely large, and it is difficult to make the same prediction between and within the surgeons.
It is an object of the present invention to provide a learning device, an estimation device, a learning method, an estimation method, and a program capable of improving the prediction accuracy of whether or not bile duct cannulation is difficult.

  One embodiment of the present invention, an image of a bile duct duodenal Fata papilla, an acquisition unit that acquires a plurality of pieces of information that associates information indicating a cannulation method that is a method of inserting a catheter into the bile duct, and a duodenal Fata papilla of the bile duct. A learning unit for machine learning based on the image indicating the cannulation method as teacher data, and a storage unit for storing the result of the machine learning performed by the learning unit and the information indicating the cannulation method in association with each other. And a learning device.

  Further, in the learning device according to one aspect of the present invention, each of the images of the duodenal Fata papillae of the plurality of bile ducts is different in at least one of the orientation and the position of the duodenal Fata papillae or both.

  Further, in the learning device according to one aspect of the present invention, each of the images of the duodenal Fata papillae of a plurality of bile ducts has a different shape of the opening of the bile duct.

  Further, in the learning device according to one aspect of the present invention, the acquisition unit, in association with the image of the duodenal Fata papilla of the bile duct and information indicating the cannulation method, further to the bile duct after the image is obtained. Obtaining information associated with the cannulation time, which is the time until the insertion of the catheter is completed, the learning unit, based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method and the cannula. Machine learning using the duration time as teacher data.

  Further, according to one aspect of the present invention, an acquisition unit that acquires an image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown, and an image of the duodenal Fata papilla of the bile duct based on the information indicating the cannulation method is taught. The result of machine learning as data, using the information associated with the information indicating the cannulation method, based on the image of the duodenal Fata papilla of the bile duct in which the cannulation method acquired by the acquisition unit is unknown, An estimation device including an estimation unit that estimates a method of inserting a catheter into a bile duct shown in the image.

  One aspect of the present invention, an image of the bile duct duodenal Fata papilla, a step of acquiring a plurality of information associated with information indicating a cannulation method that is a method of inserting a catheter into the bile duct, the bile duct duodenal Fata papilla Machine learning using information indicating the cannulation method as teacher data based on the image, and storing the result of machine learning in the machine learning step and information indicating the cannulation method in association with each other. Is a learning method executed by a learning device.

  One aspect of the present invention, a step of obtaining an image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown, and based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method is machined as teacher data. Based on the image of the bile duct duodenal fatta papilla of which the cannulation method obtained in the obtaining step is unknown, using the information associated with the learning result and the information indicating the cannulation method, the image And a step of estimating a method of inserting a catheter into the bile duct shown in FIG.

  One aspect of the present invention, a computer of the learning device, a step of acquiring a plurality of information associated with an image of the duodenal Fata papilla of the bile duct and information indicating a cannulation method that is a method of inserting a catheter into the bile duct, Based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method as teacher data, a machine learning step, and a machine learning result in the machine learning step and information indicating the cannulation method. It is a program that executes the step of storing in association with each other.

  One aspect of the present invention, the computer of the estimation device, the step of acquiring an image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown, and the cannulation method based on the image of the duodenal Fata papilla of the bile duct The image obtained by machine learning the information as teacher data, and using the information associated with the information indicating the cannulation method, the image of the bile duct duodenal fatta papillae in which the cannulation method acquired in the step of acquiring is unknown. And a step of estimating a method for inserting a catheter into the bile duct shown in the image based on the above.

  According to the present invention, it is possible to provide a learning device, an estimation device, a learning method, an estimation method, and a program capable of improving the prediction accuracy of whether or not bile duct cannulation is difficult.

It is a figure which shows an example of the image of the duodenal Fata papilla of a bile duct. It is a figure which shows an example of the hardware constitutions of the learning apparatus which concerns on embodiment of this invention. It is a figure which shows an example of a cannulation method estimation table. It is a functional block diagram of a learning device concerning an embodiment of the present invention. It is a flowchart which shows an example of operation | movement of the learning apparatus which concerns on embodiment of this invention. It is a figure which shows an example of the hardware constitutions of the learning device which concerns on the modification of this invention. It is a figure which shows an example of a cannulation method estimation table. It is a functional block diagram of the learning device which concerns on a modification. It is a flowchart which shows an example of operation | movement of the learning apparatus which concerns on a modification. It is a figure which shows an example of the hardware constitutions of the estimation apparatus which concerns on embodiment of this invention. It is a functional block diagram of the estimation device concerning the embodiment of the present invention. It is a flow chart which shows an example of operation of the estimating device concerning the embodiment of the present invention. It is a figure which shows an example of the hardware constitutions of the estimation apparatus which concerns on a modification. It is a functional block diagram of the estimation apparatus which concerns on a modification. It is a flowchart which shows an example of operation | movement of the estimation apparatus which concerns on a modification.

Hereinafter, a learning device, an estimation device, a learning method, an estimation method, and a program according to the embodiments will be described with reference to the drawings. The embodiments described below are merely examples, and the embodiments to which the present invention is applied are not limited to the following embodiments.
In all the drawings for explaining the embodiments, those having the same function are designated by the same reference numerals, and the repeated description will be omitted.
The term “based on XX” as used herein means “based on at least XX” and includes the case based on another element in addition to XX. In addition, “based on XX” is not limited to the case where XX is directly used, but also includes the case where it is based on what is calculated or processed for XX. “XX” is an arbitrary element (for example, arbitrary information).

(Embodiment)
(Learning device)
A learning device according to the embodiment will be described. The learning device according to the embodiment acquires a plurality of pieces of information in which information indicating an image of the duodenal vertebral papilla of the bile duct and information indicating a cannulation method that is a method of inserting a catheter into the bile duct are associated with each other.
The image of the duodenal Fata papilla of the bile duct is an image obtained by observing the duodenal Fata papilla of the bile duct with an endoscope when performing bile duct cannulation. More preferably, the image of the bile duct duodenum Fata papilla is a front view image obtained by observing the duodenal Fata papilla of the bile duct with an endoscope immediately before performing bile duct cannulation.
Each of the images of the duodenal Fata papillae of the plurality of bile ducts is different in at least one of the orientation and the position of the duodenal Fata papillae or both. Moreover, the shape of the opening of the bile duct is different in each of the images of the duodenal Fata papillae of the plurality of bile ducts. The shape of the bile duct opening is divided into, for example, a separable type, a septum type, and a common duct type, and the separable type is further divided into a separate open type and an onion type.

The cannulation method is a method in which a catheter is inserted into the bile duct based on the obtained image of the duodenal Fata papilla of the bile duct. Specifically, the cannulation method includes a contrast method, a wire guide method, a pancreatic duct guide wire method, and a precut method.
The contrast method is an intubation technique using a contrast catheter. The intubation technique using a contrasting catheter is classified into three methods, that is, a proximity method, a distant shadow (look-up) method, and an intermediate method positioned in between, depending on the position of the duodenal papilla observed with an endoscope. In many cases, selective bile duct intubation by the intermediate method to the proximity method, which mainly involves endoscopic operation rather than catheter operation, is tried.
The wire guide method is a method of selectively attempting intubation of the bile duct after observing a guide wire in a tapered catheter instead of performing test injection of a contrast agent.
In the pancreatic duct guide wire method, the pancreatic duct, pancreatic duct mouth, and bile duct mouth are linearized with a guide wire. Push down the pancreatic duct mouth with a guide wire. This is often used especially in cases where the nipple is highly mobile, cases where it is difficult to view the nipple frontally, and cases where the bile duct stenosis is long and bent.
The precut method is a method of directly incising the mucous membrane of the duodenal papilla and opening the bile duct mouth for deep intubation in cases where it is difficult to intubate the bile duct for various reasons. The precut includes a method using a needle (needle) knife and a method using a normal sphincterotome.
The learning device according to the embodiment, based on the image of the bile duct duodenal Fata papilla, machine learning information indicating the cannulation method as teacher data, the feature amount of the image of the bile duct duodenal Fata papilla obtained by machine learning. And the information indicating the cannulation method. The learning device according to the embodiment stores the feature amount of the image of the duodenal vertebral papilla of the bile duct and the information indicating the cannulation method in association with each other.

FIG. 1 is a diagram showing an example of an image of a duodenal Fata papilla of a bile duct. FIG. 1 shows a front view of the duodenal Fata papilla of the bile duct. The bile duct is seen in a frontal view of the duodenal Fata papillae.
Here, the first step of the ERCP-related procedure will be described.
In the first step, a catheter is inserted into the bile duct. In the first step, a side endoscope including a camera on the side surface of the endoscope is used.
First, the duodenal Fata papilla of the bile duct is viewed from the front by operating the operation unit of the endoscope. That is, the catheter is directed toward the bile duct. Here, the endoscope is moved up and down, the left and right angles are adjusted, and the standing forceps are adjusted.
Next, advance the catheter by predicting the direction of the bile duct. A contrast agent is injected and the bile duct is imaged. A guide wire is inserted after imaging the bile duct.

(Hardware configuration of learning device)
FIG. 2 is a diagram showing an example of a hardware configuration of the learning device according to the embodiment of the present invention. The learning device 100 will be described.
(Hardware configuration of learning device)
The learning device 100 includes a CPU 102, a memory 104, a non-volatile memory 108, a communication unit 106, and an internal bus 150.
The CPU 102 executes, for example, the program 1082 and the application 1084 stored in the non-volatile memory 108, and uses the memory 104 as a work memory to control each unit of the learning device 100. The memory 104 is composed of a RAM such as a volatile memory using a semiconductor element, and is used as a work memory of the CPU 102.
The non-volatile memory 108 is composed of, for example, a flash memory or the like. The non-volatile memory 108 stores a program 1082 executed by the CPU 102, an application 1084, image information 1086, cannula method information 1087, and a cannula method estimation table 1088.
The application 1084 causes the learning apparatus 100 to acquire a plurality of pieces of information in which an image of the duodenal vertebral papilla of the bile duct and information indicating a cannulation method that is a method of inserting a catheter into the bile duct are associated with each other.
The application 1084 causes the learning device 100 to use information indicating the cannulation method as teacher data based on a plurality of pieces of information in which the acquired image of the duodenal vertebral papilla of the bile duct and information indicating the cannulation method are associated with each other. Let them learn.
The application 1084 stores the result of machine learning and the information indicating the cannulation method in association with each other.
The image information 1086 stores a plurality of images of the duodenal Fata papilla of the bile duct.
The cannulation method information 1087 is associated with each of the plurality of images of the duodenal Fata papilla of the bile duct included in the image information 1086 to indicate any one of an imaging method, a wire guide method, a pancreatic duct guide wire method, and a precut method. Remember.

(Cannulation method estimation table)
FIG. 3 is a diagram showing an example of the cannulation method estimation table. The cannulation method estimation table 1088 is data in a table format in which the information indicating the feature amount of the image of the duodenal Fata papilla of the bile duct and the information indicating the cannulation method are associated with each other. The feature amount of the image of the bile duct duodenum Fata papilla is obtained by machine learning such as deep learning on a plurality of images of the bile duct duodenum Fata papilla classified for each cannulation method.
In the example shown in FIG. 3, the feature amount “AAA” of the image of the duodenal Fata papilla of the bile duct is associated with the cannulation method “imaging method”. The feature amount “BBB” of the image of the duodenal Fata papilla of the bile duct is associated with the cannulation method “wire guide method”. The feature amount “CCC” of the image of the duodenal Fata papilla of the bile duct is associated with the cannulation method “pancreatic duct guide wire method”. The feature amount “DDD” of the image of the duodenal Fata papilla of the bile duct is associated with the cannulation method “precut method”.

  The communication unit 106 communicates with an external device such as the terminal device 10 connected to the network 50 to send and receive data. Specifically, the communication unit 106 is connected to the network 50 by wire. In addition, the communication unit 106 may be configured by a wireless device that performs wireless communication using a wireless communication technology such as LTE. The internal bus 150 connects the CPU 102, the memory 104, the non-volatile memory 108, and the communication unit 106 to each other. The respective units connected to the internal bus 150 can exchange data with each other via the internal bus 150.

(Functional structure of learning device)
FIG. 4 is a functional block diagram of the learning device according to the embodiment of the present invention.
In the learning device 100, the receiving unit 152, the acquisition unit 154, the learning unit 156, and the output unit 158 are configured as the reception unit 152, the acquisition unit 154, and the output unit 158 by the CPU 102 executing the program 1082 and the application 1084 expanded from the nonvolatile memory 108 on the memory 104. Function.

(Each function configuration of the learning device)
Each functional configuration of the learning device 100 will be described in detail with reference to FIGS. 2 and 4. In the following, in explaining each function of the learning device 100, a relationship with main hardware for realizing each functional block of the learning device 100 shown in FIG. 4 will also be described.
The reception unit 152 is realized by the command from the CPU 102 and the communication unit 106. The reception unit 152 receives information in which the image of the duodenal Fata papilla of the bile duct is associated with the information indicating the cannulation method. For example, the terminal device 10 transmits information in which the image of the duodenal Fata papilla of the bile duct is associated with the information indicating the cannulation method.
The communication unit 106 of the learning device 100 receives the image of the bile duct duodenum Fata papilla transmitted by the terminal device 10 and the information that associates the information indicating the cannulation method, and the received image of the bile duct duodenum Fata papilla, Information associated with the information indicating the cannulation method is output to the CPU 102.
By functioning as the receiving unit 152, the CPU 102 receives information in which the image of the bile duct duodenal fata papillae output by the communication unit 106 is associated with information indicating the cannulation method. The reception unit 152 stores the information indicating the image of the bile duct duodenal fata papilla included in the information in which the received image of the duodenal fatta papilla of the bile duct is associated with the information indicating the cannulation method, with the image information 1086 of the nonvolatile memory 108. The information indicating the cannula method is stored in the cannula method information 1087 of the nonvolatile memory 108.

The acquisition unit 154 acquires the information indicating the image of the duodenal Fata papilla of the bile duct stored in the image information 1086 of the nonvolatile memory 108, and the acquired information indicating the image of the duodenal Fata papilla of the bile duct is input to the learning unit 156. Output. The acquisition unit 154 also acquires information indicating the cannulation method stored in the cannulation method information 1087 of the non-volatile memory 108, and outputs the acquired information indicating the cannulation method to the learning unit 156. Here, the information indicating the image of the duodenal Fata papilla of the bile duct acquired by the acquisition unit 154 is associated with the information indicating the cannulation method.
The learning unit 156 is realized by an instruction from the CPU 102. The learning unit 156 acquires a plurality of pieces of information indicating the image of the duodenal fata papilla of the bile duct stored in the image information 1086 of the non-volatile memory 108 output by the acquisition unit 154, and stores the information in the cannula method information of the non-volatile memory 108. Acquires a plurality of pieces of information indicating the cannulation method being performed. Here, each of the acquired images of the duodenal Fata papilla of the bile duct and each of the information indicating the cannulation method are associated with each other.
The learning unit 156 machine-learns information indicating the cannulation method as teacher data, based on a plurality of pieces of information in which the acquired image of the duodenal vertebral papilla of the bile duct is associated with the information indicating the cannulation method. The learning unit 156 includes a feature amount of an image of the bile duct duodenum Fata papilla obtained by machine learning a plurality of images of the bile duct duodenum Fata papilla, and a cannulation method performed on the image of the bile duct duodenum Fata papilla. And are output to the output unit 158.

  The output unit 158 is realized by an instruction from the CPU 102. The output unit 158 acquires the feature amount of the image of the bile duct duodenal Fata papillae output by the learning unit 156 and the cannulation method performed on the image of the bile duct duodenal Fata papillae, and acquires the obtained bile duct duodenal Fata papillae. Of the image of the bile duct and the cannulation method performed on the image of the duodenal fata papilla of the bile duct are stored in the cannulation method estimation table 1088 of the non-volatile memory 108 in association with each other.

(Operation of learning device)
FIG. 5: is a flowchart which shows an example of operation | movement of the learning apparatus which concerns on embodiment of this invention.
In FIG. 5, information indicating the image of the bile duct duodenal vertebral papilla of the learning device 100 is stored in the image information 1086 of the non-volatile memory 108, and information indicating the cannulation method is stored in the cannulation method information 1087 of the non-volatile memory 108. The operation after this is described. In other words, the image information 1086 of the non-volatile memory 108 stores the image information of the duodenal fata papilla of the bile duct, and the cannula method information 1087 of the non-volatile memory 108 stores the information indicating the cannula method.
(Step S1)
In the learning device 100, the acquisition unit 154 acquires a plurality of pieces of information indicating the images of the bile duct duodenal Fata papillae stored in the image information 1086 of the nonvolatile memory 108, and acquires the acquired images of the plurality of bile duct duodenal Fata papillae. The information shown is output to the learning unit 156.
(Step S2)
In the learning device 100, the acquisition unit 154 acquires a plurality of pieces of information indicating the cannulation method stored in the cannulation method information 1087 of the non-volatile memory 108, and acquires information indicating the acquired plurality of cannulation methods from the learning unit 100. Output to 156.
(Step S3)
In the learning device 100, the learning unit 156 acquires the image of the bile duct duodenal fata papillae output by the acquisition unit 154 and the information indicating the cannulation method in association with each other. The learning unit 156 machine-learns information indicating the cannulation method as teacher data, based on a plurality of pieces of information in which the acquired image of the duodenal vertebral papilla of the bile duct is associated with the information indicating the cannulation method. The learning unit 156 includes a feature amount of an image of the bile duct duodenum Fata papilla obtained by machine learning a plurality of images of the bile duct duodenum Fata papilla, and a cannulation method performed on the image of the bile duct duodenum Fata papilla. And are output to the output unit 158.

(Step S4)
In the learning device 100, the output unit 158 acquires and acquires the feature amount of the image of the duodenal Fata papilla of the bile duct output by the learning unit 156 and the cannulation method performed on the image of the duodenal Fata papilla of the bile duct. The feature of the image of the duodenal Fata papilla of the bile duct is associated with the cannulation method performed on the image of the duodenal Fata papilla of the bile duct.
(Step S5)
In the learning device 100, the output unit 158 associates information obtained by associating the acquired feature amount of the image of the duodenal Fata papilla of the bile duct with the information indicating the cannulation method performed on the image of the duodenal Fata papilla of the bile duct. The estimation method estimation table 1088 of the property memory 108 is stored.
In the flowchart shown in FIG. 5, the order of step S1 and step S2 may be exchanged.

In the above-described embodiment, the case where the learning device 100 receives the information in which the image information of the duodenal fata papilla of the bile duct and the information indicating the cannulation method are transmitted by one terminal device 10 has been described. It is not limited to this example. For example, the learning device 100 may receive information in which image information of the bile duct duodenal fata papillae transmitted by a plurality of terminal devices and information indicating a cannulation method are associated with each other. With this configuration, the learning device 100 can obtain a large amount of information in which the image information of the duodenal vertebral papilla of the bile duct and the information indicating the cannulation method are associated with each other, so that the effect of machine learning can be enhanced.
In the above-described embodiment, a case has been described in which the learning device 100 acquires a plurality of images of the duodenal Fata papilla of the bile duct and machine-learns the acquired images of the duodenal Fata papilla of the bile duct. However, the present invention is not limited to this. For example, the learning device 100 may acquire a plurality of videos of the bile duct duodenal Fata papillae and machine-learn the acquired videos of the plurality of bile duct duodenal Fata papillae.
In the above-described embodiment, the modeling method, the wire guide method, the pancreatic duct guide wire method, and the precut method are described as examples of the cannulation method, but the present invention is not limited to this example. For example, in addition to the shaping method, the wire guide method, the pancreatic duct guide wire method, and the precut method, different methods such as an ultrasonic endoscope guided rendezvous method may be included.
In the above-described embodiment, each of the shaping method, the wire guide method, the pancreatic duct guide wire method, and the precut method has been described with respect to the case where the feature amount of the image of the duodenal Fata papilla of the bile duct is associated, but in this example. Not limited. For example, each of the basic method including the shaping method and the wire guide method and the rescue method including the pancreatic duct guide wire method and the pre-cut method may be associated with the feature amount of the image of the duodenal Fata papilla of the bile duct.

According to the learning device 100 of the embodiment, the learning device 100 is a method of inserting a catheter into the bile duct using the image of the duodenal Fata papilla of the bile duct and the image of the duodenal Fata papilla of the bile duct. Based on the images of the bile duct duodenal fatta papillae acquired by the acquisition unit 154, the acquisition unit 154 acquiring a plurality of pieces of information associated with the information indicating the information indicating the cannulation method associated with the images. As data, a learning unit 156 that performs machine learning and a storage unit that stores the result of machine learning by the learning unit 156 and information indicating the cannulation method in association with each other are stored. Based on the images of the bile duct duodenum Fata papilla thus acquired, machine learning can be performed by using the information indicating the cannulation method associated with the images as teacher data, and thus the feature amount of the image of the bile duct duodenum Fata papilla And the cannulation method performed on the image of the duodenal Fata papilla of the bile duct. Predicting whether bile duct cannulation is difficult by using information that correlates the features of the image of the bile duct duodenal fatta with the method of cannulation performed on the image of the bile duct duodenal fatta The accuracy can be improved.
Further, in each of the images of the duodenal fatta papillae of the plurality of bile ducts, at least one of the orientation and the position of the duodenal fatta papillae or both of them is different. With this configuration, it is possible to enhance the learning effect of the image of the duodenal Fata papilla of the bile duct.
Furthermore, each of the images of the duodenal Fata papillae of multiple bile ducts has a different shape of the bile duct opening. With this configuration, it is possible to enhance the learning effect of the image of the duodenal Fata papilla of the bile duct.

(Modification)
A learning device 100a according to a modification will be described. The learning device 100a according to the modification, in addition to the information indicating the image of the duodenal Fata papilla of the bile duct and the information indicating the cannulation method, after the image of the duodenal Fata papilla of the bile duct is obtained, insertion of the catheter into the bile duct is not performed. Acquire multiple pieces of cannulation time information, which is the time until the end.
The learning device 100a according to the modification machine-learns information indicating a cannulation method and cannulation time information as teacher data on the basis of an image of the duodenal Fata papilla of the bile duct, and the duodenum of the bile duct obtained by the machine learning. The feature amount of the image of the papilla of Vater is associated with the information indicating the cannulation method and the cannulation time information. The learning device according to the embodiment stores the feature amount of the image of the duodenal fata papilla of the bile duct, the information indicating the cannulation method, and the cannulation time information in association with each other.

(Hardware configuration of learning device)
FIG. 6 is a diagram showing an example of a hardware configuration of a learning device according to a modification of the present invention. The learning device 100a will be described.
(Hardware configuration of learning device)
The learning device 100a includes a CPU 102, a memory 104, a non-volatile memory 108a, a communication unit 106, and an internal bus 150.
The CPU 102 executes, for example, the program 1082 and the application 1084a stored in the nonvolatile memory 108a, and uses the memory 104 as a work memory to control each unit of the learning device 100a. The memory 104 is composed of a RAM such as a volatile memory using a semiconductor element, and is used as a work memory of the CPU 102.
The non-volatile memory 108a is composed of, for example, a flash memory or the like. A program 1082 executed by the CPU 102, an application 1084a, image information 1086, cannula method information 1087, a cannula method estimation table 1088a, and cannula time information 1089 are stored in the non-volatile memory 108a. .
The application 1084a causes the learning device 100a to acquire a plurality of pieces of information in which the image of the duodenal vertebral papilla of the bile duct, the information indicating the cannulation method, and the cannulation time information are associated with each other.
The application 1084a uses the learning device 100a to acquire the information indicating the cannulation method based on a plurality of pieces of information obtained by associating the acquired image of the duodenal fatta papilla of the bile duct, the information indicating the cannulation method, and the cannulation time information. Machine learning is performed by using the and cannulation time information as teacher data.
The application 1084a stores the result of machine learning, the information indicating the cannulation method, and the cannulation time information in association with each other.

(Cannulation method estimation table)
FIG. 7 is a diagram showing an example of the cannulation method estimation table. The cannulation method estimation table 1088a is data in a table format in which information indicating the feature amount of the image of the duodenal vertebral papilla of the bile duct, information indicating the cannulation method, and cannulation time information are associated with each other. The image feature amount of the bile duct duodenum Fata papilla is obtained by performing machine learning such as deep learning on a plurality of images of the bile duct duodenum Fata papilla classified for each cannula method and cannula time information.
In the example shown in FIG. 7, the feature amount “AAA” of the image of the duodenal Fata papilla of the bile duct, the cannulation method “imaging method”, and the cannulation time information “T1” are associated with each other. The feature amount “A1A1A1” of the image of the duodenal Fata papilla of the bile duct, the cannulation method “imaging method”, and the cannulation time information “T2” are associated with each other. The feature amount "BBB" of the image of the duodenal Fata papilla of the bile duct, the cannulation method "wire guide method", and the cannulation time information "T3" are associated with each other. The feature amount “B1B1B1” of the image of the bile duct duodenal vertebral papilla, the cannulation method “wire guide method”, and the cannulation time information “T4” are associated with each other.
The feature amount “CCC” of the image of the bile duct duodenal vertebral papilla is associated with the cannulation method “pancreatic duct guidewire method” and the cannulation time information “T5”. The feature amount “C1C1C1” of the image of the duodenal vertebral papilla of the bile duct, the cannulation method “pancreatic duct guide wire method”, and the cannulation time information “T6” are associated with each other. The feature amount “DDD” of the image of the duodenal Fata papilla of the bile duct, the cannulation method “precut method”, and the cannulation time information “T7” are associated with each other. The feature amount “D1D1D1” of the image of the duodenal vertebral papilla of the bile duct, the cannulation method “precut method”, and the cannulation time information “T8” are associated with each other.
The cannulation time information 1089 stores the cannulation time information in association with each of the plurality of images of the duodenal Fata papilla of the bile duct included in the image information 1086. As described above, the cannulation time information is the time from the acquisition of the image of the duodenal fatta papilla of the bile duct to the completion of insertion of the catheter into the bile duct.

(Functional structure of learning device)
FIG. 8 is a functional block diagram of a learning device according to the modification.
The learning device 100a functions as the reception unit 152a, the acquisition unit 154a, the learning unit 156a, and the output unit 158a by the CPU 102 executing the program 1082 and the application 1084a expanded from the nonvolatile memory 108a on the memory 104.

(Each function configuration of the learning device)
Each functional configuration of the learning device 100a will be described in detail with reference to FIGS. 6 and 8. In the following, in explaining each function of the learning device 100a, a relationship with main hardware for realizing each functional block of the learning device 100a shown in FIG. 8 will also be described.
The reception unit 152a is realized by the command from the CPU 102 and the communication unit 106. The reception unit 152a receives information in which the image of the duodenal fata papilla of the bile duct, the information indicating the cannulation method, and the cannulation time information are associated with each other. For example, the terminal device 10 transmits information in which the image of the duodenal fata papilla of the bile duct, the information indicating the cannulation method, and the cannulation time information are associated with each other.
The communication unit 106 of the learning device 100 receives the image of the bile duct duodenal fatta papilla transmitted by the terminal device 10, the information showing the information indicating the cannulation method, and the cannulation time information in association with each other, and the received bile duct duodenal fatta papilla. The information indicating the image, the information indicating the cannulation method, and the cannulation time information is output to the CPU 102.
By functioning as the reception unit 152a, the CPU 102 receives the information output by the communication unit 106 that associates the image of the duodenal Fata papilla of the bile duct, the information indicating the cannulation method, and the cannulation time information. The reception unit 152a stores the information indicating the image of the duodenal fata papilla of the bile duct included in the received image of the duodenal fatta papilla of the bile duct, the information indicating the cannulation method and the information relating to the cannulation time in the nonvolatile memory 108. The image information 1086 is stored, the information indicating the cannulation method is stored in the cannulation method information 1087 of the nonvolatile memory 108, and the cannulation time information is stored in the cannulation time information 1089 of the nonvolatile memory 108.

The acquisition unit 154a acquires information indicating an image of the bile duct duodenal Fata papilla stored in the image information 1086 of the non-volatile memory 108a, and outputs the acquired information indicating the image of the duodenal Fata papilla of the bile duct to the learning unit 156a. Output. In addition, the acquisition unit 154a acquires information indicating the cannulation method stored in the cannulation method information 1087 of the nonvolatile memory 108a, and outputs the acquired information indicating the cannulation method to the learning unit 156a.
Further, the acquisition unit 154a acquires the cannulation time information stored in the cannulation time information 1089 of the nonvolatile memory 108a, and outputs the acquired cannulation time information to the learning unit 156a. Here, the information indicating the image of the duodenal fatter papilla of the bile duct acquired by the acquisition unit 154, the information indicating the cannulation method, and the cannulation time information are associated with each other.

The learning unit 156a is realized by an instruction from the CPU 102. The learning unit 156a acquires a plurality of pieces of information indicating the image of the duodenal fata papilla of the bile duct, which is stored in the image information 1086 of the non-volatile memory 108a output by the acquisition unit 154a, and stores it in the cannula method information 1087 of the non-volatile memory 108a. A plurality of pieces of information indicating the stored cannula method are acquired, and a plurality of pieces of cannula time information stored in the cannula time information 1089 of the nonvolatile memory 108a are acquired. Here, each of the acquired images of the duodenal Fata papilla of the bile duct, each piece of information indicating the cannulation method, and each piece of cannulation time information are associated with each other.
The learning unit 156a obtains the information indicating the cannulation method and the cannulation time information based on a plurality of pieces of information obtained by associating the acquired image of the duodenal furta papilla of the bile duct with the information indicating the cannulation method and the cannulation time information. Machine learning is used as teacher data. The learning unit 156a includes a feature amount of an image of the bile duct duodenal phater papilla obtained by machine learning of a plurality of images of the bile duct duodenal phater papilla, and a cannulation method performed on the image of the bile duct duodenal phater papilla. And the cannulation time information from the time when the image of the duodenal fatter papilla of the bile duct is obtained to the time when the insertion of the catheter into the bile duct is completed is output to the output unit 158a.

  The output unit 158a is realized by a command from the CPU 102. The output unit 158a ends the insertion of the catheter into the bile duct and the feature amount of the image of the duodenal Fata papilla of the bile duct output by the learning unit 156a, the information indicating the cannulation method performed in the image of the duodenal Fata papilla of the bile duct. Until the end of the insertion of the catheter into the bile duct and the cannulation method that was performed on the acquired image of the duodenal Fata papilla of the bile duct. It is stored in the cannula method estimation table 1088a of the non-volatile memory 108a in association with the cannula time information.

(Operation of learning device)
FIG. 9 is a flowchart showing an example of the operation of the learning device according to the modification.
In FIG. 9, the information indicating the image of the duodenal furta papilla of the bile duct of the learning device 100a is stored in the image information 1086 of the non-volatile memory 108a, and the information indicating the cannula method is stored in the cannula method information 1087 of the non-volatile memory 108a. Then, the operation after the cannula time information is stored in the cannula time information 1089 of the nonvolatile memory 108a will be described. In other words, the image information 1086 of the non-volatile memory 108a stores the image information of the duodenal furta papilla of the bile duct, the cannula method information 1087 of the non-volatile memory 108a stores the information indicating the cannula method, and the non-volatile memory 108a stores the information. The cannulation time information 1089 stores the cannulation time information.
(Step S11)
In the learning device 100a, the acquisition unit 154a acquires a plurality of pieces of information indicating images of the bile duct duodenal Fata papillae stored in the image information 1086 of the nonvolatile memory 108a, and acquires the acquired images of the plurality of bile duct duodenum Fata papillae. The information shown is output to the learning unit 156a.
(Step S12)
In the learning device 100a, the acquisition unit 154a acquires a plurality of pieces of information indicating the cannulation method stored in the cannulation method information 1087 of the non-volatile memory 108a, and acquires information indicating the acquired plurality of cannulation methods from the learning unit 100a. Output to 156a.
(Step S13)
In the learning device 100a, the acquisition unit 154a acquires a plurality of pieces of information indicating the cannulation time information stored in the cannulation time information of the non-volatile memory 108a, and learns the acquired information indicating the plurality of pieces of cannulation time information. It is output to the unit 156a.
(Step S14)
In the learning device 100a, the learning unit 156a acquires the image of the duodenal fata papilla of the bile duct output by the acquisition unit 154a, the information indicating the cannulation method, and the cannulation time information in association with each other. The learning unit 156a obtains the information indicating the cannulation method and the cannulation time information based on a plurality of pieces of information obtained by associating the acquired image of the duodenal furta papilla of the bile duct with the information indicating the cannulation method and the cannulation time information. Machine learning of images of the duodenal Fata papilla of the bile duct is performed as teacher data. The learning unit 156a includes a feature amount of an image of the bile duct duodenum Fata papilla obtained by machine learning a plurality of images of the bile duct duodenum Fata papilla, and a cannulation method performed on the image of the bile duct duodenum Fata papilla. And the cannulation time information are output to the output unit 158a.

(Step S15)
In the learning device 100a, the output unit 158a includes the feature amount of the image of the duodenal Fata papilla of the bile duct output by the learning unit 156a, the cannulation method performed on the image of the duodenal Fata papilla of the bile duct, and the cannulation time information. And the feature amount of the acquired image of the duodenal Fata papilla of the bile duct is associated with the cannulation method and the cannulation time information performed on the image of the duodenal Fata papilla of the bile duct.
(Step S16)
In the learning device 100a, the output unit 158a provides information that associates the acquired feature amount of the image of the duodenal Fata papilla of the bile duct, the cannulation method performed on the image of the duodenal Fata papilla of the bile duct, and the cannulation time information. , And is stored in the cannula method estimation table 1088a of the non-volatile memory 108a.
In the flowchart shown in FIG. 9, the order of steps S11 to S13 may be changed.
In the above-described modified example, a case has been described in which the image feature amount of the two types of duodenal Fata papillae and the cannulation time information are associated with the contrast method, but the present invention is not limited to this example. For example, the contrast method may be associated with one type of image feature amount of duodenal Fata papillae and cannulation time information, or may be associated with three or more types of image feature amount of duodenal Fata papillae and cannulation time information. May be. The same applies to the wire guide method, pancreatic duct guide wire method, and precut method.
According to the learning apparatus 100a of the modified example, the acquisition unit 154a associates the image of the duodenal fatta papilla of the bile duct with the information indicating the cannulation method, and terminates the insertion of the catheter into the bile duct after the further image is obtained. Acquires information related to the cannulation time, which is the time until. The learning unit 156a machine-learns, based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method and the cannulation time as teacher data. With this configuration, it is possible to derive information that associates the feature amount of the image of the duodenal Fata papilla of the bile duct with the cannulation method and the cannulation time information performed in the image of the duodenal Fata papilla of the bile duct. Bile duct cannulation is difficult by using the feature amount of the image of the duodenal Fata papilla of the bile duct and the information relating the cannulation time information with the method of cannulation performed on the image of the duodenal Fata papilla of the bile duct It is possible to improve the prediction accuracy of whether or not it is possible and the prediction accuracy of the cannulation time.

(Second embodiment)
(Estimator)
The estimation device according to the embodiment will be described. The estimation apparatus 200 according to the embodiment acquires an image of the duodenal fata papilla of the bile duct whose cannulation method is unknown. Here, the image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown is an image obtained by observing the duodenal Fata papilla of the bile duct with an endoscope when performing the bile duct cannulation. The image of the duodenal Fata papilla of the bile duct is preferably a front view image obtained by observing the duodenal Fata papilla of the bile duct with an endoscope immediately before performing cannulation.
Further, the estimation device 200 according to the embodiment, by machine learning the information indicating the image of the duodenal Fata papilla of the bile duct for which the obtained cannulation method is unknown, the cannulation method for the duodenum Fata papilla of the unknown duct method. Obtain the feature amount of the image. The estimation apparatus 200 according to the embodiment acquires, from the learning apparatus 100 described above, information that associates the feature amount of the image of the duodenal Fata papilla of the bile duct with the cannulation method from which the image of the duodenal Fata papilla of the bile duct was obtained. The estimation apparatus 200 according to the embodiment, based on the information that associates the acquired feature amount of the image of the duodenal Fata papilla of the bile duct with the cannulation method from which the image of the duodenal Fata papilla of the bile duct was associated, the cannulation method is unknown. Estimate the cannulation method related to the image features of the bile duct duodenal Fata papillae.

(Hardware configuration of the estimation device)
FIG. 10: is a figure which shows an example of the hardware constitutions of the estimation apparatus which concerns on embodiment of this invention. The estimation device 200 will be described.
The estimation device 200 includes a CPU 202, a memory 204, a communication unit 206, a non-volatile memory 208, an operation unit 210, a display unit 212, and an internal bus 250.
The CPU 202 executes, for example, the program 2082 and the application 2084 stored in the non-volatile memory 208, and uses the memory 204 as a work memory to control each unit of the estimation device 200. The memory 204 is composed of a RAM such as a volatile memory using a semiconductor element, and is used as a work memory of the CPU 202.
The communication unit 206 communicates with external devices such as the terminal device 10 and the learning device 100 connected to the network 50 to send and receive data. Specifically, the communication unit 206 is connected to the network 50 by wire. In addition, the communication unit 206 may be configured by a wireless device that performs wireless communication using a wireless communication technology such as LTE. The internal bus 250 connects the CPU 202, the memory 204, the non-volatile memory 208, and the communication unit 206 to each other. The respective parts connected to the internal bus 250 can exchange data with each other via the internal bus 250.
The non-volatile memory 208 is composed of, for example, a flash memory or the like. A program 2082 executed by the CPU 202, an application 2084, and a cannulation method estimation table 2088 are stored in the non-volatile memory 208.
The app 2084 causes the estimation device 200 to acquire an image of the duodenal fata papilla of the bile duct whose cannulation method is unknown.
The application 2084 causes the estimation device 200 to machine-learn the image of the duodenal Fata papilla of the bile duct whose acquired cannulation method is unknown.
The app 2084 uses the information in which the estimation device 200 associates the result of machine learning with the information indicating the cannulation method as the teacher data and the information indicating the cannulation method, and the acquired cannulation method is unknown. Based on the result of machine learning of the image of the duodenal Fata papilla of the bile duct, a method for inserting a catheter into the bile duct shown in the image is estimated.
As the cannulation method estimation table 2088, the above-mentioned cannulation method estimation table 1088 can be applied.
The operation unit 210 is an input device that receives a user operation, and includes a pointing device such as a touch panel, a button, a dial, a touch sensor, a touch pad, and the like.
The display unit 212 is composed of, for example, a liquid crystal display or the like, and displays a cannulation method or the like from which an image of the duodenal Fata papilla of the bile duct is obtained.

(Functional configuration of the estimation device)
FIG. 11 is a functional block diagram of the estimation device according to the embodiment of the present invention.
The estimation device 200 functions as the reception unit 252, the learning unit 254, the estimation unit 255, and the output unit 256 by the CPU 202 executing the program 2082 and the application 2084 that are loaded from the nonvolatile memory 208 on the memory 204.

(Each functional configuration of the estimation device)
Each functional configuration of the estimation device 200 will be described in detail with reference to FIGS. 10 and 11. In the following, in describing each function of the estimation apparatus 200, a relationship with main hardware for realizing each functional block of the estimation apparatus 200 illustrated in FIG. 11 will also be described.
The reception unit 252 is realized by the command from the CPU 202 and the communication unit 206. The reception unit 252 receives information indicating an image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown. Here, the information indicating the image of the bile duct duodenal Fata papilla of which the cannulation method is unknown is an image obtained by observing the duodenal Fata papilla of the bile duct with an endoscope when performing the bile duct cannulation. When the reception unit 252 acquires the information indicating the image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown, the learning unit 254 acquires the information indicating the image of the duodenal Fata papilla of the bile duct whose acquired cannulation method is unknown. Output to.
The learning unit 254 is realized by a command from the CPU 202. When the learning unit 254 acquires information indicating the image of the bile duct duodenal Fata papilla of which the cannulation method is unknown, which is output by the reception unit 252, the learning unit 254 indicates the image of the bile duct Duodenal Fata papilla of which the obtained cannulation method is unknown. Perform machine learning such as deep learning on information. The learning unit 254 outputs the feature amount of the image of the bile duct duodenal Fata papilla obtained by performing deep learning to the information indicating the image of the duodenal Fata papilla of the bile duct for which the cannulation method is unknown, to the estimation unit 255. To do.

The estimation unit 255, via the communication unit 206, from the learning device 100, a cannula method estimation table 1088 that associates the feature amount of the image of the duodenal Fata papilla of the bile duct with the cannulation method of the image of the duodenal Fata papilla of the bile duct. To get. The estimation unit 255, when acquiring the cannulation method estimation table 1088, stores it in the cannulation method estimation table 2088 of the nonvolatile memory 208.
The estimation unit 255 acquires the feature amount of the image of the duodenal Fata papilla of the bile duct for which the cannulation method output by the learning unit 254 is unknown, and the cannulation method related to the obtained feature amount of the image of the duodenal Fata papilla of the bile duct. From the cannula method estimation table 2088. For example, the estimating unit 255 obtains the image and the feature amount of the acquired duodenal vertebral papilla of the bile duct whose cannulation method is unknown, and each of the plurality of duodenal phater papilla image feature amounts stored in the cannulation method estimating table 2088. Derive the degree of similarity of. The estimation unit 255 acquires the cannulation method associated with the duodenal phater papillary image feature amount having the highest similarity based on the derived similarity. When estimating the cannulation method, the estimation unit 255 outputs the estimation result of the cannulation method to the output unit 256.
The output unit 256 is realized by a command from the CPU 202. When the output unit 256 acquires the estimation result of the cannulation method output by the estimation unit 255, the output unit 256 outputs the acquired estimation result of the cannulation method to the display unit 212.

(Operation of the estimation device)
FIG. 12 is a flowchart showing an example of the operation of the estimation device according to the embodiment of the present invention. In FIG. 12, the estimation apparatus 200 acquires, from the learning apparatus 100, a cannulation method estimation table 1088 that associates the feature amount of the image of the bile duct duodenal Fata papilla with the cannulation method of the image of the bile duct duodenal Fata papilla, The operation after storing the cannulation method estimation table 1088 in the cannulation method estimation table 2088 is shown.
(Step S21)
The reception unit 252 of the estimation apparatus 200 acquires information indicating an image of the duodenal fata papilla of the bile duct whose cannulation method is unknown. The reception unit 252 outputs information indicating the image of the duodenal fatta papilla of the bile duct whose acquired cannulation method is unknown to the learning unit 254.
(Step S22)
The learning unit 254 of the estimation device 200 acquires information indicating an image of the bile duct duodenal fatta papilla of which the cannulation method is unknown, which is output by the reception unit 252, and the acquired duodenal fatter papilla of the bile duct whose cannulation method is unknown. Machine learning such as deep learning is performed on the information indicating the image. The learning unit 254 outputs the feature amount of the image of the bile duct duodenal Fata papilla obtained by performing deep learning to the information indicating the image of the duodenal Fata papilla of the bile duct for which the cannulation method is unknown, to the estimation unit 255. To do.

(Step S23)
The estimating unit 255 of the estimating apparatus 200 acquires the feature amount of the image of the bile duct duodenal fatta papillae of which the cannulation method is unknown and which is output by the learning unit 254, and the acquired duodenal fatter papilla of the bile duct whose unknown cannulation method is unknown. The cannulation method associated with the feature amount of the image is acquired from the cannulation method estimation table 2088. The estimation unit 255 outputs the acquired cannulation method to the output unit 256.
(Step S24)
The output unit 256 of the estimation apparatus 200 acquires the estimation result of the cannulation method output by the estimation unit 255 and outputs the acquired cannulation method to the display unit 212.

In the above-described embodiment, the estimation device 200 obtains from the learning device 100 the cannulation method estimation table 1088 in which the feature amount of the image of the bile duct duodenal Fata papilla and the cannulation method of the image of the bile duct duodenal Fata papilla are associated with each other. The case where the obtained cannulation method estimation table 1088 is stored in the cannulation method estimation table 2088 has been described, but the present invention is not limited to this example. For example, the estimation device 200 may inquire the learning device 100 about the cannulation method related to the feature amount of the image of the duodenal fata papilla of the bile duct whose cannulation method is unknown. In response to the inquiry from the estimation device 200, the learning device 100 notifies the estimation device 200 of the cannulation method associated with the feature amount of the image of the duodenal fata papilla of the bile duct whose cannulation method is unknown.
In the above-described embodiment, the estimation device 200 acquires an image of the bile duct duodenal Fata papilla of which the cannulation method is unknown, and machine-learns the acquired image of the bile duct duodenal Fata papilla of which the cannulation method is unknown. By using, the feature quantity of the image of the duodenal Fata papilla of the bile duct for which the cannulation method is unknown is derived. Furthermore, a case has been described where the estimation device 200 acquires the cannulation method associated with the feature amount of the image of the duodenal phatera papilla of the bile duct whose derived cannulation method is unknown from the cannulation method estimation table 2088. It is not limited to this example.
For example, the estimation device 200 acquires a video of a bile duct duodenal Fata papilla with an unknown cannulation method, and machine-learns the acquired video of a bile duct duodenal Fata papilla with an unknown cannulation method to perform cannulation. The feature amount of the moving image of the duodenal Fata papilla of the bile duct whose method is unknown may be derived. Furthermore, the estimation device 200 may acquire, from the cannulation method estimation table 2088, the cannulation method associated with the feature amount of the moving image of the duodenal Fata papilla of the bile duct whose derived cannulation method is unknown.

  According to the estimation device 200 according to the embodiment, based on the cannulation method estimation table 1088 acquired from the learning device 100, the cannulation method relating to the feature amount of the image of the duodenal fata papilla of the bile duct whose cannulation method is unknown. Can be estimated. Based on the cannulation method estimation table 1088 obtained from the learning device 100, the cannulation method from which an image of the duodenal Fata papilla of an unknown bile duct is obtained can be obtained. Therefore, difficult cases of cannulation can be predicted regardless of the experience value of the operator. In other words, an operator with any experience value can set a strategy for bile duct cannulation. In addition, in the ERCP-related procedure, whether or not it is a difficult case can be determined when the bile duct cannulation is started. Since it is possible to actually perform cannulation based on the acquired cannulation method, in the case of a difficult case, it is possible to promptly introduce a rescue method at an early stage. With such a configuration, accidental accidents can be prevented. Further, the time required for the ERCP-related procedure can be shortened.

(Modification)
(Estimator)
An estimation device according to the modification will be described. The estimation apparatus 200a according to the modification acquires an image of the duodenal fata papilla of the bile duct whose cannulation method is unknown.
Further, the estimation apparatus 200a according to the modification performs machine learning such as deep learning on information indicating an image of the duodenal vertebral papilla of the bile duct for which the obtained cannulation method is unknown. The feature amount of the image of the duodenal fata papilla is obtained.
The estimation device 200a according to the modified example associates the feature amount of the image of the bile duct duodenum Fata papilla, the cannulation method and the cannulation time information from which the image of the bile duct duodenum Fata papilla was obtained from the learning device 100a described above. Get information.
The estimation device 200a according to the modified example is based on the information in which the feature amount of the acquired image of the duodenal Fata papilla of the bile duct and the cannulation method and the cannula time information in which the image of the duodenal Fata papilla of the bile duct is obtained is associated with each other, The cannulation method and the cannulation time related to the feature quantity of the image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown are estimated.

(Hardware configuration of the estimation device)
FIG. 13: is a figure which shows an example of the hardware constitutions of the estimation apparatus which concerns on a modification. The estimation device 200a will be described.
The estimation device 200a includes a CPU 202, a memory 204, a communication unit 206, a nonvolatile memory 208a, an operation unit 210, a display unit 212, and an internal bus 250.
The CPU 202 executes, for example, the program 2082 and the application 2084a stored in the non-volatile memory 208a, and uses the memory 204 as a work memory to control each unit of the estimation device 200a.
The non-volatile memory 208a is composed of, for example, a flash memory or the like. A program 2082 executed by the CPU 202, an application 2084a, and a cannula method estimation table 2088a are stored in the non-volatile memory 208a.
The application 2084a causes the estimation device 200a to acquire an image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown.
The application 2084a causes the estimation device 200a to machine-learn the acquired image of the duodenal fata papilla of the bile duct whose cannulation method is unknown.
The app 2084a uses the estimation device 200a to acquire the cannula obtained by using the result of machine learning with the information indicating the cannulation method as teacher data, the information indicating the cannulation method, and the information relating the cannulation time information. Based on the result of machine learning of the image of the duodenal Fata papilla of the bile duct whose method is unknown, a method for inserting a catheter into the bile duct shown in the image and a cannulation time are estimated.
The above-mentioned cannulation method estimation table 1088a can be applied to the cannulation method estimation table 2088a.

(Functional configuration of the estimation device)
FIG. 14 is a functional block diagram of an estimation device according to the modification.
The estimation device 200a functions as the reception unit 252, the learning unit 254, the estimation unit 255a, and the output unit 256a by the CPU 202 executing the program 2082 and the application 2084a expanded from the nonvolatile memory 208 on the memory 204.

(Each functional configuration of the estimation device)
Each functional configuration of the estimation device 200a will be described in detail with reference to FIGS. 13 and 14. Note that, in the following, in explaining each function of the estimation device 200a, a relationship with main hardware for realizing each functional block of the estimation device 200a shown in FIG. 14 will also be described.
The estimation unit 255a associates the learning device 100a with the feature amount of the image of the bile duct duodenal Fata papilla, the cannula method of the image of the duodenal Fata papilla of the bile duct, and the cannulation time information via the communication unit 206. Obtains the correlation method estimation table 1088a. When the estimation unit 255a acquires the cannula method estimation table 1088a, the estimation unit 255a stores the cannula method estimation table 1088a in the cannula method estimation table 2088a.
The estimation unit 255a acquires the feature amount of the image of the duodenal Fata papilla of the bile duct for which the cannulation method output by the learning unit 254 is unknown, and the cannulation method related to the feature amount of the image of the duodenal Fata papilla of the obtained bile duct. And the cannulation time are acquired from the cannulation method estimation table 2088a. The estimation unit 255a, when estimating the cannulation method and the cannulation time, outputs the estimation result of the cannulation method and the cannulation time to the output unit 256a.
The output unit 256a is realized by a command from the CPU 202. When the output unit 256a acquires the estimation result of the cannula method and the cannula time output by the estimating unit 255a, the output unit 256a outputs the acquired estimation result of the cannula method to the display unit 212.

(Operation of the estimation device)
FIG. 15 is a flowchart showing an example of the operation of the estimation device according to the modified example. In FIG. 15, the estimation device 200a uses the learning device 100a to associate the feature amount of the image of the bile duct duodenal Fata papilla, the cannulation method of the image of the bile duct duodenal Fata papilla, and the cannulation time estimation table 1088a. The operation after the acquisition of the cannula method estimation table 1088a is stored in the cannula method estimation table 2088a.
The steps S21-S22 described with reference to FIG. 12 can be applied to the steps S31-S32.
(Step S33)
The estimation unit 255a of the estimation device 200a acquires the feature amount of the image of the bile duct duodenal vertebral papilla of which the cannulation method output by the learning unit 254 is unknown, and the obtained cannulation method of the bile duct duodenal phater papilla of unknown type. The cannulation method and the cannulation time associated with the feature amount of the image are acquired from the cannulation method estimation table 2088a. The estimation unit 255a outputs the acquired cannulation method and cannulation time to the output unit 256a.
(Step S34)
The output unit 256a of the estimation device 200a acquires the estimation result of the cannula method and the cannula time output by the estimating unit 255a, and outputs the acquired cannula method and the cannula time to the display unit 212.

  In the above-described modified example, the estimation device 200a uses the learning device 100a to perform a cannulation method that associates the feature amount of the image of the bile duct duodenal Fata papilla, the cannulation method of the image of the bile duct duodenal Fata papilla, and the cannulation time information. The case where the estimation table 1088a is acquired and the acquired cannulation method estimation table 1088a is stored in the cannulation method estimation table 2088a has been described, but the present invention is not limited to this example. For example, the estimation device 200a may inquire the learning device 100a about the cannulation method and the cannulation time related to the feature amount of the image of the duodenal fata papilla of the bile duct whose cannulation method is unknown. In response to the inquiry from the estimation device 200a, the learning device 100a determines the cannulation method and the cannulation time associated with the feature amount of the image of the duodenal phatera papilla of the bile duct whose cannulation method and cannulation time are unknown. , And notifies the estimation device 200a.

  According to the estimation device 200a according to the modified example, based on the cannulation method estimation table 1088a acquired from the learning device 100a, the feature amount of the image of the duodenal fata papilla of the bile duct whose cannulation method and cannulation time are unknown. The associated cannulation method and cannulation time can be estimated. The cannulation method and cannulation time from which the image of the duodenal Fata papilla of an unknown bile duct was obtained can be acquired. Therefore, difficult cases of cannulation and the time required for cannulation can be predicted regardless of the experience value of the operator. In other words, an operator with any experience value can set a strategy for bile duct cannulation. In addition, in the ERCP-related procedure, whether or not it is a difficult case can be determined when the bile duct cannulation is started. Cannulation can be actually performed based on the acquired cannulation method and cannulation time. Therefore, in the case of a difficult case, early rescue method introduction can be promoted. With such a configuration, accidental accidents can be prevented. Further, the time required for the ERCP-related procedure can be shortened.

Although the embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and includes design changes and the like within a range not departing from the gist of the present invention.
Further, a computer program for realizing the functions of the learning device 100, the learning device 100a, the estimation device 200, and the estimation device 200a is recorded in a computer-readable recording medium, and the program recorded in this recording medium is stored in a computer. It may be read by the system and executed. The "computer system" may include an OS and hardware such as peripheral devices.
The "computer-readable recording medium" means a writable non-volatile memory such as a flexible disk, a magneto-optical disk, a ROM, a flash memory, a portable medium such as a DVD (Digital Versatile Disk), and a built-in computer system. A storage device such as a hard disk.

Further, the "computer-readable recording medium" means a volatile memory (for example, a DRAM (Dynamic) in a computer system that serves as a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. Random Access Memory)), which also holds the program for a certain period of time.
The program may be transmitted from a computer system that stores the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
Further, the program may be for realizing a part of the functions described above.
Further, it may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

10 ... Terminal device, 50 ... Network, 100, 100a ... Learning device, 102 ... CPU, 104 ... Memory, 106 ... Communication part, 108, 108a ... Non-volatile memory, 1082 ... Program, 1084, 1084a ... App, 1086 ... Image Information, 1087 ... Cannulation method information, 1088, 1088a ... Cannulation method estimation table, 152, 152a ... Receiving section, 154, 154a ... Acquisition section, 156, 156a ... Learning section, 158, 158a ... Output section, 200, 200a ... estimating device, 202 ... CPU, 204 ... memory, 206 ... communication unit, 208, 208a ... non-volatile memory, 2082 ... program, 2084, 2084a ... application, 2088, 2088a ... cannulation method estimation table, 252, ... receiving unit 254 ... Learning unit, 55,255a ... estimation unit, 256,256a ... the output unit

Claims (9)

An image of the duodenal Fata papilla of the bile duct, and an acquisition unit that acquires a plurality of pieces of information that associates information indicating a cannulation method that is a method of inserting a catheter into the bile duct,
Based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method as teacher data, a learning unit for machine learning, and
A learning device, comprising: a storage unit that stores a result of machine learning performed by the learning unit and information indicating the cannulation method in association with each other.   The learning device according to claim 1, wherein in each of the images of the duodenal fatter papillae of the plurality of bile ducts, at least one of the orientation and the position of the duodenal fatter papillae or both of them are different.   The learning device according to claim 1 or 2, wherein each of the images of the duodenal Fata papillae of a plurality of bile ducts has a different shape of the opening of the bile duct. The acquisition unit is associated with the image of the duodenal Fata papilla of the bile duct and the information indicating the cannulation method, and is a time period from the time when the image is obtained until the insertion of the catheter into the bile duct is completed. Information that is associated with the
4. The learning unit performs machine learning based on the image of the duodenal Fata papilla of the bile duct, the information indicating the cannulation method, and the cannulation time as teacher data. The learning device according to item 1. An acquisition unit that acquires an image of the duodenal Fata papilla of the bile duct whose cannulation method is unknown,
Based on the image of the duodenal Fata papilla of the bile duct, using the information obtained by associating the result of machine learning with the information indicating the cannulation method as teacher data and the information indicating the cannulation method, the acquisition unit acquires. And an estimating unit that estimates a method of inserting a catheter into the bile duct shown in the image based on the image of the duodenal vertebral papilla of the bile duct for which the cannulation method is unknown. An image of the duodenal Fata papilla of the bile duct and a step of acquiring a plurality of pieces of information associated with information indicating a cannulation method that is a method of inserting a catheter into the bile duct,
Based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method as teacher data, machine learning, and
A learning method executed by a learning device, comprising: storing the result of machine learning in the machine learning step in association with information indicating the cannulation method. Acquiring an image of the duodenal Fata papilla of the bile duct for which the cannulation method is unknown;
Based on the image of the duodenal Fata papilla of the bile duct, the result of machine learning as information indicating the cannulation method as machine data, and using information that associates the information indicating the cannulation method with the acquisition step. Estimating the method of inserting a catheter into the bile duct shown in the image based on the image of the duodenal Fata papilla of the bile duct for which the obtained cannulation method is unknown. On the computer of the learning device,
An image of the duodenal Fata papilla of the bile duct and a step of acquiring a plurality of pieces of information associated with information indicating a cannulation method that is a method of inserting a catheter into the bile duct,
Based on the image of the duodenal Fata papilla of the bile duct, information indicating the cannulation method as teacher data, machine learning, and
A program for executing the step of storing the result of machine learning in the machine learning step and the information indicating the cannulation method in association with each other. In the computer of the estimation device,
Acquiring an image of the duodenal Fata papilla of the bile duct for which the cannulation method is unknown;
Based on the image of the duodenal Fata papilla of the bile duct, the result of machine learning as information indicating the cannulation method as machine data, and using information that associates the information indicating the cannulation method with the acquisition step. A step of estimating a method of inserting a catheter into the bile duct shown in the image based on the image of the duodenal vertebral papilla of the bile duct for which the obtained cannulation method is unknown.