Pain estimation device, pain estimation method, and recording medium

Abstract

A pain estimation device has an information acquisition section configured to perform processing for acquiring estimation insertion condition information including at least one of estimation insertion shape information related to an insertion shape of an insertion portion of an endoscope inserted into a subject's body in one endoscopic examination and estimation operation force information related to a force applied to the insertion portion in the one endoscopic examination, and a pain estimation processing section configured to perform processing by applying the estimation insertion condition information to an estimation model to generate pain information related to a degree of pain of the subject, the estimation model being made using beforehand collected information including information on a relationship between at least one of beforehand collected insertion shape information and beforehand collected operation force information and beforehand collected pain information.

Description

Technical Field

[0001] The invention relates to a pain estimation device, a pain estimation method, and a recording medium used during endoscopic examinations. Background Technology

[0002] In endoscopic examinations in the medical field, an insertion procedure is performed to insert a slender insertion part of an endoscope deep into the body of a patient or other subject. Furthermore, in endoscopic examinations in the medical field, methods have been proposed to obtain information that facilitates the insertion of the endoscope's insertion part.

[0003] Specifically, for example, International Publication No. 2018 / 135018 discloses a method for calculating the force applied to the insertion part inserted into the body of a subject, as information for the insertion operation of an endoscope insertion part.

[0004] Here, in the field of medical endoscopy, a method for estimating the degree of pain in a subject (who is undergoing endoscopy) whose insertion site of an endoscope has been inserted is investigated.

[0005] However, International Publication No. 2018 / 135018 does not disclose a specific method for estimating the degree of pain of a subject based on the force applied to the insertion part inserted into the subject's body. Therefore, the structure disclosed in International Publication No. 2018 / 135018 presents problems such as the surgeon or other user performing the insertion operation of the endoscope's insertion part being unable to ascertain the degree of pain of the subject, sometimes causing excessive burden on the user.

[0006] The present invention was made in view of the above circumstances, and its object is to provide a pain estimation device, a pain estimation method, and a recording medium that can reduce the burden on users performing insertion operations of the insertion part of an endoscope. Summary of the Invention

[0007] means for solving problems

[0008] One aspect of the pain estimation device of the present invention includes: an information acquisition unit configured to perform processing for acquiring estimation examination status information, the estimation examination status information including estimation insertion status information, the estimation insertion status information including at least one of estimation insertion shape information relating to the insertion shape of an insertion portion of an endoscope inserted into the patient's body during an endoscopic examination, and estimation operating force information relating to the force applied to the insertion portion during the endoscopic examination, the estimation examination status information further including information different from both the estimation insertion shape information and the estimation operating force information; and a pain estimation processing unit configured to generate pain information related to the patient's pain based on the estimation examination status information.

[0009] One aspect of the pain estimation method of the present invention includes the following steps: obtaining estimation examination status information, the estimation examination status information including estimation insertion status information, the estimation insertion status information including at least one of estimation insertion shape information related to the insertion shape of the insertion portion of an endoscope inserted into the patient's body during an endoscopic examination, and estimation operating force information related to the force applied to the insertion portion during the endoscopic examination, the estimation examination status information further including information different from both the estimation insertion shape information and the estimation operating force information; and generating pain information related to the patient's pain based on the estimation examination status information.

[0010] One aspect of the present invention includes a recording medium recording a program that causes a computer to perform the following processes: acquiring estimated examination status information, the estimated examination status information including estimated insertion status information, the estimated insertion status information including at least one of estimated insertion shape information relating to the insertion shape of an endoscope inserted into a patient's body during an endoscopic examination, and estimated operating force information relating to the force applied to the insertion portion during the endoscopic examination, the estimated examination status information further including information different from both the estimated insertion shape information and the estimated operating force information; and generating pain information related to the patient's pain based on the estimated examination status information. Attached Figure Description

[0011] Figure 1 This is a diagram showing the structure of the main parts of an endoscopic system containing an embodiment of a pain estimation device.

[0012] Figure 2 This is a block diagram illustrating the specific structure of the endoscope system for implementing the method.

[0013] Figure 3This is a schematic diagram illustrating an example of the estimation model used in the processing of the pain estimation device in an embodiment. Detailed Implementation

[0014] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0015] For example, such as Figure 1 As shown, the endoscope system 1 is configured to include: an endoscope 10, a main body device 20, an insertion shape observation device 30, an operating force measuring device 40, an input device 50, and a display device 60. Figure 1 This is a diagram showing the structure of the main parts of an endoscopic system containing an embodiment of a pain estimation device.

[0016] The endoscope 10 is configured to include: an insertion section 11, which is inserted into the body of a patient or other subject; an operating section 16, which is disposed at the base of the insertion section 11; and a universal flexible cable 17, which extends from the operating section 16. Furthermore, the endoscope 10 is configured to be detachably connected to the main body device 20 via a scope connector (not shown) disposed at the end of the universal flexible cable 17. Additionally, a light guide 110 for transmitting illumination light supplied from the main body device 20 is provided inside the insertion section 11, the operating section 16, and the universal flexible cable 17. Figure 1 (Not shown in the image).

[0017] The insertion portion 11 is configured to have a flexible and elongated shape. Furthermore, the insertion portion 11 is configured by sequentially providing a rigid front end portion 12, a freely bendable curved portion 13, and a flexible, elongated tube portion 14 from the front end side. Additionally, within the front end portion 12, the curved portion 13, and the flexible tube portion 14, multiple source coils 18 are arranged at predetermined intervals along the length direction of the insertion portion 11 to generate a magnetic field corresponding to the coil drive signal supplied from the main body device 20. Furthermore, an air supply channel 120 is provided inside the insertion portion 11, which serves as a conduit for allowing gas supplied from the main body device 20 to flow and discharge forward of the front end portion 12. Figure 1 (Not shown in the figure). Furthermore, within a rigidly variable range provided in at least a portion of the insertion portion 11, a rigidly variable mechanism 130 is provided along the length direction of the insertion portion 11. Figure 1 (Not shown in the figure), the variable rigidity mechanism 130 is configured to change the bending rigidity within the variable range according to the control of the main body device 20. Furthermore, for ease of explanation, "bending rigidity" will be appropriately referred to as "rigidity" below.

[0018] An illumination window (not shown) is provided at the front end 12, which is used to emit illumination light transmitted by the light guide 110 provided inside the insertion part 11 onto the subject. Additionally, a camera part 140 is provided at the front end 12. Figure 1 (Not shown in the figure), the camera unit 140 is configured to perform an operation corresponding to the camera control signal provided from the main unit 20, and to capture images of the subject illuminated by the illumination light emitted through the illumination window and output a camera signal.

[0019] The bending portion 13 is configured to be able to bend according to the operation of the angle knob (not shown) provided on the operation portion 16.

[0020] The operating unit 16 is configured to be shaped so that a user can hold and operate it. Additionally, the operating unit 16 is provided with an angle knob, which is configured to bend the bending section 13 in four directions—up, down, left, and right—intersecting the length axis of the insertion section 11. Furthermore, the operating unit 16 is provided with one or more mirror switches (not shown) that can provide indications corresponding to user input operations.

[0021] The main unit 20 is configured to have one or more processors 20P and a non-transitory storage medium 20M. Furthermore, the main unit 20 is configured to be detachably connected to the endoscope 10 via a universal flexible cable 17. Additionally, the main unit 20 is configured to be detachably connected to each of the insertion shape observation device 30, the input device 50, and the display device 60. Furthermore, the main unit 20 is configured to perform actions corresponding to instructions from the input device 50. Moreover, the main unit 20 is configured to generate an endoscopic image based on the camera signal output from the endoscope 10, and to perform actions to display the generated endoscopic image on the display device 60. Furthermore, the main unit 20 is configured to generate and output various control signals for controlling the operation of the endoscope 10. Additionally, the main unit 20 is configured to function as a pain estimation device, processing data to estimate the pain level of the patient undergoing endoscopic examination and obtaining the estimation result, generating pain level information representing the obtained estimation result. In addition, the main unit 20 is configured to perform actions to cause the display device 60 to display pain level information generated as described above.

[0022] The insertion shape observation device 30 is configured to detect the magnetic field emitted from each of the source coils 18 provided in the insertion section 11, and to obtain the position of each of the multiple source coils 18 based on the strength of the detected magnetic field. In addition, the insertion shape observation device 30 is configured to generate insertion position information representing the position of each of the multiple source coils 18 obtained as described above, and to output the insertion position information to the main unit 20.

[0023] The force measurement device 40 is configured, for example, to include an electromyography (EMG) sensor capable of measuring the muscle potential generated in the hand or arm of the user operating the endoscope 10. Furthermore, the force measurement device 40 is configured to measure the voltage value generated based on the force applied by the user operating the endoscope 10 to the insertion portion 11, generate force information representing the measured voltage value, and output the force information to the main unit 20.

[0024] Furthermore, in this embodiment, for example, the operating force measuring device 40 may be configured to acquire the measurement result obtained by measuring the voltage value, etc., and generate operating force information representing the acquired measurement result and output it to the main device 20. The voltage value, etc., is generated based on the operating force applied to the insertion part 11 by a robot (not shown) capable of operating the endoscope 10.

[0025] The input device 50 is configured to have one or more user-operated input interfaces, such as a mouse, keyboard, and touch panel. Furthermore, the input device 50 is configured to output information and instructions input according to user operations to the main unit 20.

[0026] The display device 60 is configured to include, for example, a liquid crystal monitor. Furthermore, the display device 60 is configured to display endoscopic images, etc., output from the main unit 20 on a screen.

[0027] like Figure 2 As shown, the endoscope 10 is configured to include: a source coil 18, a light guide 110, an air delivery channel 120, a rigid variable mechanism 130, and an imaging unit 140. Figure 2 This is a block diagram illustrating the specific structure of the endoscope system for implementing the method.

[0028] The camera unit 140 is configured, for example, to have: an observation window into which reflected light from the subject illuminated by the illumination light is incident; and an image sensor such as a color CCD that captures the reflected light and outputs an image signal.

[0029] like Figure 2 As shown, the main device 20 is configured to include: a light source unit 210, an air supply unit 220, a rigidity control unit 230, an image processing unit 240, a coil drive signal generation unit 250, a display control unit 260, and a system control unit 270.

[0030] The light source unit 210 is configured to have one or more LEDs or lamps as the light source. Furthermore, the light source unit 210 is configured to generate illumination light for illuminating the body of the subject through which the insertion part 11 is inserted, and to supply this illumination light to the endoscope 10. Additionally, the light source unit 210 is configured to change the amount of illumination light according to a system control signal provided from the system control unit 270.

[0031] The gas delivery unit 220 is configured, for example, to include a gas delivery pump and a gas storage cylinder. Furthermore, the gas delivery unit 220 is configured to supply gas stored in the gas storage cylinder to the gas delivery channel 120 based on a system control signal provided from the system control unit 270.

[0032] The rigidity control unit 230 is configured to include, for example, a rigidity control circuit. Furthermore, the rigidity control unit 230 is configured to control the drive state of the rigidity variable mechanism 130 based on system control signals provided from the system control unit 270, thereby performing an operation to set the rigidity of the insertion part 11 within its rigidity variable range.

[0033] The image processing unit 240 is configured, for example, to have an image processing circuit. Furthermore, the image processing unit 240 is configured to generate an endoscopic image by performing prescribed processing on the camera signal output from the endoscope 10, and to output the generated endoscopic image to the display control unit 260 and the system control unit 270.

[0034] The coil drive signal generation unit 250 is configured to include a drive circuit, for example. Furthermore, the coil drive signal generation unit 250 is configured to generate and output a coil drive signal for driving the source coil 18 based on a system control signal provided from the system control unit 270.

[0035] The display control unit 260 is configured to perform processing for generating a display image, including an endoscopic image output from the image processing unit 240, and to perform processing for displaying the generated display image on the display device 60. Additionally, the display control unit 260 is configured to perform processing for displaying pain level information output from the system control unit 270 on the display device 60. Various information, including pain level information, displayed on the display device 60 is transmitted to the user (e.g., a doctor, nurse, or other medical professional).

[0036] The system control unit 270 is configured to generate and output system control signals for performing actions corresponding to instructions from the operation unit 16 and the input device 50. Furthermore, the system control unit 270 is configured to include an information acquisition unit 271 and a pain estimation processing unit 272.

[0037] The information acquisition unit 271 is configured to perform processing to acquire insertion status information (estimation insertion status information) that is equivalent to information indicating the insertion status of the insertion part 11 inserted into the subject's body, based on the insertion position information output from the insertion shape observation device 30.

[0038] Specifically, the information acquisition unit 271 is configured, for example, to calculate multiple curvatures corresponding to the respective positions of the multiple source coils 18 provided on the insertion part 11 based on multiple 3D coordinate values ​​(described later) included in the insertion position information output from the insertion shape observation device 30, and to obtain the calculation results, and to generate insertion status information including the obtained calculation results. That is, the information acquisition unit 271 is configured to perform processing to calculate multiple curvatures corresponding to the multiple positions of the insertion part 11, as a process for obtaining information (estimation insertion shape information) related to the insertion shape of the insertion part 11 inserted into the body of a subject undergoing an endoscopy.

[0039] The pain estimation processing unit 272 is configured to perform processing to obtain an estimation result of the subject's pain level based on the insertion status information generated by the information acquisition unit 271. The estimated result of the subject's pain level is obtained, for example, as one of a plurality of defined pain levels. In addition, the pain estimation processing unit 272 is configured to generate pain level information (pain information) representing the estimation result obtained as described above, and output the generated pain level information to the display control unit 260.

[0040] Here, a specific example of the structure of the pain estimation processing unit 272 in this embodiment will be described.

[0041] The pain estimation processing unit 272 is configured to process the data using an estimator CLP to obtain an estimation result that estimates the pain level of the subject corresponding to the insertion status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLP is created by learning the connection coefficients (weights) of a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0042] When creating the aforementioned estimator CLP, machine learning is performed, for example, using training data. This training data includes insertion status information identical to the insertion status information generated by the information acquisition unit 271, and labels representing classification results. The classification results are obtained by classifying the pain level of the subject corresponding to the insertion status information into one of a predetermined multiple pain levels. In other words, the estimator CLP is created, for example, by performing machine learning using pre-collected information as training data. This pre-collected information includes pre-collected insertion shape information and pre-collected pain information representing the pain level of the subject corresponding to the pre-collected insertion shape. Furthermore, the predetermined multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, for example, the following operation is performed: labels corresponding to the evaluation results obtained based on the subject's subjective evaluation criteria for assessing the pain level are assigned to the insertion status information. The subject's subjective evaluation criteria are, for example, the pressed state of a button switch with multiple switches, which are pressed by the subject based on the actual pain level generated during the endoscopic examination. Alternatively, when creating the training data mentioned above, the following tasks may be performed: the labels corresponding to the evaluation results obtained by evaluating the degree of pain based on objective evaluation criteria of experts or other persons different from the subject may be assigned to the insertion status information. The objective evaluation criteria of experts or other persons different from the subject may be, for example, the analysis results of waveforms obtained by an electroencephalogram (EEG) measuring the brain waves emitted by the subject during an endoscopy, and the analysis results of waveforms obtained by an electromyography (EMG) sensor measuring the myoelectric potentials generated by the subject, etc.

[0043] Therefore, according to the aforementioned estimator CLP, for example, by inputting multiple curvatures contained in the insertion status information generated by the information acquisition unit 271 as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the subject corresponding to the insertion status information. Furthermore, according to the processing using the aforementioned estimator CLP, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimate of the subject's pain level.

[0044] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the insertion status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLP for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLP is equivalent to an estimation model created by machine learning using the same information obtained before the single endoscopy as the insertion status information.

[0045] In this embodiment, at least a portion of the functions of the main device 20 can be implemented by the processor 20P. Furthermore, in this embodiment, at least a portion of the main device 20 can be configured as various electronic circuits, or as circuit blocks in an integrated circuit such as an FPGA (Field Programmable Gate Array). Additionally, by appropriately modifying the structure of this embodiment, for example, a computer can read a program from a storage medium 20M, such as a memory, to execute at least a portion of the functions of the main device 20, and perform actions corresponding to the read program.

[0046] like Figure 2 As shown, the insertion shape observation device 30 is configured to have a receiving antenna 310 and an insertion position information acquisition unit 320.

[0047] The receiving antenna 310 is configured, for example, to have multiple coils for 3D detection of magnetic fields emitted from multiple source coils 18. Furthermore, the receiving antenna 310 is configured to detect magnetic fields emitted from the multiple source coils 18, generate a magnetic field detection signal corresponding to the strength of the detected magnetic field, and output the magnetic field detection signal to the insertion position information acquisition unit 320.

[0048] The insertion position information acquisition unit 320 is configured to acquire the positions of each of the plurality of source coils 18 based on the magnetic field detection signal output from the receiving antenna 310. Furthermore, the insertion position information acquisition unit 320 is configured to generate insertion position information representing the positions of each of the plurality of source coils 18 acquired as described above, and output this insertion position information to the system control unit 270.

[0049] Specifically, the insertion position information acquisition unit 320 acquires, for example, multiple 3D coordinate values ​​in a hypothetical spatial coordinate system with a predetermined position (such as the anus) of the subject to which the insertion unit 11 is inserted serving as the origin or reference point, and uses these values ​​as the positions of the multiple source coils 18. Furthermore, the insertion position information acquisition unit 320 generates insertion position information including the multiple 3D coordinate values ​​acquired as described above, and outputs this insertion position information to the system control unit 270.

[0050] In this embodiment, at least a portion of the insertion shape observation device 30 can be configured as an electronic circuit, or as a circuit block in an integrated circuit such as an FPGA (Field Programmable Gate Array). Additionally, in this embodiment, for example, the insertion shape observation device 30 can also be configured to have one or more processors (CPUs, etc.).

[0051] Next, the function of this embodiment will be explained.

[0052] After the surgeon or other user connects the various parts of the endoscope system 1 and turns on the power, for example, the insertion part 11 is configured such that its front end 12 is located near the anus or rectum of the subject.

[0053] The information acquisition unit 271 performs processing to generate insertion status information based on the insertion position information output from the insertion shape observation device 30. The insertion status information includes calculation results of multiple curvatures corresponding to the respective positions of the multiple source coils 18 provided on the insertion unit 11.

[0054] The pain estimation processing unit 272 inputs multiple curvatures contained in the insertion status information generated by the information acquisition unit 271 into the estimator CLP for processing, thereby obtaining an estimation result of the subject's pain level corresponding to the insertion status information, and generating pain level information representing the obtained estimation result. Then, the pain estimation processing unit 272 outputs the pain level information generated as described above to the display control unit 260.

[0055] Specifically, the pain estimation processing unit 272 obtains an estimation result of the pain level of the examinee as any of the following pain levels: pain level PH, which is equivalent to a situation where the pain of the examinee is relatively large; pain level PL, which is equivalent to a situation where the pain of the examinee is relatively small; and pain level PN, which is equivalent to a situation where the pain of the examinee is not present.

[0056] The display control unit 260 performs processing to enable the display device 60 to display pain level information output from the pain estimation processing unit 272.

[0057] Specifically, for example, if the pain level indicated by the pain level information output from the pain estimation processing unit 272 is PH, the display control unit 260 generates a string indicating that the pain experienced by the subject is relatively high, and performs processing to display the generated string on the display device 60. Furthermore, for example, if the pain level indicated by the pain level information output from the pain estimation processing unit 272 is PL, the display control unit 260 generates a string indicating that the pain experienced by the subject is relatively low, and performs processing to display the generated string on the display device 60. Additionally, for example, if the pain level indicated by the pain level information output from the pain estimation processing unit 272 is PN, the display control unit 260 generates a string indicating that no pain is experienced by the subject, and performs processing to display the generated string on the display device 60.

[0058] As described above, according to this embodiment, it is possible to estimate the level of pain experienced by a patient undergoing an endoscopy, and to provide the user with information indicating the level of pain. Therefore, according to this embodiment, the burden on the user performing the insertion of the endoscope's insertion section can be reduced.

[0059] According to this embodiment, for example, the information acquisition unit 271 may also generate an insertion shape image that represents the insertion shape of the insertion part 11 inserted into the subject during endoscopy in a 2D manner, based on multiple 3D coordinate values ​​included in the insertion position information output from the insertion shape observation device 30, and perform processing for generating insertion status information including the generated insertion shape image. In other words, in this embodiment, the information acquisition unit 271 may also be configured to perform processing to generate an insertion shape image that represents the insertion shape of the insertion part 11 in a 2D manner, as a process for obtaining information related to the insertion shape of the insertion part 11.

[0060] Furthermore, in the case described above, the pain estimation processing unit 272 can be configured to use the estimator CLQ to process and obtain the estimation result. The estimation result is obtained by estimating the pain level of the subject corresponding to the insertion status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLQ is created by learning the connection coefficients (weights) of a multi-layer neural network including an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0061] When creating the aforementioned estimator CLQ, machine learning is performed, for example, using training data. This training data includes insertion status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the insertion status information into one of a set of predetermined pain levels. Furthermore, the predetermined multiple pain levels are, for example, set as multi-stage levels such as high, low, and none. Additionally, when creating the training data, a process is performed to assign labels corresponding to evaluation results to the insertion status information. These evaluation results are obtained by evaluating the pain level based on either a subjective evaluation criterion from the subject undergoing endoscopy or an objective evaluation criterion from an expert or other individual different from the subject.

[0062] Therefore, according to the aforementioned estimator CLQ, for example, by acquiring multidimensional data such as the pixel values ​​of each pixel of the insertion shape image included in the insertion status information generated by the information acquisition unit 271, and inputting this multidimensional data as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the subject corresponding to the insertion status information. Furthermore, according to the processing using the aforementioned estimator CLQ, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods included in the output data output from the output layer of the neural network can be obtained as an estimation result of the subject's pain level.

[0063] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the insertion status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLQ for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLQ is equivalent to an estimation model created by machine learning using the same information obtained before the single endoscopy as the insertion status information.

[0064] According to this embodiment, for example, the information acquisition unit 271 can acquire information on the estimated operating force as insertion status information. This estimated operating force information is information related to the force applied to the insertion unit 11 during endoscopic examination. For example, time-series data having multiple voltage values ​​can be acquired as estimated operating force information by continuously recording a voltage value output from the operating force measuring device 40 for a certain period of time, and processing can be performed to acquire insertion status information including the acquired time-series data.

[0065] Furthermore, in the case described above, the estimation result can be obtained by the pain estimation processing unit 272 using the estimator CLW. The estimation result is obtained by estimating the pain level of the subject corresponding to the insertion status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLW is created by learning the connection coefficients (weights) of a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0066] When creating the aforementioned estimator CLW, machine learning is performed, for example, using training data. This training data includes insertion status information identical to the insertion status information generated by the information acquisition unit 271, and labels representing classification results. The classification results are obtained by classifying the pain level of the subject corresponding to the insertion status information into one of a predetermined multiple pain levels. In other words, the estimator CLW is created, for example, by performing machine learning using pre-collected information as training data. This pre-collected information includes pre-collected operational force information and pre-collected pain information representing the pain level of the subject corresponding to the pre-collected operational force information. Furthermore, the predetermined multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopy or the objective evaluation criterion of an expert or other person different from the subject.

[0067] Therefore, according to the estimator CLW described above, for example, by inputting multiple voltage values ​​contained in the time series data of the insertion status information generated by the information acquisition unit 271 as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the subject corresponding to the insertion status information. Furthermore, according to the processing using the estimator CLW described above, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the subject's pain level.

[0068] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the insertion status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLW for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLW is equivalent to an estimation model created by machine learning using the same information obtained before the single endoscopy as the insertion status information.

[0069] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for acquiring examination status information (estimation examination status information). This examination status information, as information representing the examination status during an endoscopy, includes insertion status information and subject information (estimation subject information). The insertion status information is generated based on multiple 3D coordinate values ​​contained in the insertion position information output from the insertion shape observation device 30. The subject information is equivalent to information related to the subject obtained by detecting information input into the input device 50. Furthermore, the subject information may include, for example, any one of the following: information indicating the gender of the subject undergoing the endoscopy; information indicating the age of the subject; information indicating the body type of the subject; information indicating whether the subject has intestinal adhesions; and information indicating whether the subject has used sedatives.

[0070] Furthermore, in the case described above, the estimation result can be obtained by the pain estimation processing unit 272 using the estimator CLR. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLR is created by learning the connection coefficients (weights) in a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0071] When creating the aforementioned estimator CLR, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of multiple pain levels. In other words, the estimator CLR is created, for example, by using machine learning as training data: information containing the relationship between pre-collected subject information and pre-collected pain information, and pre-collected information containing the relationship between pre-collected insertion shape information and pre-collected pain information. Furthermore, the aforementioned multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either a subjective evaluation criterion from the subject undergoing the endoscopy or an objective evaluation criterion from an expert or other person different from the subject.

[0072] Therefore, according to the aforementioned estimator CLR, for example, by inputting multiple curvatures contained in the insertion status information of the examination status information generated by the information acquisition unit 271 and the values ​​in the examination status information corresponding to the subject information as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the subject corresponding to the examination status information. Furthermore, according to the processing using the aforementioned estimator CLR, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the subject's pain level.

[0073] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLR for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLR is created by machine learning using the same information obtained before the single endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0074] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for acquiring examination status information. This examination status information, which represents the examination status during a single endoscopic examination, includes insertion status information and estimated endoscopic image information. The insertion status information is generated based on multiple 3D coordinate values ​​included in the insertion position information output from the insertion shape observation device 30. The estimated endoscopic image information is related to the endoscopic image output from the image processing unit 240. The estimated endoscopic image information may, for example, be analysis information representing the analysis result obtained by performing analysis processing on the endoscopic image output from the image processing unit 240. In addition, the analysis information may include, for example, any one of the following: information indicating whether an over-proximity state has occurred, and information indicating whether there is a diverticulum in the patient's intestine. The over-proximity state corresponds to a state where the distance from the intestinal wall of the patient (where the insertion part 11 has been inserted) to the anterior end face 12 during the endoscopic examination is 0 or approximately 0. Furthermore, whether an excessive proximity state has occurred can be detected, for example, based on the proportion of red areas in the entire area of ​​the endoscopic image output from the image processing unit 240.

[0075] Furthermore, in the case described above, the estimation result can be obtained by the pain estimation processing unit 272 using the estimator CLS. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLS is created by learning the connection coefficients (weights) of a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0076] When creating the aforementioned estimator CLS, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of predetermined pain levels. In other words, the estimator CLS is created, for example, by using machine learning with the following information as training data: information containing the relationship between pre-collected endoscopic image information and pre-collected pain information, and pre-collected information containing the relationship between pre-collected insertion shape information and pre-collected pain information. Furthermore, the aforementioned predetermined multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopic examination or the objective evaluation criterion of an expert or other person different from the subject.

[0077] Therefore, according to the estimator CLS described above, for example, by inputting multiple curvatures contained in the insertion status information of the examination status information generated by the information acquisition unit 271 and the values ​​corresponding to the parsing information in the examination status information as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the examinee corresponding to the examination status information. Furthermore, according to the processing using the estimator CLS described above, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the examinee's pain level.

[0078] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLS for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLS is created by machine learning using the same information obtained before the single endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0079] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for generating inspection status information. This inspection status information, which represents the inspection status during a single endoscopic examination, includes insertion status information and air supply information (estimated air supply information). The insertion status information is generated based on multiple 3D coordinate values ​​included in the insertion position information output from the insertion shape observation device 30. The air supply information represents the detection result obtained by detecting the operating state of the air supply unit 220. Furthermore, the air supply information described above may simply include information indicating whether the supply of gas from the air supply unit 220 to the air supply channel 120 has been continuous for a certain period of time.

[0080] Furthermore, in the case described above, the estimation result can be obtained by configuring the pain estimation processing unit 272 to perform processing using the estimator CLT. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLT is created by learning the connection coefficients (weights) in a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0081] When creating the aforementioned estimator CLT, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of multiple pain levels. In other words, the estimator CLT is created, for example, by using machine learning as training data: information containing the relationship between pre-collected air delivery information and pre-collected pain information, and pre-collected information containing the relationship between pre-collected insertion shape information and pre-collected pain information. Furthermore, the aforementioned multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopy or the objective evaluation criterion of an expert or other person different from the subject.

[0082] Therefore, according to the aforementioned estimator CLT, for example, by inputting multiple curvatures contained in the insertion condition information of the examination condition information generated by the information acquisition unit 271 and the value corresponding to the air supply information in the examination condition information as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the examinee corresponding to the examination condition information. Furthermore, according to the processing using the aforementioned estimator CLT, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the examinee's pain level.

[0083] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLT for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLT is created by machine learning using the same information obtained before the single endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0084] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for generating inspection status information. This inspection status information, which represents the inspection status during a single endoscopic examination, includes insertion status information and rigidity control information (estimated rigidity variable part operation information). The insertion status information is generated based on multiple 3D coordinate values ​​included in the insertion position information output from the insertion shape observation device 30, and the rigidity control information represents the detection result obtained by detecting the operation state of the rigidity control unit 230. In other words, the rigidity control information is information related to the operation of the rigidity variable part provided on the insertion unit 11. Furthermore, the rigidity control information described above may simply include, for example, information indicating the set value of the rigidity set by the rigidity control unit 230.

[0085] Furthermore, in the case described above, the estimation result can be obtained by the pain estimation processing unit 272 performing processing using the estimator CLU. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLU is created by learning the connection coefficients (weights) in a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0086] When creating the aforementioned estimator CLU, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of multiple pain levels. In other words, the estimator CLU is created, for example, by using machine learning as training data: information containing the relationship between previously collected rigid variable part motion information and previously collected pain information, and prior collection information containing the relationship between previously collected insertion shape information and previously collected pain information. Furthermore, the aforementioned multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopy or the objective evaluation criterion of an expert or other person different from the subject.

[0087] Therefore, according to the aforementioned estimator CLU, for example, by inputting multiple curvatures contained in the insertion condition information of the examination condition information generated by the information acquisition unit 271 and the value corresponding to the rigidity control information in the examination condition information (the setting value of the rigidity magnitude) as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the examinee corresponding to the examination condition information. Furthermore, according to the processing using the aforementioned estimator CLU, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the examinee's pain level.

[0088] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLU for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLU is created by machine learning using the same information obtained before the single endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0089] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for generating examination status information, which is information representing the examination status in a single endoscopy examination, including insertion status information and endoscope information (estimated number of uses information). The insertion status information is generated based on multiple 3D coordinate values ​​contained in the insertion position information output from the insertion shape observation device 30, and the endoscope information is equivalent to information representing the number of uses of the endoscope 10 obtained by detecting the information input into the input device 50.

[0090] Furthermore, in the case described above, the estimation result can be obtained by configuring the pain estimation processing unit 272 to perform processing using the estimator CLV. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLV is created by learning the connection coefficients (weights) in a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0091] When creating the aforementioned estimator CLV, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of predetermined pain levels. In other words, the estimator CLV is created, for example, by using machine learning as training data: information containing the relationship between pre-collected usage frequency information and pre-collected pain information, and pre-collected information containing the relationship between pre-collected insertion shape information and pre-collected pain information. Furthermore, the aforementioned predetermined multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopy or the objective evaluation criterion of an expert or other person different from the subject.

[0092] Therefore, according to the aforementioned estimator CLV, for example, by inputting multiple curvatures contained in the insertion status information of the examination status information generated by the information acquisition unit 271 and the values ​​corresponding to the endoscopic information in the examination status information as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the examinee corresponding to the examination status information. Furthermore, according to the processing using the aforementioned estimator CLV, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the examinee's pain level.

[0093] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLV for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLV is created by machine learning using the same information obtained before the single endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0094] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for generating inspection status information. This inspection status information, which represents the inspection status during a single endoscopic examination, includes insertion status information and insertion part rigidity information (estimated insertion part rigidity information). The insertion status information is generated based on multiple 3D coordinate values ​​included in the insertion position information output from the insertion shape observation device 30, and the insertion part rigidity information represents the rigidity of the insertion part 11. Furthermore, the insertion part rigidity information represents the magnitude of rigidity predetermined based on the material, length, etc., of the insertion part 11. That is, the insertion part rigidity information represents a predetermined design value based on the material, length, etc., of the insertion part 11, and is different from rigidity control information that changes the magnitude of rigidity based on user operation and the operating state of the rigidity control unit 230.

[0095] Furthermore, in the case described above, the estimation result can be obtained by configuring the pain estimation processing unit 272 to perform processing using the estimator CLY. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLY is created by learning the connection coefficients (weights) in a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0096] When creating the estimator CLY, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of multiple pain levels. In other words, the estimator CLY is created, for example, by using machine learning with the following data as training data: information containing the relationship between pre-collected insertion rigidity information and pre-collected pain information, and pre-collected information containing the relationship between pre-collected insertion shape information and pre-collected pain information. Furthermore, the aforementioned multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopic examination or the objective evaluation criterion of an expert or other person different from the subject.

[0097] Therefore, according to the estimator CLY described above, for example, by inputting the multiple curvatures contained in the insertion status information in the examination status information generated by the information acquisition unit 271 and the values ​​corresponding to the insertion stiffness information in the examination status information as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the examinee corresponding to the examination status information. Furthermore, according to the processing using the estimator CLY described above, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the examinee's pain level.

[0098] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLY for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the one endoscopy to one of a number of prescribed pain levels. The estimator CLY is created by machine learning using the same information obtained before the one endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0099] According to this embodiment, for example, the information acquisition unit 271 may also perform processing for generating examination status information. This examination status information, which represents the examination status during a single endoscopic examination, includes insertion status information and insertion length information (estimated insertion length information). The insertion status information is generated based on multiple 3D coordinate values ​​included in the insertion position information output from the insertion shape observation device 30, and the insertion length information indicates the insertion length of the insertion part 11 into the subject. Furthermore, the insertion length information is acquired by the insertion shape observation device 30 and input to the information acquisition unit 271 of the system control unit 270. In other words, the insertion length information indicates the location (e.g., sigmoid colon, splenic flexure) of the anterior end portion 12 of the insertion part 11 within the intestine.

[0100] Furthermore, in the case described above, the estimation result can be obtained by the pain estimation processing unit 272 using the estimator CLX. The estimation result is obtained by estimating the pain level of the examinee corresponding to each piece of information contained in the examination status information generated by the information acquisition unit 271 as one of a number of prescribed pain levels. The estimator CLX is created by learning the connection coefficients (weights) of a multi-layer neural network containing an input layer, a hidden layer, and an output layer using learning methods such as deep learning.

[0101] When creating the estimator CLX, machine learning is performed, for example, using training data. This training data includes examination status information identical to that generated by the information acquisition unit 271, and labels representing classification results. These classification results categorize the pain level of the subject corresponding to the examination status information into one of a set of multiple pain levels. In other words, the estimator CLX is created, for example, by using machine learning with the following information as training data: information containing the relationship between pre-collected insertion length information and pre-collected pain information, and pre-collected information containing the relationship between pre-collected insertion shape information and pre-collected pain information. Furthermore, the aforementioned multiple pain levels are, for example, set as multi-stage levels such as large, small, and none. Additionally, when creating the training data, a task is performed to assign labels corresponding to evaluation results to the examination status information. These evaluation results are obtained by evaluating the pain level based on any evaluation criterion, either the subjective evaluation criterion of the subject undergoing the endoscopy or the objective evaluation criterion of an expert or other person different from the subject.

[0102] Therefore, according to the estimator CLX described above, for example, by inputting multiple curvatures contained in the insertion status information of the examination status information generated by the information acquisition unit 271 and the value corresponding to the insertion length information in the examination status information as input data into the input layer of the neural network, multiple likelihoods can be obtained as output data output from the output layer of the neural network. These multiple likelihoods each correspond to a level that can be estimated as the pain level of the examinee corresponding to the examination status information. Furthermore, according to the processing using the estimator CLX described above, for example, a pain level corresponding to the highest likelihood among the multiple likelihoods contained in the output data output from the output layer of the neural network can be obtained as an estimation result of the examinee's pain level.

[0103] That is, according to the processing described above, the pain estimation processing unit 272 is configured to apply the examination status information generated by the information acquisition unit 271 during an endoscopy to the estimator CLX for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the single endoscopy as one of a number of prescribed pain levels. The estimator CLX is created by machine learning using the same information obtained before the single endoscopy as the examination status information, and is made into an estimation model different from the estimator CLP.

[0104] According to this embodiment, for example, the information acquisition unit 271 can be configured to perform processing for generating insertion status information, which includes multiple curvatures obtained based on insertion position information output from the insertion shape observation device 30, and time-series data obtained based on operating force information output from the operating force measuring device 40. In other words, in this embodiment, as long as the information acquisition unit 271 is configured to perform processing for generating insertion status information, which is information representing the insertion status of the insertion part 11 inserted into the body of a subject undergoing an endoscopy, it can include at least one of information related to the insertion shape of the insertion part 11 and information obtained based on the force applied to the insertion part 11 during the endoscopy. In addition, in such a case, the pain estimation processing unit 272 can also be configured to perform processing such as estimating the subject's pain level and obtaining an estimation result based on the output data obtained by inputting the multiple curvatures included in the insertion status information generated by the information acquisition unit 271 into the estimator CLP, and the output data obtained by inputting the time-series data included in the insertion status information into the estimator CLW.

[0105] According to this embodiment, the pain estimation processing unit 272 is not limited to outputting pain level information generated based on the estimation result of the subject's pain level to the display control unit 260; for example, it may also output the pain level information to a speaker (not shown). Furthermore, in this case, for example, different warning tones or sounds can be output from the speaker based on the pain level information generated by the pain estimation processing unit 272.

[0106] According to this embodiment, the pain estimation processing unit 272 is not limited to outputting pain level information generated based on the estimation result of the subject's pain level to the display control unit 260; for example, it may also output the pain level information to a lamp (not shown). Furthermore, in this case, for example, the lamp can be illuminated with different flashing intervals based on the pain level information generated by the pain estimation processing unit 272.

[0107] According to this embodiment, the pain estimation processing unit 272 is not limited to generating pain level information representing the estimation result and outputting it to the display control unit 260. For example, it may further generate operation guidance information based on the estimation result to guide the insertion operation of the insertion unit 11 performed by the user, and output the operation guidance information to the display control unit 260. The estimation result is obtained by estimating the pain level of the examinee as one of a number of prescribed pain levels.

[0108] Specifically, the pain estimation processing unit 272, for example, when it obtains the pain level PH as an estimate of the subject's pain level, generates operation guidance information that causes the insertion unit 11 to temporarily stop, and outputs this operation guidance information to the display control unit 260. Furthermore, when it obtains the pain level PL as an estimate of the subject's pain level, the pain estimation processing unit 272 generates operation guidance information that causes adjustments to the insertion speed and force of the insertion unit 11, and outputs this operation guidance information to the display control unit 260. Additionally, when it obtains the pain level PN as an estimate of the subject's pain level, the pain estimation processing unit 272 generates operation guidance information that causes the insertion speed and force of the insertion unit 11 to be maintained, and outputs this operation guidance information to the display control unit 260. In addition, the pain estimation processing unit 272, for example, when it obtains the pain level PO as the estimated result of the subject's pain level, generates operation guidance information that prompts to perform an insertion operation different from the insertion operation of the insertion unit 11 currently being performed by the user, and outputs the operation guidance information to the display control unit 260 so as to change to a pain level that represents pain that is smaller than the pain level PO.

[0109] According to this embodiment, the pain estimation processing unit 272 is not limited to outputting operation guidance information generated based on the estimation result of the subject's pain level to the display control unit 260. For example, the operation guidance information may also be output to a speaker (not shown). Furthermore, in this case, for example, a sound prompting the user to perform an operation corresponding to the operation guidance information generated by the pain estimation processing unit 272 can be output from the speaker.

[0110] According to this embodiment, the pain estimation processing unit 272 is not limited to outputting operation guidance information generated based on the estimation result of the subject's pain level to the display control unit 260; for example, it may also output the operation guidance information to a lamp (not shown). Furthermore, in such a case, for example, the lamp can be illuminated in a state that prompts the performance of an operation corresponding to the operation guidance information generated by the pain estimation processing unit 272.

[0111] According to this embodiment, the pain estimation processing unit 272 is not limited to outputting pain level information and operation guidance information generated based on the estimation result of the subject's pain level to the display control unit 260. It can also use the pain level information and operation guidance information in the control of an automatic insertion device configured to automatically perform the insertion action of the insertion unit 11. In such a case, for example, if the pain level PP is obtained as the estimation result of the subject's pain level, operation guidance information is generated to prompt an insertion action different from the current insertion action of the insertion unit 11 performed by the automatic insertion device, and this operation guidance information is output to the automatic insertion device so as to change to a pain level representing pain lower than the pain level PP.

[0112] According to this embodiment, the pain estimation processing unit 272 is not limited to using an estimation model created through machine learning to perform processing related to the estimation of the subject's pain level. For example, it may also use an estimation model represented by a polynomial to perform processing related to the estimation of the subject's pain level. An example of such a case will be described below.

[0113] The pain estimation processing unit 272 calculates the pain value Pa by applying multiple curvatures contained in the insertion status information generated by the information acquisition unit 271 to an estimation model represented by a polynomial as shown in the following mathematical formula (1), and obtains an estimation result that estimates the subject's pain level based on the magnitude of the calculated pain value Pa. Furthermore, in the following mathematical formula (1), A1, A2, A3, ..., A... s A s+1 Representing approximate parameters, X1, X2, X3, ..., X s This represents the s curvatures contained in the insertion status information generated by the information acquisition unit 271.

[0114] Pa = A1X1 + A2X2 + A3X3 + ... + A s X s +A s+1 …(1)

[0115] The approximate parameters A1, A2, A3, ..., A in the above mathematical formula (1) are... s A s+1 For example, it can be calculated by performing the operations involved in the determinant represented by the following mathematical expression (2). Furthermore, in the following mathematical expression (2), P1, P2, P3, ..., P... m Let X represent m known pain values, which are equivalent to any evaluation criterion used to assess pain intensity based on either the subjective evaluation criterion of the subject undergoing endoscopy or the objective evaluation criterion of an expert or other person different from the subject. Furthermore, in the following mathematical formula (2), X... 1m X 2m X 3m ... X sm This represents the known s curvatures obtained based on the pain value Pm.

[0116]

[0117] That is, according to the example described above, the pain estimation processing unit 272 is configured to apply the insertion status information generated by the information acquisition unit 271 during an endoscopy to the polynomial in (1) above for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the endoscopy to one of a number of prescribed pain levels. The polynomial in (1) above is equivalent to an estimation model made using the same information as the insertion status information obtained before the endoscopy.

[0118] The estimation model represented by the polynomial as described above can achieve the same effect as the estimation model made through machine learning.

[0119] Furthermore, the estimation model used in the processing of the pain estimation processing unit 272 is not limited to being made into a first-order polynomial like the mathematical formula (1) above. For example, it can also be made into a polynomial with a degree of 2 or higher, which can apply multiple curvatures contained in the insertion status information generated by the information acquisition unit 271.

[0120] Furthermore, the estimation model used in the pain estimation processing unit 272 is not limited to being made into a polynomial like the mathematical formula (1) above. For example, it can also be made into a polynomial that can apply multiple curvatures contained in the insertion status information in the examination status information generated by the information acquisition unit 271 and the values ​​corresponding to the subject information in the examination status information.

[0121] According to this embodiment, the pain estimation processing unit 272 is not limited to using an estimation model created through machine learning to perform processing related to estimating the subject's pain level. For example, it may also use an estimation model obtained through statistical methods to perform processing related to estimating the subject's pain level. An example of such a case will be described below. In addition, the processing involved in creating the estimation model described below is not limited to being performed in the pain estimation processing unit 272, but may also be performed in a device different from the main device 20, such as a computer.

[0122] The pain estimation processing unit 272 generates a matrix C by arranging q (q≥2) curvatures corresponding to each of the p (p≥2) pain values, and performs singular value decomposition on the generated matrix C as shown in the following mathematical formula (3), where the p (p≥2) pain values ​​are equivalent to values ​​obtained by evaluating the degree of pain based on any evaluation criterion among the subjective evaluation criteria of the subject who underwent endoscopy and the objective evaluation criteria of experts or other persons different from the subject. In addition, in the following mathematical formula (3), V represents the left singular vector, S represents the singular value matrix, and U represents the left singular vector. T The matrix represents the transpose of the right singular vector.

[0123] c = VSU T …(3)

[0124] The pain estimation processing unit 272 obtains the largest value among the q components contained in the q-row, 1-column left singular vector V obtained by performing the singular value decomposition shown in the above mathematical formula (3) as the first component Vx, and obtains the second largest value among the elements contained in the left singular vector V as the second component Vy. That is, the first component Vx is obtained as the component that is estimated to have the greatest impact in the evaluation of each of the p pain values. In addition, the second component Vy is obtained as the component that is estimated to have the second largest impact in the evaluation of each of the p pain values.

[0125] The pain estimation processing unit 272 obtains the curvature Cx corresponding to the first component Vx and the curvature Cy corresponding to the second component Vy from the q curvatures corresponding to one of the p pain values ​​Px. That is, the curvature Cx and the curvature Cy can be represented as coordinate values ​​(Cx, Cy) in a 2D coordinate system defined by the first component Vx and the second component Vy.

[0126] The pain estimation processing unit 272 processes each of the p pain values ​​to obtain the coordinate value (Cx, Cy) corresponding to the pain value Px, thereby producing a... Figure 3 The estimated model shown is CMA. Figure 3 This is a schematic diagram illustrating an example of the estimation model used in the processing of the pain estimation device in the embodiment.

[0127] The pain estimation processing unit 272 obtains two curvatures corresponding to the first component Vx and the second component Vy from the multiple curvatures contained in the insertion status information in the examination status information generated by the information acquisition unit 271. The obtained two curvatures (coordinate values ​​determined by the two curvatures) are applied to the estimation model CMA, and clustering processing based on k-nearest neighbor method or the like is performed to obtain the estimation result of the subject's pain level.

[0128] That is, according to the example described above, the pain estimation processing unit 272 is configured to apply the insertion status information generated by the information acquisition unit 271 during an endoscopy to the estimation model CMA for processing, thereby obtaining an estimation result that estimates the pain level of the subject in the endoscopy as one of a number of prescribed pain levels. The estimation model CMA is created using the same information obtained before the endoscopy as the insertion status information.

[0129] The estimation model obtained using the statistical methods described above can achieve the same effect as the estimation model created through machine learning.

[0130] Furthermore, the present invention is not limited to the above-described embodiments, and various modifications, combinations, and applications can be made without departing from the spirit of the invention.

Claims

1. A pain estimation device, characterized in that, have: The information acquisition unit is configured to perform processing for acquiring estimated examination status information, which includes estimated insertion status information and estimated subject information related to the subject. The estimated insertion status information includes at least one of estimated insertion shape information related to the insertion shape of the insertion part of the endoscope inserted into the subject's body during the endoscopy and estimated operating force information related to the force applied to the insertion part during the endoscopy. as well as The pain estimation processing unit is configured to input the estimation insertion status information and the estimation subject information into an estimation model, obtain the subject's pain level corresponding to the estimation insertion status information and the estimation subject information, and use this as an estimation result of the subject's pain level. The estimation model is trained using training data, which includes pre-collected insertion status information and subject information, as well as classification labels representing the pain level corresponding to the insertion status information and subject information. The information acquisition unit is configured to perform processing to calculate multiple curvatures corresponding to multiple positions of the insertion part, or to perform processing to generate an insertion shape image representing the insertion shape of the insertion part, as a process for acquiring the estimated insertion shape information.

2. The pain estimation device according to claim 1, characterized in that, The pain estimation device also has an input device, wherein... The estimated subject information includes information related to the subject obtained by detecting information input into the input device.

3. The pain estimation device according to claim 1, characterized in that, The estimated subject information includes one of the following: information indicating the gender of the subject undergoing the endoscopy; information indicating the age of the subject; information indicating the body type of the subject; information indicating whether the subject has intestinal adhesions; and information indicating whether the subject has used sedatives.

4. A pain estimation device, characterized in that, have: The information acquisition unit is configured to perform processing for acquiring estimated examination status information, the estimated examination status information including estimated insertion status information and estimated endoscopic image information related to endoscopic images obtained by endoscopic imaging inside the patient's body, the estimated insertion status information including at least one of estimated insertion shape information related to the insertion shape of the insertion portion of the endoscope inserted into the patient's body during the endoscopic examination, and estimated operating force information related to the force applied to the insertion portion during the endoscopic examination; as well as The pain estimation processing unit is configured to input the estimation insertion status information and the estimation endoscopic image information into an estimation model, and obtain the pain level of the subject corresponding to the estimation insertion status information and the estimation endoscopic image information, as the estimation result of the subject's pain level. The estimation model is trained using training data, which includes pre-collected insertion status information and endoscopic image information, as well as classification labels representing the pain level corresponding to the insertion status information and endoscopic image information.

5. The pain estimation device according to claim 4, characterized in that, The pain estimation device also includes an image processing unit, wherein... The estimated endoscopic image information includes analytical information representing the analytical result obtained by performing analytical processing on the endoscopic image output from the image processing unit.

6. The pain estimation device according to claim 4, characterized in that, The estimated endoscopic image information includes one of information indicating whether an over-proximity state has occurred and information indicating whether there are diverticula in the subject's intestine.

7. A pain estimation device, characterized in that, have: The information acquisition unit is configured to perform processing for acquiring estimated examination status information, which includes estimated insertion status information, and one of estimated air supply information, estimated rigid variable part operation information, estimated number of uses information, estimated insertion part rigidity information, or estimated insertion length information. The estimated insertion status information includes at least one of estimated insertion shape information related to the insertion shape of the insertion part of the endoscope inserted into the patient's body during the endoscopy, and estimated operating force information related to the force applied to the insertion part during the endoscopy. as well as The pain estimation processing unit is configured to input one of the following into an estimation model: the estimation insertion status information, the estimation air delivery information, the estimation rigidity variable part operation information, the estimation usage frequency information, the estimation insertion part rigidity information, or the estimation insertion length information; and obtain the pain level of the subject corresponding to one of the following: the estimation insertion status information, the estimation air delivery information, the estimation rigidity variable part operation information, the estimation usage frequency information, the estimation insertion part rigidity information, or the estimation insertion length information. This pain level is then used as an estimation result for the subject's pain level. in, The estimated gas supply information is related to the operational state of the gas supply unit that performs the action of supplying gas to the endoscope. The estimated motion information of the rigid variable part is related to the motion of the rigid variable part provided in the insertion part. The estimated usage information is related to the number of times the endoscope is used. The estimation of the stiffness information of the insertion part is related to the stiffness of the insertion part. The estimated insertion length information is related to the insertion length of the insertion part into the subject. The estimation model is trained using training data, which includes pre-collected insertion status information, and one of the following: estimation air delivery information, estimation rigid variable part movement information, estimation usage frequency information, estimation insertion part rigidity information, or estimation insertion length information; and a classification label representing the pain level corresponding to the insertion status information, the estimation air delivery information, the estimation rigid variable part movement information, the estimation usage frequency information, the estimation insertion part rigidity information, or the estimation insertion length information.

8. The pain estimation device according to any one of claims 4 to 7, characterized in that, The information acquisition unit is configured to perform a process of calculating multiple curvatures corresponding to multiple positions of the insertion unit, as a process for acquiring the estimated insertion shape information.

9. The pain estimation device according to any one of claims 4 to 7, characterized in that, The information acquisition unit is configured to perform a process of generating an insertion shape image representing the insertion shape of the insertion part, as a process for acquiring the estimated insertion shape information.

10. The pain estimation device according to any one of claims 1 to 7, characterized in that, The pain estimation processing unit further generates operation guidance information based on the obtained pain level to guide the insertion operation of the insertion unit.

11. A recording medium having a program recorded thereon that causes a computer to perform the following processes: Obtaining estimated examination status information, the estimated examination status information including estimated insertion status information and estimated subject information related to the subject, the estimated insertion status information including at least one of estimated insertion shape information related to the insertion shape of the insertion portion of the endoscope inserted into the subject's body during the endoscopic examination, and estimated operating force information related to the force applied to the insertion portion during the endoscopic examination; and The estimation model is input with the estimated insertion status information and the estimated subject information to obtain the subject's pain level corresponding to the estimated insertion status information and the estimated subject information. This level is used as the estimated result of the subject's pain level. in, The estimation model is trained using training data, which includes pre-collected insertion status information and subject information, as well as classification labels representing the pain level corresponding to the insertion status information and subject information. The process of obtaining the estimated insertion shape information includes either calculating multiple curvatures corresponding to multiple positions of the insertion part or generating an insertion shape image representing the insertion shape of the insertion part.

12. A recording medium having a program recorded thereon that causes a computer to perform the following processes: Obtaining estimated examination status information, the estimated examination status information including estimated insertion status information and estimated endoscopic image information related to endoscopic images obtained by endoscopic imaging of the patient's body, the estimated insertion status information including at least one of estimated insertion shape information related to the insertion shape of the insertion portion of the endoscope inserted into the patient's body during the endoscopic examination, and estimated operating force information related to the force applied to the insertion portion during the endoscopic examination; and The insertion status information and the endoscopic image information used for estimation are input into the estimation model to obtain the pain level of the subject corresponding to the insertion status information and the endoscopic image information used for estimation. This level is used as the estimation result of the subject's pain level. in, The estimation model is trained using training data, which includes pre-collected insertion status information and endoscopic image information, as well as classification labels representing the pain level corresponding to the insertion status information and endoscopic image information.

13. A recording medium having a program recorded thereon that causes a computer to perform the following processes: Obtaining estimation examination status information, which includes estimation insertion status information, and one of estimation air supply information, estimation rigidity variable part operation information, estimation usage count information, estimation insertion part rigidity information, or estimation insertion length information, wherein the estimation insertion status information includes at least one of estimation insertion shape information related to the insertion shape of the insertion part of the endoscope inserted into the patient's body during endoscopy, and estimation operating force information related to the force applied to the insertion part during the endoscopy; and The estimation model is input with one of the following: the estimation insertion status information, the estimation air delivery information, the estimation rigid variable part movement information, the estimation usage frequency information, the estimation insertion part rigidity information, or the estimation insertion length information. The pain level of the subject corresponding to one of these three information is then used as the estimation result for the subject's pain level. in, The estimated gas supply information is related to the operational state of the gas supply unit that performs the action of supplying gas to the endoscope. The estimated motion information of the rigid variable part is related to the motion of the rigid variable part provided in the insertion part. The estimated usage information is related to the number of times the endoscope is used. The estimation of the stiffness information of the insertion part is related to the stiffness of the insertion part. The estimated insertion length information is related to the insertion length of the insertion part into the subject. The estimation model is trained using training data, which includes pre-collected insertion status information, and one of the following: estimation air delivery information, estimation rigid variable part movement information, estimation usage frequency information, estimation insertion part rigidity information, or estimation insertion length information; and a classification label representing the pain level corresponding to the insertion status information, the estimation air delivery information, the estimation rigid variable part movement information, the estimation usage frequency information, the estimation insertion part rigidity information, or the estimation insertion length information.

Images