Bone processing estimation device, learning device, bone processing estimation method, and learning model generation method

Abstract

The present invention implements a bone processing estimation device capable of appropriately estimating the position of a tool in a to-be-processed part of a bone. The bone processing estimation device (3) is provided with: a first acquisition unit (31) that acquires first information pertaining to a bone before processing; a second acquisition unit (32) that acquires second information that is time-series information sensed during processing of the bone; and an estimation unit (33) that estimates the position of the tool in the to-be-processed part of the bone on the basis of the first information and the second information.

Description

Bone modification estimation device, learning device, bone modification estimation method, and learning model generation method

[0001] The present invention relates to a bone modification estimation device, a learning device, a bone modification estimation method, and a learning model generation method.

[0002] Patent Document 1 discloses a surgical drilling device having a drill. The surgical drilling device outputs a signal related to the drilling status based on changes in the load current of a rotary motor, and when a penetration signal is output, the rotary motor is rotated in reverse to apply regenerative braking. This allows the surgical drilling device to control the drive of the cutting tool so as not to cut too much biological tissue.

[0003] Japanese Patent Application Publication No. 2012-161401

[0004] However, bone hardness varies depending on the patient's condition, and the torque applied to the cutting tool also varies depending on the bone hardness. Therefore, the above-mentioned conventional technology may not be able to properly determine whether the hole has been penetrated.

[0005] An object of one aspect of the present invention is to provide a bone processing estimation device that can appropriately estimate the remaining thickness of a processed portion of a bone, i.e., the position of a tool.

[0006] A bone processing estimation device according to one aspect of the present invention is configured to include a first acquisition unit that acquires first information regarding the bone before processing, a second acquisition unit that acquires second information which is time-series information sensed during processing of the bone, and an estimation unit that estimates the position of a tool in the processed portion of the bone based on the first information and the second information.

[0007] A learning device according to one aspect of the present invention is configured to generate a learning model using training data in which first information about a bone before processing and second information, which is time-series information sensed during processing of the bone, are input, and third information about the position of a tool in the processed portion of the bone is output.

[0008] A bone processing estimation method according to one aspect of the present invention includes a first acquisition step of acquiring first information about the bone before processing, a second acquisition step of acquiring second information which is time-series information sensed during processing of the bone, and an estimation step of estimating the position of a tool in the processed portion of the bone based on the first information and the second information.

[0009] A learning model generation method according to one aspect of the present invention is a method including an acquisition step of acquiring training data that uses as input first information about a bone before processing and second information, which is time-series information sensed during processing of the bone, and outputs third information about the position of a tool in the processed portion of the bone, and a learning step of generating a learning model using the training data.

[0010] According to one aspect of the present invention, the remaining thickness of the processed portion of the bone, i.e., the position of the tool, can be appropriately estimated.

[0011] FIG. 1 is a diagram showing multiple stages of processing a bone. FIG. 2 is a block diagram showing the configuration of a bone processing system according to an embodiment of the present invention. FIG. 3 is a diagram showing data related to bones and data sensed during bone processing. FIG. 4 is a flow diagram showing the processing and estimation process by the bone processing system. FIG. 5 is a diagram showing the processing of a bone sample when obtaining second information to be used as training data. FIG. 6 is a diagram showing the processing of a bone sample when obtaining second information to be used as training data.

[0012] [Embodiment 1] Fig. 1 is a diagram showing multiple stages in processing bone. Fig. 1 shows a longitudinal cross section of a portion of bone. Bone 10 has cortical bone 11 covering the outside and cancellous bone 12 covered by cortical bone 11. Cortical bone 11 is dense and relatively hard. Cancellous bone 12 includes a medullary cavity and is softer than cortical bone 11. A surgical procedure is performed in which bone 10 is milled with a cutting tool 21 (tool). The rotating cutting tool 21 is advanced in the axial direction indicated by the arrow to form a hole in bone 10.

[0013] Reference numeral 101 indicates a state in which the cutting tool 21 has not yet come into contact with the bone 10. Reference numeral 102 indicates a state in which the cutting tool 21 is cutting the cortical bone 11 on the front side. Reference numeral 103 indicates a state in which the cutting tool 21 is cutting the cancellous bone 12. Reference numeral 104 indicates a state in which the cutting tool 21 is cutting the cortical bone 11 on the back side. Reference numeral 105 indicates a state in which a hole has been drilled. Depending on the purpose of the surgical operation, a hole may be drilled as indicated by reference numeral 105, or a portion of the bone 10 may be thinned and left as indicated by reference numeral 104.

[0014] Thinning and leaving a portion of the bone 10 is a technically difficult procedure. If too much bone 10 is removed, a hole will form, but it has not been possible to accurately determine the remaining thickness of the processed portion of the bone 10. For this reason, thinning has traditionally been a procedure that relied on the skills of an experienced surgeon.

[0015] Furthermore, when forming a through-hole in the bone 10, if the bone 10 is scraped away too much, the tissue surrounding the bone 10 will be damaged.

[0016] The bone processing estimation device of this embodiment assists such surgical operations by estimating the remaining thickness of the processed portion of the bone 10 during processing.

[0017] (Configuration of bone processing system) Fig. 2 is a block diagram showing the configuration of a bone processing system 1 according to this embodiment. The bone processing system 1 includes a robot 2 and a bone processing estimation device 3. The bone processing system 1 is a system that processes a patient's bone 10 in collaboration with a surgeon or automatically.

[0018] The robot 2 is capable of moving the cutting tool in at least one axial direction. Here, the robot 2 is an articulated robot having multiple joints (axes). The robot 2 includes a cutting tool 21, a motor 22, a motor control unit 23, a motion control unit 24, and a sensor 25. The cutting tool 21 is, for example, a surgical drill or a surgical end mill that cuts the bone 10.

[0019] The motor 22 is a motor that rotates the cutting tool 21. The motor 22 detects the current flowing through the motor 22 in accordance with the torque, and outputs a load signal indicating the magnitude of the current to the motor control unit 23.

[0020] The motor control unit 23 drives the motor 22. The motor control unit 23 outputs rotational direction load information indicating the torque applied to the cutting tool 21 by the motor 22 to the bone processing estimation device 3 based on the load signal.

[0021] The motion control unit 24 controls the motion of each joint of the robot 2 and the cutting tool 21. The motion control unit 24 acquires information indicating the estimated remaining thickness of the processed portion of the bone 10 from the bone processing estimation device 3. The motion control unit 24 determines the speed at which the cutting tool 21 is moved in the axial direction of the cutting tool 21 and the number of rotations of the cutting tool 21 based on the information indicating the estimated remaining thickness of the processed portion of the bone 10. The motion control unit 24 instructs the motor control unit 23 on the number of rotations at which to operate the motor 22. The motion control unit 24 also controls each joint of the robot 2 that moves the cutting tool 21.

[0022] The sensor 25 senses physical quantities related to the processing of the bone 10 during processing of the bone 10. The sensor 25 outputs information indicating the sensed physical quantities to the bone processing estimation device 3. Here, the sensor 25 is a force sensor that senses the force applied to the cutting tool 21 in the axial direction of the cutting tool 21. The force applied in the axial direction of the cutting tool 21 is, in other words, a force applied to the cutting tool 21 in the direction of travel of the cutting tool 21 during processing of the bone 10. The sensor 25 outputs axial load information indicating the force applied in the axial direction of the cutting tool 21 to the bone processing estimation device 3.

[0023] The bone processing estimation device 3 (learning device) includes a first acquisition unit 31, a second acquisition unit 32, an estimation unit 33, an output unit 34, and a learning unit 35.

[0024] The first acquisition unit 31 acquires first information about the bone 10 before processing. Here, the first information includes information about CT values ​​(pixel values) obtained by computed tomography (CT) of the bone 10 before processing. For example, the first acquisition unit 31 acquires data on the distribution of CT values ​​(part of a CT image) of the planned processing location of the bone 10, which has been measured in advance, from an external information processing device. The first acquisition unit 31 outputs the first information to the estimation unit 33.

[0025] The second acquisition unit 32 acquires second information, which is time-series information sensed during processing of the bone 10. Here, the second acquisition unit 32 acquires, as the second information, rotational direction load information from the motor control unit 23 and axial direction load information from the sensor 25 at predetermined time intervals (at a predetermined sampling period). As a result, the second acquisition unit 32 obtains time-series rotational direction load information and axial direction load information. The second acquisition unit 32 outputs the second information to the estimation unit 33.

[0026] The estimation unit 33 estimates the remaining thickness of the processed portion of the bone 10 during processing based on the first information and the second information. The estimation unit 33 has a trained learning model 36. The learning model 36 is a learning model that receives the first information and the second information as input and outputs the remaining thickness of the processed portion of the bone 10. The estimation unit 33 inputs the first information and the second information into the learning model 36 and obtains the remaining thickness of the processed portion of the bone 10 as output. The estimation unit 33 outputs information indicating the estimated remaining thickness of the processed portion of the bone 10 to the output unit 34.

[0027] The output unit 34 outputs information indicating the estimated remaining thickness of the processed portion of the bone 10 to the robot 2. The output unit 34 may notify the user of the estimated remaining thickness of the processed portion of the bone 10 via a display device or a speaker. When the estimated remaining thickness of the processed portion of the bone 10 is equal to or less than the target thickness, the output unit 34 may notify the user that the remaining thickness of the processed portion has reached the target thickness.

[0028] The learning unit 35 generates a learning model 36 using training data that has as input the first information and the second information and as output the third information related to the remaining thickness of the processed portion of the bone 10. The learning unit 35 passes the generated learning model 36 to the estimation unit 33.

[0029] (Learning process) Figure 3 shows data related to the bone 10 and data sensed during machining of the bone 10. The left vertical axis of the upper graph indicates the force (advancement resistance) (N) applied to the cutting tool 21 in the direction of travel of the cutting tool 21. The left vertical axis of the lower graph indicates the torque (N·m) applied to the cutting tool 21 in the direction of the rotation axis. The right vertical axis indicates the CT value (HU) of the processed portion of the bone 10 before machining. The horizontal axis indicates the position (advancement distance) (mm) of the tip of the cutting tool 21. The dotted lines indicate the advancement resistance and torque. The solid line indicates the CT value.

[0030] Sections 301 to 305 in Figure 3 correspond to stages 101 to 105 in Figure 1, respectively. Section 301 is a state in which the cutting tool 21 has not yet come into contact with the bone 10. Section 302 is a state in which the cutting tool 21 is cutting the cortical bone 11 on the front side. Section 303 is a state in which the cutting tool 21 is cutting the cancellous bone 12. Section 304 is a state in which the cutting tool 21 is cutting the cortical bone 11 on the back side. Section 305 is a state in which the hole has penetrated and the tip of the cutting tool 21 has emerged on the back side of the bone 10. The boundary between Section 304 and Section 305 corresponds to the time when the tip of the cutting tool 21 has penetrated the cortical bone 11 on the back side, i.e., the time when the hole has penetrated. When the cutting tool 21 is cutting the cortical bone 11, both the resistance in the direction of travel and the torque are large. When the cutting tool 21 is cutting the cancellous bone 12 , the resistance in the direction of advance and the torque are smaller than when the cutting tool 21 is cutting the cortical bone 11 .

[0031] The bone processing system 1 performs learning in advance using training data to generate a learning model 36. The first information included in the training data is obtained, for example, by performing CT imaging on a plurality of bone 10 samples before processing.

[0032] The first information is information about the distribution of CT values ​​in the direction of travel of the cutting tool 21 (the direction in which the bone is machined). Each CT value along the direction of travel may be the CT value at the center of the machining area perpendicular to the direction of travel, or may be the average value of the CT values ​​in the machining area perpendicular to the direction of travel. The CT value of the bone 10 is closely related to bone density. Bone density is closely related to the strength of the bone 10. The resistance and torque in the direction of travel applied to the cutting tool 21 may change depending on the strength (hardness) of the bone 10.

[0033] The second time-series information included in the training data is obtained by processing multiple bone 10 samples and sensing the resistance and torque in the forward direction during processing. Because the robot 2 can know the position of the tip of the cutting tool 21 while processing the samples, the resistance and torque in the forward direction during processing of the samples can be correlated with the position of the tip of the cutting tool 21 at that time. In other words, data can be obtained in which the axes (positions) of the second time-series information and the first time-series information are aligned. Note that processing of the bone 10 samples can be performed while the bone 10 is fixed more firmly than during surgery to prevent displacement. Therefore, the relationship between the position of the cutting tool 21 and the bone 10 can be accurately obtained.

[0034] The sensed data used as the second information may be subjected to appropriate preprocessing such as noise removal, etc. For example, data obtained by calculating a moving average of the sensed data or data obtained by processing the sensed data with a low-pass filter may be used as the second information.

[0035] The training data is obtained by associating the first information and the second information as input with the third information relating to the remaining thickness of the processed portion of the bone 10 as output. As shown in Fig. 3, one training data set includes the second information in time series, the first information associated with the second information, and the third information associated with the second information. For example, one training data set includes the forward resistance and torque at each time point, the CT value associated with the position at each time point, and the remaining thickness associated with the position at each time point.

[0036] For example, if the goal is to penetrate a hole, the user may set the target thickness to 0. In this case, the data in sections 301 to 304 of the training data are labeled "non-penetration," and the data in section 305 is labeled "penetration." The point of penetration (the boundary between sections 304 and 305) can be identified, for example, by capturing an image of the backside of the sample being processed with an imaging device and observing the moment the cutting tool 21 appears. Note that the remaining thickness corresponding to each time point may be determined based on the point of penetration. The robot 2 has data indicating the position of the tip of the cutting tool 21 at each time point. Therefore, the remaining thickness corresponding to each time point can be determined by setting the thickness at the time of penetration to 0. The user may assign a numerical value for the remaining thickness as a label (output) to the data in the training data at each time point. The data in section 305 corresponds to a state in which the tip of the cutting tool 21 protrudes from the back surface of the bone 10. In section 305, the remaining thickness is a negative value. For example, data indicating that the remaining thickness is −1 mm corresponds to a state in which the tip of the cutting tool 21 is positioned protruding 1 mm from the surface of the back side of the bone 10. The remaining thickness, which is the label (third information) of the data, may include a negative numerical value.

[0037] The learning unit 35 performs learning using multiple pieces of training data obtained from multiple samples, and generates a trained learning model 36. As a learning algorithm, for example, an RNN or the like can be used, but this is not limiting and any known learning algorithm can be used. The estimation unit 33 acquires the generated learning model 36 from the learning unit 35.

[0038] For example, the learning unit 35 may receive the first information and the second information for the period from time t' to time t as input, and output the third information for time t, where t'<t.

[0039] (Estimation Process) Figure 4 is a flow diagram showing the flow of the processing and estimation process by the bone processing system 1. The first acquisition unit 31 acquires first information about the bone 10 before processing (S1). The robot 2 starts processing the bone 10 in response to an instruction from the surgeon (S2). During processing of the bone 10, the second acquisition unit 32 acquires sensed second information (S3).

[0040] The estimation unit 33 inputs the first information and the second information into the learning model 36, and obtains information on the remaining thickness of the processed portion of the bone 10 being processed as an output from the learning model 36. In this way, the estimation unit 33 estimates the remaining thickness of the processed portion of the bone 10 (S4).

[0041] The estimation unit 33 estimates whether the remaining thickness of the processed portion of the bone 10 has reached the target thickness (S5). For example, if the learning model 36 outputs "penetration," the estimation unit 33 estimates that the target thickness (= 0) has been reached. If the output indicates "non-penetration," the estimation unit 33 estimates that the target thickness (= 0) has not been reached. Alternatively, if the estimated remaining thickness of the processed portion of the bone 10 is equal to or less than the target thickness (≠ 0), the estimation unit 33 estimates that the target thickness has been reached, i.e., that the remaining thickness of the processed portion of the bone 10 has been thinned to the target thickness. If the estimated remaining thickness of the processed portion of the bone 10 is greater than the target thickness (≠ 0), the estimation unit 33 estimates that the target thickness has not been reached.

[0042] If the remaining thickness of the processed portion of the bone 10 has not reached the target thickness (No in S5), the operation control unit 24 causes the robot 2 to continue processing the bone 10. The process flow returns to S3, and the second acquisition unit 32 acquires the second information at the next time. The estimation unit 33 estimates the remaining thickness of the processed portion of the bone 10 based on the newly acquired second information at the next time (S4).

[0043] When the remaining thickness of the processed portion of the bone 10 reaches the target thickness (Yes in S5), the operation control unit 24 causes the robot 2 to stop processing the bone 10 (S6). Specifically, the operation control unit 24 stops the advancement and rotation of the cutting tool 21.

[0044] The bone processing system 1 can estimate whether the remaining thickness of the processed portion of the bone 10 has reached a target thickness based on first information about the bone 10 before processing and second information about physical quantities sensed during processing. The bone processing estimation device 3 has a learning model 36 that is trained using the second information in association with the density or strength of the bone 10 before processing. Therefore, the bone processing estimation device 3 can estimate the remaining thickness of the processed portion of the bone 10 with higher accuracy depending on the condition of the bone 10, which varies from patient to patient. Therefore, the bone processing system 1 can reduce damage to tissue adjacent to the bone 10. Alternatively, the bone processing system 1 can thin a portion of the bone 10 to a desired thickness.

[0045] Estimating the remaining thickness of the processed portion of the bone 10 means estimating the position of the cutting tool 21 in the processed portion of the bone 10. It can be said that the estimation unit 33 estimates the position (tip position) of the cutting tool 21 in the bone 10 being processed based on the first information and the second information. For example, if it is desired to form a hole that is completely through, the target position of the cutting tool 21 may be a position where the tip of the cutting tool 21 protrudes a predetermined distance from the back surface of the bone 10. In this case, the target thickness of the processed portion of the bone 10 becomes a negative value (= -predetermined protruding distance). For example, the estimation unit 33 may estimate the distance where the tip of the cutting tool 21 protrudes from the back surface of the bone 10.

[0046] (Variation) The estimation unit 33 and the learning unit 35 may use, as the second information, time-series information sensed during the processing of the bone 10, vibrations of the bone 10 caused by the processing, vibrations of the cutting tool 21, vibrations of the robot 2, sounds caused by the processing, or images of the processed portion. The vibrations and sounds generated may differ depending on whether hard cortical bone 11 or soft cancellous bone 12 is being cut. Furthermore, the amount or scattering of powder (bone powder) or bleeding caused by the processing may differ depending on whether the processed portion is cortical bone 11 or cancellous bone 12. Therefore, the learning unit 35 can generate a learning model 36 that estimates the remaining thickness of the processed portion of the bone 10 by using time-series images including powder or bleeding for learning. The sensor for sensing sound or the imaging device for acquiring images may be provided on the robot 2 or may be located at a location separate from the robot 2. The sensor for sensing the vibration of the bone 10 may be placed on the patient.

[0047] The estimation unit 33 and the learning unit 35 may use, as the second information, a force acting on the cutting tool 21 in any direction. That is, the estimation unit 33 and the learning unit 35 may use, as the second information, a force acting on the cutting tool 21 in a direction other than the direction of travel of the cutting tool 21 or a force in a direction other than the direction of the rotation axis. In the example of the above-described embodiment, the axial direction of the cutting tool 21 and the direction of travel of the cutting tool 21 coincide with each other, but they do not have to coincide with each other. For example, an end mill may be used as the cutting tool 21, and the cutting tool 21 may be advanced in a direction different from the axial direction of the cutting tool 21.

[0048] In the above-described embodiment, the estimation unit 33 and the learning unit 35 used two types of information (traveling direction resistance and torque) as the second information, but they may use only one type of information, or three or more types of information.

[0049] The estimation unit 33 and the learning unit 35 may use, as the first information about the bone 10 before processing, information about the strength of the bone 10, information about the density of the bone 10, information about the quality of the bone 10, information about the bone density of the bone 10, information about the CT value, information about the age, gender, height, or weight of the patient who owns the bone 10, or background information about the patient who owns the bone 10. The strength of the bone 10 is mainly determined by the density and quality of the bone 10. For example, the bone density information may be information about the distribution of bone density in the direction in which the bone is processed, a representative value of the bone density of the bone 10, or a representative value of the bone density of other bones of the patient. The bone density of one bone is correlated with the bone density of other bones. The patient's age, gender, height, and weight also correlate to some extent with bone density. Age and gender are particularly highly correlated with bone density. The patient's age, gender, height, weight, and background information (presence or absence of a specific chronic illness) are also related to the quality of the bone 10. For example, the amount of collagen contained in the bone 10 affects the quality of the bone 10. Even if the amount of bone mineral is the same, if there is less collagen binding them together, the strength of the bone 10 will decrease. The CT value is closely related to the bone density of the bone 10. For example, a representative value of the CT value in the cortical bone 11 may be used as the first information. The representative value of the CT value may be, for example, an average value of the CT values ​​in the processed region of the cortical bone 11. Furthermore, the thickness of the cortical bone 11 can be identified from the distribution of the CT values. For example, the thickness of the cortical bone 11 with high strength may be used as the first information.

[0050] The estimation unit 33 and the learning unit 35 may use only one of the above-mentioned pieces of information related to the density or strength of the bone 10 as the first information, or may use a combination of the above-mentioned pieces of information.

[0051] In the above-described embodiment, the training data includes the second information for sections 301 to 305 shown in FIG. 3 , but is not limited to this. The training data may include the second information for a portion of a section that includes the position of the target thickness. For example, the training data may include the second information for a portion of a section that includes the boundary between sections 304 and 305. For example, the second information for section 302 or section 303 may not be used for learning.

[0052] In the above-described embodiment, the training data was generated using the second information sensed during processing by the robot 2. However, this is not limiting. The learning unit 35 may also use the second information sensed during processing by a user, such as a doctor, as training data. For example, a doctor processes the bone 10 using a hand drill operated by hand. The hand drill includes a cutting tool 21, a motor 22, a motor control unit 23, and a sensor 25. The second acquisition unit 32 of the bone processing estimation device 3 acquires the second information sensed from the hand drill during processing. The learning unit 35 generates a learning model 36 using training data including the second information sensed during processing by the doctor. The bone processing estimation device 3 using the learning model 36 thus learned may estimate the remaining thickness of the processed portion of the bone 10 when another doctor processes the bone 10, or may estimate the remaining thickness of the processed portion of the bone 10 when the robot 2 processes the bone 10. This allows the skills and experiential knowledge of experienced doctors to be preserved as data (learning model) and passed on to younger doctors or robots.

[0053] The learning unit 35 may generate the learning model 36 using training data including both the second information sensed during processing by the doctor and the second information sensed during processing by the robot 2. The bone processing estimation device 3 using this learning model 36 may estimate the remaining thickness of the processed portion of the bone 10 when another doctor processes the bone 10, or may estimate the remaining thickness of the processed portion of the bone 10 when the robot 2 processes the bone 10.

[0054] Furthermore, as in the above-described embodiment, the bone processing estimation device 3 having a learning model 36 learned using the second information sensed during processing by the robot 2 may estimate the remaining thickness of the processed portion of the bone 10 when a doctor processes the bone 10.

[0055] In addition to the first information and the second information, the estimation unit 33 and the learning unit 35 may use parameters of the cutting tool 21 as input to the learning model 36. The parameters of the cutting tool 21 may include, for example, the tip shape, diameter, grain size, and / or usage time of the cutting tool 21. The type and condition of the cutting tool 21 may affect the second information, such as resistance in the direction of travel, torque, vibration, sound, or images. By using the parameters of the cutting tool 21 for learning, the estimation unit 33 can estimate the remaining thickness of the processed portion of the bone 10 with higher accuracy.

[0056] [Embodiment 2] Another embodiment of the present invention will be described below. For ease of explanation, components having the same functions as those described in the above embodiment will be denoted by the same reference numerals, and their description will not be repeated. The configuration of the bone processing system 1 of this embodiment is the same as that of embodiment 1.

[0057] 5 is a diagram showing the processing of a bone 10 sample when obtaining second information to be used as training data. A longitudinal cross section of a portion of the bone is shown in FIG. Reference numeral 501 indicates a state in which the cutting tool 21 has not yet come into contact with the bone 10. Here, the bone 10 sample is supported and fixed by a plurality of fixing parts 4.

[0058] The fixing portion 4 is deformable by the force applied by the cutting tool 21 during processing. The fixing portion 4 may be an elastic body that elastically deforms by the force applied by the cutting tool 21, or may be a material that plastically deforms. For example, the fixing portion 4 may be clay. The fixing portion 4 loosely fixes the bone 10 sample so as to allow displacement during processing.

[0059] Reference numeral 502 indicates a state in which the cutting tool 21 is cutting the cortical bone 11 on the rear side. In the state shown by reference numeral 502, the fixing part 4 is deformed by the force applied by the cutting tool 21, and the bone 10 sample is displaced in the direction of movement of the cutting tool 21.

[0060] In this way, the learning unit 35 uses as training data the second information sensed in a state allowing displacement of the bone 10. The learning unit 35 performs learning using as input the first information and the second information sensed in a state allowing displacement of the bone 10, and generates a learned learning model. The estimation unit 33 uses the learning model learned in this way to estimate the remaining thickness of the processed portion of the bone 10.

[0061] In an actual surgery, since the bone 10 is inside the patient's body, it is difficult to fix the bone 10 so that it does not move at all. During surgery, the bone 10 may be displaced. If the bone 10 is displaced in the direction of travel of the cutting tool 21, it is not possible to accurately calculate the remaining thickness of the processed portion of the bone 10 from the position coordinates of the cutting tool 21 that are grasped by the robot 2.

[0062] The training data in this embodiment includes second information obtained during processing of the bone 10 that is fixed in a state similar to the state of the bone 10 during actual surgery. The learning unit 35 performs learning using this training data. By using the learning model obtained in this manner, the estimation unit 33 can estimate the remaining thickness of the processed portion of the bone 10 with higher accuracy, taking into account the displacement of the bone 10.

[0063] The third information used as a label of the training data can be determined from the measured value of the displacement of the bone 10 during processing of the bone 10 sample and the position coordinates of the cutting tool 21 grasped by the robot 2. The displacement of the bone 10 can be measured by capturing an image using a distance measuring device or an imaging device.

[0064] [Embodiment 3] Another embodiment of the present invention will be described below. For ease of explanation, components having the same functions as those described in the above embodiment will be denoted by the same reference numerals, and their description will not be repeated. The configuration of the bone processing system 1 of this embodiment is the same as that of embodiment 1.

[0065] 6 is a diagram showing how a bone 10 sample is processed when obtaining second information to be used as training data. One end of the bone 10 sample is connected to a wall 5 by a wire 6. The other end of the bone 10 sample is connected to a weight 8 by a wire 7. The wire 7 is hung on a pulley 9. The bone 10 sample is pulled laterally by the weight of the weight 8. The bone 10 sample is loosely fixed so as to allow vertical displacement during processing by being pulled in a direction (lateral direction) intersecting the direction of travel of the cutting tool 21.

[0066] A bone 10 in a living body is connected to other bones or muscles by ligaments or tendons. In this embodiment, the bone 10 sample is fixed in a state that more closely resembles a bone 10 in a living body. The learning unit 35 uses the second information sensed in a state that allows displacement of the bone 10 as training data. Therefore, the estimation unit 33 can estimate the remaining thickness of the processed portion of the bone 10 with higher accuracy, taking into account the displacement of the bone 10.

[0067] For example, the force applied to the cutting tool 21 in the axial direction of the cutting tool 21 may vary depending on whether the bone 10 being machined is fixed so as to be displaceable. By using the force applied to the cutting tool 21 in the axial direction of the cutting tool 21, which is sensed in a state in which displacement of the bone 10 is permitted, as the second information, the estimation unit 33 can estimate the remaining thickness of the machined portion of the bone 10 with higher accuracy.

[0068] The fixing portion 4 of the second embodiment may be further added to support the bone 10 sample in the vertical direction as well. The bone 10 sample may be tensioned and fixed by an elastic body such as a spring or rubber, instead of the wire 7 and the weight 8.

[0069] [Example of implementation by software] The functions of the robot 2 and the bone processing estimation device 3 (hereinafter referred to as the "device") can be realized by a program for causing a computer to function as the device, and a program for causing a computer to function as each control block of the device (motor control unit 23, operation control unit 24, first acquisition unit 31, second acquisition unit 32, estimation unit 33, output unit 34, and learning unit 35).

[0070] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The functions described in each of the above embodiments are realized by executing the program using the control device and storage device.

[0071] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0072] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0073] [Summary] The bone processing estimation device according to aspect 1 of the present invention is configured to include a first acquisition unit that acquires first information about the bone before processing, a second acquisition unit that acquires second information that is time-series information sensed during processing of the bone, and an estimation unit that estimates the position of a tool in the processed portion of the bone based on the first information and the second information.

[0074] A bone processing estimation device according to a second aspect of the present invention may be configured in the above-mentioned first aspect so that the estimation unit estimates a remaining thickness of the processed portion.

[0075] A bone processing estimation device according to a third aspect of the present invention may be configured in accordance with the first or second aspect above, wherein the first information includes information relating to the strength of the bone.

[0076] A bone processing estimation device according to Aspect 4 of the present invention may be configured in any one of Aspects 1 to 3 above, wherein the first information includes information on the density of the bone.

[0077] A bone processing estimation device according to Aspect 5 of the present invention may be configured in any one of Aspects 1 to 4 above, wherein the first information includes information regarding the quality of the bone.

[0078] A bone processing estimation device according to a sixth aspect of the present invention may be configured in any one of the first to fifth aspects, wherein the first information includes information regarding a CT value obtained by computer tomography of the bone.

[0079] A bone processing estimation device according to Aspect 7 of the present invention may be configured in accordance with Aspect 6 above, wherein the first information includes information regarding the distribution of the CT values ​​in the direction in which the bone is processed.

[0080] A bone processing estimation device according to aspect 8 of the present invention may be configured such that, in any of aspects 1 to 7 above, the first information includes the age, sex, height, or weight of a patient having the bone, or background information of the patient.

[0081] A bone processing estimation device according to Aspect 9 of the present invention may be configured in any one of Aspects 1 to 8 above, wherein the second information includes a force applied to a tool that processes the bone.

[0082] A bone processing estimation device according to Aspect 10 of the present invention may be configured in any one of Aspects 1 to 9 above, wherein the second information includes a torque applied to a tool that processes the bone.

[0083] A bone processing estimation device according to aspect 11 of the present invention may be configured in any one of aspects 1 to 10 above, wherein the second information includes vibrations of a tool that processes the bone, vibrations of the bone, or sound.

[0084] A bone processing estimation device according to aspect 12 of the present invention may be configured in any one of aspects 1 to 11 above, wherein the second information includes an image including processing powder or bleeding caused by processing the bone.

[0085] A bone processing estimation device according to aspect 13 of the present invention may be configured such that, in any of aspects 1 to 12 above, the estimation unit performs estimation using a learning model trained using as input the first information and the second information sensed during processing of a bone fixed to allow displacement.

[0086] A bone processing estimation device according to aspect 14 of the present invention may be configured such that, in any of aspects 1 to 13 above, the estimation unit estimates whether the remaining thickness of the processed portion of the bone has been thinned to a target thickness.

[0087] The learning device according to aspect 15 of the present invention is configured to generate a learning model using training data that takes as input first information about a bone before processing and second information, which is time-series information sensed during processing of the bone, and outputs third information about the position of a tool in the processed portion of the bone.

[0088] A bone processing estimation method according to aspect 16 of the present invention is a method including a first acquisition step of acquiring first information about the bone before processing, a second acquisition step of acquiring second information which is time-series information sensed during processing of the bone, and an estimation step of estimating the position of a tool in the processed portion of the bone based on the first information and the second information.

[0089] The learning model generation method according to aspect 17 of the present invention is a method including an acquisition step of acquiring training data that uses as input first information about a bone before processing and second information which is time-series information sensed during processing of the bone, and outputs third information about the position of a tool in the processed portion of the bone, and a learning step of generating a learning model using the training data.

[0090] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0091] REFERENCE SIGNS LIST 1 Bone processing system 2 Robot 3 Bone processing estimation device (learning device) 4 Fixation unit 6, 7 Wire 8 Sinker 10 Bone 11 Cortical bone 12 Cancellous bone 21 Cutting tool 22 Motor 23 Motor control unit 24 Operation control unit 25 Sensor 31 First acquisition unit 32 Second acquisition unit 33 Estimation unit 34 Output unit 35 Learning unit 36 ​​Learning model

Claims

1. A bone processing estimation device comprising: a first acquisition unit that acquires first information regarding a bone before processing; a second acquisition unit that acquires second information which is time-series information sensed during processing of the bone; and an estimation unit that estimates the position of a tool on the processed portion of the bone based on the first information and the second information.

2. A bone processing estimation device according to claim 1, wherein the estimation unit estimates the remaining thickness of the processed portion.

3. A bone processing estimation device according to claim 1, wherein the first information includes information relating to the strength of the bone.

4. A bone processing estimation device according to claim 1, wherein the first information includes information relating to the density of the bone.

5. A bone processing estimation device according to claim 1, wherein the first information includes information relating to the quality of the bone.

6. A bone processing estimation device according to claim 1, wherein the first information includes information on a CT value obtained by computer tomography of the bone.

7. A bone processing estimation device according to claim 6, wherein the first information includes information relating to the distribution of the CT values ​​in the direction in which the bone is processed.

8. The bone processing estimation device according to claim 1, wherein the first information includes the age, sex, height, or weight of a patient having the bone, or background information of the patient.

9. A bone processing estimation device according to any one of claims 1 to 8, wherein the second information includes a force applied to a tool that processes the bone.

10. A bone processing estimation device according to any one of claims 1 to 8, wherein the second information includes a torque applied to a tool that processes the bone.

11. A bone processing estimation device according to any one of claims 1 to 8, wherein the second information includes vibrations of a tool that processes the bone, vibrations of the bone, or sound.

12. A bone processing estimation device according to any one of claims 1 to 8, wherein the second information includes an image including processing powder or bleeding resulting from processing the bone.

13. A bone processing estimation device according to any one of claims 1 to 8, wherein the estimation unit performs estimation using a learning model trained using as input the first information and the second information sensed during processing of a bone fixed to allow displacement.

14. A bone processing estimation device according to any one of claims 1 to 8, wherein the estimation unit estimates whether the remaining thickness of the processed portion of the bone has been thinned to a target thickness.

15. A learning device that generates a learning model using training data that takes as input first information about a bone before processing and second information which is time-series information sensed during processing of the bone, and outputs third information about the position of a tool in the processed portion of the bone.

16. A bone processing estimation method comprising: a first acquisition step of acquiring first information about a bone before processing; a second acquisition step of acquiring second information which is time-series information sensed during processing of the bone; and an estimation step of estimating the position of a tool on a processed portion of the bone based on the first information and the second information.

17. A learning model generation method comprising: an acquisition step of acquiring training data that uses as input first information about a bone before processing and second information which is time-series information sensed during processing of the bone, and outputs third information about the position of a tool in the processed portion of the bone; and a learning step of generating a learning model using the training data.

Images