Information processing method, program, information processing device, model generation method, and balloon catheter manufacturing method

Abstract

Provided are an information processing method and the like that make it possible to easily measure the strength of a connection part of a balloon and a shaft constituting a balloon catheter. According to said method, a computer acquires spectral data of light from a connection part generated via the connection part of the balloon and the shaft constituting the balloon catheter, or a connection part of the shaft which is continuous therewith, being irradiated with near infrared light. Subsequently, the computer determines information regarding the connection strength of the connection part on the basis of the acquired spectral data.

Description

Information processing method, program, information processing device, model generation method, and balloon catheter manufacturing method

[0001] The present disclosure relates to an information processing method, a program, an information processing device, a model generation method, and a balloon catheter manufacturing method.

[0002] Intravascular treatment is performed as a treatment method for lesions such as stenosis or occlusion in blood vessels, in which a treatment device such as a balloon catheter or a stent is percutaneously inserted into the blood vessel to dilate the lesion. Patent Document 1 discloses a technology for manufacturing a balloon catheter with low sliding resistance and excellent operability.

[0003] International Publication No. 2010 / 131554

[0004] Balloon catheters are constructed by connecting multiple shaft tubes and balloons by fusion, welding, adhesives, or other methods. Because catheters are inserted into narrow, complexly curved blood vessels to reach lesions, the connections between the shaft tubes and balloons must be strong. Therefore, there is a need for a technology that can easily measure the connection strength of the connections, for example, during the assembly process of balloon catheters.

[0005] In one aspect, an object of the present invention is to provide an information processing method and the like that can easily measure the strength of the connection between the shaft and the balloon that constitutes a balloon catheter.

[0006] (1) The present disclosure relates to an information processing method in which a computer acquires spectroscopic data of light generated by irradiating a connecting portion of a shaft and a balloon that constitute a balloon catheter, or a connecting portion of a continuous shaft, with near-infrared light, and determines information regarding the connection strength of the connecting portion based on the acquired spectroscopic data.

[0007] (2) In the information processing method of (1) above, it is preferable that the computer further performs a process of performing multivariate analysis on the spectroscopic data to obtain a distribution of each wavelength component contained in the light from the connecting portion, and identifying information regarding the connection strength of the connecting portion based on the distribution of each wavelength component.

[0008] (3) In the information processing method of (2) above, it is preferable that the computer further executes a process of identifying information regarding the connection strength of the connection portion based on the distribution of a specific wavelength component in the distribution of each wavelength component.

[0009] (4) In the information processing method of (2) above, it is preferable that the computer further executes a process of inputting the distribution of each wavelength component obtained by performing multivariate analysis on the spectroscopic data into a learning model that has been trained to output information regarding the connection strength of the connection part when the distribution of each wavelength component contained in the light from the connection part is input, and acquiring information regarding the connection strength of the connection part.

[0010] (5) In the information processing method of (4) above, it is preferable that a plurality of the learning models corresponding to a plurality of the connection parts are stored in a memory unit, and the computer further executes a process of identifying the learning model corresponding to the position of the connection part and using the identified learning model to obtain information regarding the connection strength of the connection part.

[0011] (6) In the information processing method described in any one of (1) to (5) above, it is preferable that the computer further executes preprocessing on the spectroscopic data, including normalization, smoothing, and differentiation.

[0012] (7) In the information processing method described in any one of (1) to (6) above, it is preferable that the computer further executes a process of acquiring the spectroscopic data using a near-infrared spectrometer or a hyperspectral camera.

[0013] (8) The present disclosure also provides a program that causes a computer to execute a process of acquiring spectroscopic data of light generated from a connecting portion of a shaft and a balloon that constitute a balloon catheter, or a connecting portion of a continuous shaft, when near-infrared light is irradiated onto the connecting portion, and identifying information regarding the connection strength of the connecting portion based on the acquired spectroscopic data.

[0014] (9) The present disclosure also provides an information processing device including a control unit, which acquires spectroscopic data of light from a connecting portion of a shaft and a balloon that constitute a balloon catheter, or a connecting portion of a continuous shaft, generated by irradiating the connecting portion with near-infrared light, and identifies information regarding the connection strength of the connecting portion based on the acquired spectroscopic data.

[0015] (10) The present disclosure also provides a model generation method in which a computer executes a process to acquire training data correlating the distribution of wavelength components contained in light from a connecting portion of a shaft and a balloon that constitutes a balloon catheter, or a connecting portion of a continuous shaft, generated by irradiating the connecting portion with near-infrared light, with information regarding the connection strength of the connecting portion, and to generate a learning model that outputs information regarding the connection strength of the connecting portion when the distribution of wavelength components contained in light from the connecting portion is input using the acquired training data.

[0016] (11) The present disclosure also provides a balloon catheter manufacturing method including the steps of: connecting a shaft and a balloon for a balloon catheter, and / or connecting consecutive shafts; irradiating a connecting portion of the shaft and the balloon, and / or connecting portions of the consecutive shafts, with near-infrared light using a light-emitting unit; acquiring spectroscopic data of the light from the connecting portion; identifying information relating to the connection strength of the connecting portion based on the acquired spectroscopic data; and determining whether the identified information relating to the connection strength is equal to or greater than a reference value.

[0017] In one aspect, the strength of the connection between the shaft and the balloon that constitutes the balloon catheter can be easily measured.

[0018] FIG. 1 is a block diagram showing an example of the configuration of a measurement system. FIG. 2 is an explanatory diagram of a balloon catheter to be measured. FIG. 3 is an explanatory diagram showing an example of a record layout of an analysis result DB. FIG. 4 is an explanatory diagram showing an example of the configuration of a learning model. FIG. 5 is an explanatory diagram of an example of the configuration of a learning model. FIG. 6 is an explanatory diagram of the component analysis process. FIG. 7 is an explanatory diagram of the component analysis process. FIG. 8 is an explanatory diagram of the component analysis process. FIG. 9 is an explanatory diagram of the component analysis process. FIG. 10 is a flowchart showing an example of a processing procedure for generating a learning model. FIG. 11 is a flowchart showing an example of a processing procedure for estimating connection strength.

[0019] Hereinafter, the information processing method, program, information processing device, model generation method, and balloon catheter manufacturing method of the present disclosure will be described in detail with reference to the drawings showing embodiments thereof.

[0020] (Embodiment 1) An information processing device for measuring the connection strength of a balloon catheter configured by connecting multiple shaft tubes and balloons will be described. The balloon catheter in this embodiment is a catheter that is inserted into a hollow organ such as a blood vessel, bile duct, pancreatic duct, bronchi, esophagus, intestine, or urethra. In the balloon catheter, the connection between the shaft tube and balloon, and the connection between successive shaft tubes, are connected by a method such as fusion, welding, or adhesive bonding.

[0021] 1 is a block diagram showing an example of the configuration of a measurement system. The measurement system of this embodiment includes an information processing device 10 and a spectroscopic device 20, which are communicatively connected by wired communication via a cable or wireless communication. The information processing device 10 and the spectroscopic device 20 may also be communicatively connected via a network, which may be the Internet, a public communication line, or a LAN (Local Area Network) constructed within a facility where the measurement system is installed.

[0022] The spectroscopic device 20 is a device that irradiates a measurement object (in this embodiment, the connecting portion between the shaft and balloon that constitute the balloon catheter, and the connecting portion between the continuous shafts) with near-infrared light, receives light from the measurement object, and acquires spectroscopic data (spectral data) by decomposing the received light into wavelengths. For example, a near-infrared spectroscopic device, a hyperspectral camera, or the like can be used. The spectroscopic device 20 includes a light-emitting unit 21, a light-receiving unit 22, a signal processing unit 23, and a communication unit 24. The light-emitting unit 21 emits near-infrared light having a wavelength in the range of approximately 750 nm to 2500 nm. The light-receiving unit 22 receives light (reflected light or transmitted light) from the measurement object of the near-infrared light emitted by the light-emitting unit 21. The spectroscopic device 20 may be a reflection-type spectroscopic device that receives reflected light from the measurement object of the near-infrared light irradiated by the light-emitting unit 21, or a transmission-type spectroscopic device that receives transmitted light of the irradiated near-infrared light that passes through the measurement object. The signal processing unit 23 generates spectral data by decomposing the light received by the light receiving unit 22 into wavelengths. The communication unit 24 is a communication module for directly communicating with the information processing device 10 and transmits the spectral data generated by the signal processing unit 23 to the information processing device 10. Note that the communication unit 24 may also be a communication module for connecting to a network, in which case the spectral data is transmitted to the information processing device 10 via the network.

[0023] In addition to the above-described configuration, the spectroscopic device 20 may be configured to include an input unit that accepts user input and a display unit such as a liquid crystal display or an organic EL display. The spectroscopic device 20 may also be configured to include a polarizing plate on the light-receiving surface side of the light-receiving unit 22, and to receive light (reflected light or transmitted light) from the measurement object through the polarizing plate. Because the spectroscopic data is proportional to the absorbance of the measurement object, qualitative and quantitative evaluations of the characteristics of the measurement object can be performed from the obtained spectroscopic data (the absorbance indicated by the spectroscopic data). In this embodiment, the information processing device 10 performs qualitative and quantitative evaluations of the coupling strength of the measurement object based on the spectroscopic data acquired by the spectroscopic device 20.

[0024] The information processing device 10 is a device capable of various information processing and information transmission and reception, and is composed of, for example, a personal computer, a server computer, etc. The information processing device 10 has a control unit 11, a memory unit 12, a communication unit 13, an input unit 14, a display unit 15, a reading unit 16, etc., and these units are connected via a bus. The control unit 11 has one or more processors, such as a central processing unit (CPU), a micro-processing unit (MPU), a graphics processing unit (GPU), a general-purpose computing on graphics processing units (GPGPU), or a tensor processing unit (TPU). The control unit 11 appropriately executes a program P stored in the memory unit 12 to perform various information processing, control processing, etc., that the information processing device 10 should perform. Note that if the control unit 11 has multiple processors, each process may be executed by the same processor, or each process may be executed by a different processor.

[0025] The storage unit 12 includes RAM (Random Access Memory), flash memory, a hard disk, an SSD (Solid State Drive), etc. The storage unit 12 pre-stores the program P (program product) executed by the control unit 11 and various data necessary for executing the program P. The storage unit 12 temporarily stores data generated when the control unit 11 executes the program P. The storage unit 12 also stores multiple learning models M, each of which has been trained with training data through, for example, machine learning, for each of the connecting portions of the balloon catheter. The learning models M are machine learning models trained with predetermined training data. For example, the learning models M receive input information indicating the results of component analysis (spectral analysis) of spectroscopic data acquired by the spectroscopic device 20 and output data regarding the connection strength of the measurement object (the measurement object that reflects or transmits light related to the spectroscopic data). The learning models M are expected to be used as program modules constituting artificial intelligence software. The learning models M perform predetermined calculations on input values ​​and output the calculation results. The storage unit 12 stores data, such as coefficients and thresholds of functions that define the calculations, as the learning models M. The storage unit 12 also stores an analysis result DB 12 a. The learning model M and / or the analysis result DB 12 a may be stored in another storage device connected to the information processing device 10 or in another storage device with which the information processing device 10 can communicate.

[0026] The communication unit 13 has a configuration similar to the communication unit 24 of the spectroscopic device 20, and receives spectroscopic data acquired by the spectroscopic device 20 from the spectroscopic device 20. The communication unit 13 may be configured to receive spectroscopic data from the spectroscopic device 20 via a network. The input unit 14 accepts operation input by a user and sends a control signal corresponding to the operation content to the control unit 11. The display unit 15 is a liquid crystal display, an organic EL display, or the like, and displays various information in accordance with instructions from the control unit 11. The input unit 14 and the display unit 15 may be a touch panel configured as an integrated unit. The input unit 14 and the display unit 15 are not essential, and the information processing device 10 may be configured to accept operations through a connected terminal device and output information to be displayed to an external display device.

[0027] The reading unit 16 reads information stored in a portable storage medium 10a, such as a CD (Compact Disc), a DVD (Digital Versatile Disc), a USB (Universal Serial Bus) memory, or an SD (Secure Digital) card. The program P and various data stored in the storage unit 12 may be read by the control unit 11 from the portable storage medium 10a via the reading unit 16 and stored in the storage unit 12. The program P and various data may be written to the storage unit 12 during the manufacturing stage of the information processing device 10, or may be downloaded by the control unit 11 from another device via the communication unit 13 and stored in the storage unit 12.

[0028] The information processing device 10 is not limited to a single computer, but may be a multi-computer including multiple computers. Furthermore, the information processing device 10 may be a virtual machine virtually constructed within a single device by software, or may be a cloud server. In the following description, the information processing device 10 is described as a single computer. Furthermore, the program P may be executed on a single computer, or on multiple computers interconnected via a network.

[0029] 2A and 2B are explanatory diagrams of a balloon catheter to be measured. The balloon catheter has a branched hub 31, a kink-resistant protector 32, and a shaft 33. The branched hub 31 and the shaft 33 are connected via the kink-resistant protector 32. The shaft 33 is configured by connecting shaft tubes 33a-33e, which are resin tubes manufactured by extrusion molding, for example, to a balloon 34, which is expanded by blowing air into the resin tube from the inside. Specifically, the first shaft tube 33a and one end of the balloon 34 are connected by a first connecting portion C1, and the other end of the balloon 34 and one end of the second shaft tube 33b are connected by a second connecting portion C2. An inner tube 33c is inserted into the balloon 34 and the second shaft tube 33b, and one end of the inner tube 33c is connected to the first shaft tube 33a. The other ends of the second shaft tube 33b and the inner tube 33c are connected to one end of the third shaft tube 33d by a third connecting portion C3. At the third connecting portion C3, the other end of the second shaft tube 33b is partially open, and a reinforcing member 36 reinforcing the opening 35 is provided inside the second shaft tube 33b and the third shaft tube 33d. The other end of the third shaft tube 33d is connected to one end of the fourth shaft tube 33e by a fourth connecting portion C4. Furthermore, the other end of the fourth shaft tube 33e is connected to the anti-kink protector 32. The first connecting portion C1 to the third connecting portion C3 are connected to each other by, for example, a fusion process, and the fourth connecting portion C4 is connected to each other by, for example, a welding process. Therefore, in this embodiment, the tip fusion strength at the first connecting portion C1, the base fusion strength at the second connecting portion C2, the opening fusion strength at the third connecting portion C3, and the shaft fusion strength at the fourth connecting portion C4 are measured (estimated).

[0030] The inner tube 33c is provided with contrast markers 37, which do not transmit X-rays, at appropriate locations within the region where the balloon 34 is inserted. For example, the positions of the contrast markers 37 are visualized in fluoroscopic images taken during catheterization, allowing the position of the balloon 34 to be determined. Furthermore, depth markers 38, which do not transmit X-rays, are provided with appropriate locations on the fourth shaft tube 33e. The positions of the depth markers 38 are visualized in fluoroscopic images taken during catheterization, allowing the length of the shaft 33 inserted into the blood vessel to be determined. FIG. 2B shows a method for measuring each of the connecting portions C1 to C4 of the balloon catheter using the transmission-type spectrometer 20. FIG. 2B illustrates the state in which the first connecting portion C1 between the first shaft tube 33a and the balloon 34 and the second connecting portion C2 between the balloon 34 and the second shaft tube 33b are measured. A predetermined number of balloon catheters (three in FIG. 2B ) are arranged and held in a holding member that holds the distal end of the first shaft tube 33a, and the measurement is performed using the spectrometer 20. The position at which the holding member holds the balloon catheter can be appropriately changed depending on the connection portions C1 to C4 to be measured. The measurement process using the spectroscopic device 20 may be performed during the manufacturing process of the balloon catheter, for example, after each of the shaft tubes 33a to 33e is connected to the balloon 34.

[0031] FIG. 3 is an explanatory diagram showing an example of the record layout of the analysis results DB 12a. The analysis results DB 12a is a database that stores the results of analysis (measuring connection strength) of the connecting portions C1 to C4 of the balloon catheter using the spectroscopic device 20. The analysis results DB 12a shown in FIG. 3 includes a catheter ID column, a connecting portion ID column, a spectroscopic data column, a component analysis result column, a connecting strength column, and other columns, and stores information about each connecting portion C1 to C4 of the balloon catheter in association with the catheter ID. The catheter ID column stores identification information (catheter ID) uniquely assigned to the manufactured balloon catheter. The connecting portion ID column stores identification information (connecting portion ID) uniquely assigned to the connecting portions C1 to C4 of the balloon catheter. The spectroscopic data column stores spectroscopic data acquired from the connecting portions C1 to C4 of the balloon catheter through measurement processing by the spectroscopic device 20. The component analysis result column stores information about the results of component analysis performed on the spectroscopic data. The component analysis performed on the spectroscopic data will be described later. The spectroscopic data and component analysis results may be stored in a predetermined area of ​​the storage unit 12 rather than in the analysis result DB 12a. In this case, the spectroscopic data column and component analysis result column store information for reading each data (e.g., a folder name and file name indicating the data storage location). The connection strength column stores information regarding the connection strength of the connection portions C1 to C4 of the balloon catheter. The connection strength information may include the connection strength of the connection portions C1 to C4, the results (pass / fail) of a connection strength test conducted on the connection portions C1 to C4, and the yield of the balloon catheter manufacturing process. The yield indicates the percentage of manufactured balloon catheters that passed the connection strength test. The contents stored in the analysis result DB 12a are not limited to the example shown in FIG. 3.

[0032] 4A and 4B are explanatory diagrams showing an example configuration of the learning model M. The learning model M shown in FIG. 4A is trained to receive input information indicating the results of component analysis of spectroscopic data acquired from a balloon catheter connection (any of connection C1 to C4) using the spectroscopic device 20, perform a calculation to estimate the connection strength of the connection based on the input data, and output the calculation result. The learning model M can be configured using algorithms such as a support vector machine (SVM), a decision tree, a random forest, a convolutional neural network (CNN), or a transformer, or may be configured by combining multiple algorithms. The information indicating the component analysis results, which is input data, includes the types and contributions of a predetermined number (approximately 2 to 7) of components in descending order of contribution (content ratio). The connection strength, which is output data, may be any numerical value or may be within a range of values ​​that can be assumed as connection strength.

[0033] The learning model M is generated by machine learning using training data that associates information indicating the results of component analysis of training spectroscopic data with correct connection strengths. The correct connection strengths can be measurement results (connection strengths) obtained by performing connection strength measurements on the connection parts of actually manufactured balloon catheters. The learning model M learns to output the correct connection strengths when information indicating the results of component analysis of the spectroscopic data included in the training data is input. Specifically, the learning model M performs calculations based on the input information, estimates the connection strengths of the connection parts from which the spectroscopic data was obtained, and obtains the estimated results. The learning model M then compares the estimated connection strengths with the correct connection strengths and optimizes parameters such as the weights (coupling coefficients) between neurons so that the estimated connection strengths approximate the correct connection strengths. While the parameter optimization method is not particularly limited, examples include steepest descent and backpropagation. As a result, a learning model M that outputs the connection strengths of the connection parts of a balloon catheter when information indicating the results of component analysis of the spectroscopic data is input is obtained. The information processing device 10 generates a learning model M for each of the connection parts C1 to C4.

[0034] The connection strength output by the learning model M shown in Figure 4A is a numerical value expressed as a continuous value, and the learning model M outputs one of the continuous values ​​as a so-called regression problem. However, it may also be configured to consider it as a classification problem and determine one of selectable numerical values ​​or numerical ranges (numerical values ​​or numerical ranges that can be taken as connection strength) for the connection strength. In this case, the learning model M determines the optimal numerical value or numerical range for the connection strength from multiple preset numerical values ​​or numerical ranges and outputs the determination result.

[0035] FIG. 4B shows a modified example of the learning model M. The learning model M1 shown in FIG. 4B receives input data similar to that of the learning model M shown in FIG. 4A, performs a calculation to estimate the results of a connection strength test for the link based on the input data, and outputs the calculation results. The learning model M1 has multiple output nodes, each of which is associated with a label indicating whether the connection strength test result for the link is a pass or a fail, and outputs the probability (certainty) that the connection strength test result should be determined to be a pass or a fail. The output value of each output node is, for example, a value between 0 and 1, and the sum of the probabilities output from each output node is 1.0 (100%). With the above-described configuration, when information indicating the component analysis results of the spectroscopic data is input, the learning model M1 outputs information regarding whether the connection strength test result for the balloon catheter link is a pass or a fail (certainty regarding pass or fail). The information processing device 10 specifies, as information on the connection strength to be predicted, the label (pass or fail) associated with the output node that outputs the largest output value (certainty) among the output values ​​from each output node in the learning model M1. Note that the learning model M1 may be configured to have a single output node that outputs the label (pass or fail) with the higher certainty, instead of having two output nodes that output certainty for pass or fail.

[0036] The learning model M1 is generated by machine learning using training data that associates information indicating the component analysis results of the training spectroscopic data with correct connection strength test results. The correct connection strength test results can be information indicating the test results (pass or fail) obtained by conducting a connection strength test on the connection portion of an actually manufactured balloon catheter. When information indicating the component analysis results of the spectroscopic data included in the training data is input, the learning model M1 learns so that the output value from the output node corresponding to the correct label (correct test result) approaches 1.0 and the output value from the other output node approaches 0.0. In the learning process, the learning model M1 performs calculations based on the information indicating the component analysis results of the input spectroscopic data to calculate output values ​​from each output node. The learning model M1 then compares the calculated output values ​​of each output node with values ​​corresponding to the correct label (1 for the output node corresponding to the correct label and 0 for the other output nodes) and optimizes parameters used in the calculation process so that the calculated output values ​​approximate the correct label. Here, parameters such as weights between neurons in the learning model M1 are optimized using backpropagation, steepest descent, or the like. This results in a learning model M1 that, when information indicating the results of component analysis of spectroscopic data is input, estimates information indicating the connection strength test results (pass or fail) for the connection part of the balloon catheter and outputs the estimated result.

[0037] The learning models M and M1 may be trained on another learning device. The trained learning models M and M1 generated by training on another learning device are downloaded from the learning device to the information processing device 10 via a network or via a portable storage medium 10a and stored in the storage unit 12. The learning models M and M1 are not limited to the configurations shown in Figures 4A and 4B and may be configured to estimate the yield in the manufacturing process of a balloon catheter based on information indicating the component analysis results of the spectroscopic data, for example.

[0038] The information processing device 10 prepares the above-described learning models M and M1 in advance for each of the connecting portions C1 to C4 to be measured, and when conducting a connection strength test for the connecting portions C1 to C4 of the balloon catheter, the information processing device 10 uses the learning models M and M1 to estimate the connection strength of the connecting portions C1 to C4 or the connection strength test results. The following describes an example of a configuration in which the information processing device 10 estimates the connection strength of the connecting portions C1 to C4 using the learning model M.

[0039] Here, we will explain the component analysis process (spectral analysis process) performed on spectroscopic data acquired using the spectroscopic device 20. FIGS. 5A to 6C are explanatory diagrams of the component analysis process. The left side of FIGS. 5A and 5B shows spectroscopic data, which are graphs with wavelength on the horizontal axis and absorbance on the vertical axis. The spectroscopic data acquired by the spectroscopic device 20 has overlapping absorbance (spectral) peaks for each component being measured, resulting in the peaks of each component being obscured, as shown on the left side of FIGS. 5A and 5B. Furthermore, as shown on the left side of FIG. 5B, the spectroscopic data exhibits a fluctuating baseline as the absorbance of each component changes. The baseline is expressed as a linear function (y = ax + b), and the example on the left side of FIG. 5B shows two baselines, one solid and one dashed. Furthermore, because spectroscopic data cannot capture the characteristics of each component based solely on wavelength changes at a single wavelength, it is necessary to analyze the change in absorbance (spectral change) across the entire wavelength range. Therefore, in this embodiment, preprocessing is performed on the spectroscopic data to expose hidden peaks and correct the baselines to eliminate variations between the baselines, and then multivariate analysis processing is performed to analyze the components contained in the spectroscopic data.

[0040] Preprocessing includes smoothing, normalization, and derivative processing. Smoothing reduces noise contained in the spectroscopic data, while normalization reduces the effect of the characteristics of the object being measured on the spectroscopic data. Second-order differentiation, as shown by the dashed circle on the right side of FIG. 5A, makes it possible to reveal absorbance peaks hidden in the spectroscopic data. Furthermore, as shown in FIG. 5B, the baseline represented by a linear function becomes y = 0, eliminating fluctuations between baselines.

[0041] Examples of the smoothing process that can be used include a moving average calculation process, a Gaussian filter process, a median filter process, and a Savitzky-Golay filter process. Examples of the normalization process that can be used include a standard normal variate (SNV) process and a multiplicative scatter correction (MSC) process. Examples of the differentiation process that can be used include a gap-segment process and a Savitzky-Golay filter process. Note that SNV processing is known to be able to appropriately suppress the influence of the film thickness of the measurement target on the spectroscopic data. In this embodiment, since the measurement target is the shaft 33, it is preferable to use SNV processing as the normalization process. Furthermore, since Savitzky-Golay filtering can simultaneously perform two filter processes, smoothing and differentiation, it is preferable to use Savitzky-Golay filtering as the smoothing process and differentiation process.

[0042] Examples of multivariate analysis include principal component analysis, cluster analysis, partial least squares discriminant analysis (PLS-DA), soft independent modeling of class analogy (SIMCA) classification, SVM, multiple regression analysis, principal component regression analysis, and PLS regression analysis. Figures 6A and 6B show the results of component analysis performed on the preprocessed spectroscopic data. Figure 6A shows a score plot representing the relationship between the score of the first principal component (PC1) and the score of the second principal component (PC2) obtained by the component analysis, with the score of the first principal component on the horizontal axis and the score of the second principal component on the vertical axis. In the score plot of Figure 6A, squares indicate unconnected connections (unfused connections), and circles indicate normal connections (fused connections). As can be seen from Figure 6A, the relationship between the first principal component score and the second principal component score can be used to determine whether the connections being measured are normal (fused or not). FIG. 6B shows a graph representing the change in loading value of the first principal component (PC1), with wavelength on the horizontal axis and loading value of the first principal component on the vertical axis. As can be seen from FIG. 6B, the peak of the first principal component (the distribution of specific wavelength components) can be identified as indicated by the dashed circle. In particular, CH group peaks are detected near 750 nm, 900 nm, 1200 nm, 1400 nm, and 1700 nm. Based on the differences between these peaks, the characteristics of the object can be determined, and the level of linkage strength (e.g., whether or not the object passes a linkage strength test) can be estimated (sorted). As described above, the characteristics of the object can be determined based on the scores and loading values ​​of each component contained in the spectroscopic data. In this embodiment, the linkage strength of the linkage portion of the object can be estimated from the results of component analysis of the spectroscopic data. FIG. 6C shows the absorbance of the peak wavelength extracted from the spectroscopic data when the peak wavelength of the first principal component shown in FIG. 6B is identified based on the spectroscopic data acquired by a hyperspectral camera, and image data is generated based on the extracted absorbance. Figure 6C shows a connection C2 that passed the connection strength test. If the peak wavelength of a principal component (e.g., the first principal component) can be identified, an image showing the characteristics of the object to be measured (the level of connection strength in Figure 6C) can be displayed based on the absorbance at the identified wavelength.

[0043] Next, the generation process of the learning model M will be described. FIG. 7 is a flowchart showing an example of the generation process procedure of the learning model M. The following process is performed by the control unit 11 of the information processing device 10 according to the program P stored in the storage unit 12, but may also be performed by another learning device. In the process shown in FIG. 7, the control unit 11 of the information processing device 10 first generates training data and then uses the generated training data to train the learning model M. Note that the information processing device 10 previously collects and stores in the storage unit 12 spectroscopic data to be used in the training data for each of the connection portions C1 to C4 and the connection strengths of the connection portions of the measurement targets from which the spectroscopic data was obtained. The information processing device 10 performs the following process for each of the connection portions C1 to C4 to generate a learning model M corresponding to each of the connection portions C1 to C4.

[0044] The control unit 11 reads one set of spectral data and link strengths for training data stored in the memory unit 12 (S11). The control unit 11 performs the above-described preprocessing on the acquired spectral data (S12), and then performs multivariate analysis (S13). For example, the control unit 11 performs SNV processing (normalization processing) as preprocessing, followed by Savitzky-Golay filtering (smoothing processing and differentiation processing), and then performs principal component analysis as multivariate analysis. This allows the control unit 11 to obtain component analysis results for the spectral data (distribution of each wavelength component contained in the light from the linking portion). The control unit 11 generates training data by assigning the link strengths obtained in step S11 to the acquired component analysis results as correct link strengths, and stores the training data in the memory unit 12 (S14). The control unit 11 stores the generated training data, for example, in a training DB (not shown) provided in the memory unit 12.

[0045] The control unit 11 determines whether or not there is unprocessed data that has not been used to generate the training data among the sets of spectral data and link strengths for the training data (S15). If it determines that there is unprocessed data (S15: YES), the control unit 11 returns to step S11. The control unit 11 then executes the processes of steps S11 to S14 for the sets of unprocessed spectral data and link strengths. The control unit 11 repeats the processes of steps S11 to S15 until it determines that there is no unprocessed data. As a result, training data to be used for learning the learning model M is generated and accumulated based on the spectral data and link strengths prepared for the training data.

[0046] If it is determined that there is no unprocessed data (S15: NO), the control unit 11 performs training on the learning model M using the accumulated training data. Specifically, the control unit 11 reads one piece of training data stored in the training DB (S16) and extracts the component analysis result for one pixel included in the read training data (S17). The control unit 11 then performs a training process on the learning model M using the component analysis result for one pixel and the connection strength included in the training data (S18). Here, the control unit 11 inputs information indicating the component analysis result for one pixel into the learning model M and obtains an output value from the learning model M. The control unit 11 then compares the output value from the learning model M with the correct connection strength included in the training data and optimizes parameters such as the weights between neurons in the learning model M using, for example, backpropagation so that the two values ​​approximate each other.

[0047] The control unit 11 determines whether there are any component analysis results for unprocessed pixels among the component analysis results included in the training data read out in step S16 (S19). If it determines that there are any (S19: YES), the control unit 11 returns to step S17. The control unit 11 then executes the processes of steps S17 to S18 for the component analysis results for the unprocessed pixels. That is, the control unit 11 extracts the component analysis result for one unprocessed pixel and performs a learning process for the learning model M using the component analysis result for the extracted one pixel and the connection strength included in the training data. The control unit 11 repeats the processes of steps S17 to S19 until it determines that there are no unprocessed pixels. This allows the component analysis results for each pixel included in the component analysis result of one piece of spectral data prepared as training data to be used for learning the learning model M.

[0048] If the control unit 11 determines that there are no component analysis results for unprocessed pixels (S19: NO), it determines whether there is any unprocessed training data that has not been subjected to learning processing among the training data stored in step S14 (S20). If it determines that there is unprocessed training data (S20: YES), the control unit 11 returns to the processing of step S16 and executes the processing of steps S16 to S19 for the unprocessed training data. If it determines that there is no unprocessed training data (S20: NO), the control unit 11 ends the series of processing.

[0049] The above-described process generates a learning model M that is trained to output the connection strength of the measurement object (connection portion) of the spectroscopic data when the results of component analysis processing on the spectroscopic data are input. Thus, the information processing device 10 can use the learning model M to estimate the connection strength of the connection portion from the component analysis results on the spectroscopic data acquired from the connection portion of the measurement object using the spectroscopic device 20. In the above-described process, the training data generation processing in steps S11 to S15 and the learning model M generation processing in steps S16 to S20 may be performed by separate devices. The learning model M can be further optimized by repeatedly performing the learning processing using the training data described above. Furthermore, the already-trained learning model M may be fine-tuned (transfer learning) using training data for each balloon catheter manufacturing factory or training data for each balloon catheter (shaft) size. In this case, a learning model M tailored to each manufacturing factory or each size is generated, enabling appropriate connection strength estimation for each manufacturing factory or each size.

[0050] The following describes a process for estimating the connection strength of a balloon catheter's connection portion using the learning model M configured as described above. FIG. 8 is a flowchart showing an example of the connection strength estimation process. The control unit 11 of the information processing device 10 determines whether there is a balloon catheter connection portion for which spectroscopic data is stored in the analysis result DB 12a but component analysis results are not stored. If it is determined that there is a connection portion, the control unit 11 executes the following process for that connection portion. Specifically, the following process is executed after the connection work (fusion work) of each of the connection portions C1 to C4 is completed during the manufacturing process of the balloon catheter. Note that the following process may be executed for each of the connection portions C1 to C4 after the connection work for all of the connection portions C1 to C4 is completed, and the connection strengths may be measured collectively.

[0051] The control unit 11 acquires spectroscopic data of the connection portion to be processed from the analysis result DB 12a (S31). The control unit 11 performs preprocessing on the acquired spectroscopic data (S32), and then performs multivariate analysis processing (S33). Steps S32 to S33 are the same as steps S12 to S13 in Fig. 7, and the control unit 11 thereby acquires component analysis results for the spectroscopic data (distribution of each wavelength component contained in the light from the connection portion) and stores the acquired component analysis results in the analysis result DB 12a in association with the catheter ID and connection portion ID of the connection portion.

[0052] The control unit 11 reads from the storage unit 12 the learning model M corresponding to the connection portion (one of the connection portions C1 to C4) to be processed (S34). The control unit 11 then extracts the component analysis result for one pixel from the component analysis result for the spectroscopic data (S35). The control unit 11 then estimates the connection strength of the measurement object corresponding to the extracted pixel (S36). Here, the control unit 11 inputs information indicating the component analysis result for one pixel into the learning model M corresponding to the connection portion (one of the connection portions C1 to C4) of the measurement object, and obtains the connection strength as an output value from the learning model M. Note that the process of estimating the connection strength from the component analysis result is not limited to processing using the learning model M, but may also be rule-based processing. For example, the connection strength corresponding to the component analysis result can be estimated by previously associating the component analysis result for the spectroscopic data with the connection strength. Alternatively, the distribution of a specific wavelength component (e.g., peak wavelength) with the connection strength may be previously associated, and the connection strength corresponding to the distribution of a specific wavelength component in the component analysis result for the measurement object may be estimated. The control unit 11 stores the estimated connection strength in the storage unit 12 in association with, for example, the identification information of the corresponding pixel.

[0053] The control unit 11 determines whether there are any component analysis results for unprocessed pixels among the component analysis results acquired in step S33 (S37), and if it determines that there are (S37: YES), the process returns to step S35. The control unit 11 then executes the processes of steps S35 to S36 for the component analysis results for the unprocessed pixels. The control unit 11 repeats the processes of steps S35 to S37 until it determines that there are no unprocessed pixels. As a result, the connection strength based on the component analysis results for each pixel in the spectral data is estimated and stored in the storage unit 12.

[0054] If it is determined that there are no component analysis results for unprocessed pixels (S37: NO), the control unit 11 identifies the connection strength of the connection part based on the connection strength estimated for each pixel in step S36 (S38). For example, the control unit 11 calculates the proportion of the connection strengths estimated for each pixel and identifies the connection strength with the largest proportion as the connection strength of the connection part. Alternatively, the control unit 11 may identify the smallest connection strength among the connection strengths estimated for each pixel as the connection strength of the connection part. Alternatively, the control unit 11 may identify the average value of the connection strengths estimated for each pixel as the connection strength of the connection part. The control unit 11 stores the identified connection strength of the connection part in the analysis result DB 12a in association with the catheter ID and connection part ID of the connection part (S39), and ends the process.

[0055] Through the above-described process, the information processing device 10 of this embodiment can estimate the coupling strength of the coupling portion based on the spectroscopic data acquired from the coupling portion of the balloon catheter being measured using the spectroscopic device 20 and store the estimated coupling strength in the analysis result DB 12a. Note that, after the process of step S38, the control unit 11 may output an alert to notify a user (e.g., a balloon catheter manufacturer) if, for example, the identified coupling strength is less than a predetermined value. Furthermore, upon receiving a request to view the coupling strengths of each coupling portion stored in the analysis result DB 12a, the control unit 11 may display a screen on the display unit 15 showing the coupling strength of each coupling portion in association with the catheter ID, for example. In this case, the control unit 11 may highlight coupling strengths that are less than the predetermined value to provide a warning. With this configuration, the user can determine whether the estimated coupling strengths of each coupling portion of the balloon catheter are appropriate.

[0056] In this embodiment, the process of measuring the connection strength of the connecting portions of the balloon catheter using the spectroscopic device 20 and the information processing device 10 may be performed after each connecting portion is connected during the manufacturing process of the balloon catheter. For example, each time the connecting operation (fusion operation) of any of the connecting portions C1 to C4 is completed, the light-emitting unit 21 of the spectroscopic device 20 irradiates the connecting portion C1 to C4 with near-infrared light, and the light-receiving unit 22 receives the light from the connecting portion C1 to C4 to obtain spectroscopic data. The control unit 11 of the information processing device 10 may then identify the connection strength of the connecting portions C1 to C4 based on the spectroscopic data acquired by the spectroscopic device 20, determine whether the identified connection strength is equal to or greater than a reference value (predetermined value), and if the connection strength is equal to or greater than the reference value, determine that the connecting portion C1 to C4 has passed the connection strength test. This manufacturing process makes it possible to manufacture balloon catheters whose connection strengths at each of the connecting portions C1 to C4 pass the connection strength test. Furthermore, the process of measuring the connection strength of the connecting portions of the balloon catheter may be performed after the connection (fusion) of all connecting portions C1-C4 is completed. In this case, after the balloon catheter is manufactured by connecting all connecting portions C1-C4, the light-emitting unit 21 of the spectroscopic device 20 irradiates each connecting portion C1-C4 of the balloon catheter with near-infrared light, and the light-receiving unit 22 receives the light from the connecting portions C1-C4 to obtain spectroscopic data. The control unit 11 of the information processing device 10 then determines the connection strength of each connecting portion C1-C4 based on the spectroscopic data obtained by the spectroscopic device 20 from each connecting portion C1-C4 and determines whether the determined connection strength is equal to or greater than a reference value (predetermined value). Even in this manufacturing process, the balloon catheter may be usable if the connection strength of each connecting portion C1-C4 is equal to or greater than the reference value, thereby enabling the manufacture of balloon catheters that pass the connection strength test.

[0057] In this embodiment, the spectroscopic device 20 may be either a near-infrared spectroscopic device or a hyperspectral camera, and may be either a reflective or transmissive type. Furthermore, it is preferable to use preprocessing and multivariate analysis performed on the spectroscopic data that have high accuracy in estimating the connection strength of the connection parts based on the results of component analysis performed on the spectroscopic data. For example, a generated learning model M with high estimation accuracy is used.

[0058] In this embodiment, the process of estimating the connection strength of each connection part from the component analysis result is not limited to being performed locally by the information processing device 10. For example, a server may be provided to execute the process of estimating the connection strength of each connection part from the component analysis result. In this case, the information processing device 10 may be configured to transmit information indicating the component analysis result for one pixel extracted in step S35 of FIG. 8 to the server, and receive the estimated connection strength from the server.

[0059] The embodiments disclosed herein are illustrative in all respects and should not be considered limiting. The scope of the present invention is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims.

[0060] REFERENCE SIGNS LIST 10 Information processing device 11 Control unit 12 Storage unit 13 Communication unit 20 Spectroscopic device 21 Light emitting unit 22 Light receiving unit 23 Signal processing unit 24 Communication unit M Learning model

Claims

1. An information processing method in which a computer executes a process of acquiring spectroscopic data of light generated from a connecting portion of a shaft and a balloon that constitute a balloon catheter, or a connecting portion of a continuous shaft, generated by irradiating the connecting portion with near-infrared light, and identifying information regarding the connection strength of the connecting portion based on the acquired spectroscopic data.

2. The information processing method according to claim 1, wherein the computer executes a process of performing a multivariate analysis on the spectroscopic data to obtain a distribution of each wavelength component contained in the light from the connecting portion, and identifying information relating to the connection strength of the connecting portion based on the distribution of each wavelength component.

3. The information processing method according to claim 2, wherein the computer executes a process of identifying information relating to the connection strength of the connection portion based on a distribution of a specific wavelength component in the distribution of each of the wavelength components.

4. The information processing method according to claim 2, wherein the computer executes a process of inputting the distribution of each wavelength component obtained by performing multivariate analysis on the spectroscopic data into a learning model that has been trained to output information regarding the connection strength of the connection part when the distribution of each wavelength component contained in the light from the connection part is input, and acquiring information regarding the connection strength of the connection part.

5. The information processing method according to claim 4, wherein the computer executes the process of: storing a plurality of the learning models corresponding to a plurality of the connection parts in a memory unit; identifying the learning model corresponding to the position of the connection part; and using the identified learning model to obtain information regarding the connection strength of the connection part.

6. The information processing method according to any one of claims 1 to 5, wherein the computer executes a process of performing pre-processing including normalization, smoothing, and differentiation on the spectroscopic data.

7. The information processing method according to any one of claims 1 to 5, wherein the computer executes a process of acquiring the spectroscopic data using a near-infrared spectrometer or a hyperspectral camera.

8. A program that causes a computer to execute a process of acquiring spectroscopic data of light generated from a connecting portion of a shaft and a balloon that constitute a balloon catheter, or a connecting portion of a continuous shaft, generated by irradiating the connecting portion with near-infrared light, and identifying information regarding the connection strength of the connecting portion based on the acquired spectroscopic data.

9. An information processing device having a control unit, wherein the control unit acquires spectroscopic data of light generated from a connecting portion of a shaft and a balloon that constitute a balloon catheter, or a connecting portion of a continuous shaft, generated by irradiating the connecting portion with near-infrared light, and identifies information regarding the connection strength of the connecting portion based on the acquired spectroscopic data.

10. A model generation method in which a computer executes a process of acquiring training data that associates the distribution of each wavelength component contained in the light from a connecting portion generated by irradiating near-infrared light onto the connecting portion of a shaft and a balloon that constitutes a balloon catheter, or the connecting portion of a continuous shaft, with information regarding the connection strength of the connecting portion, and using the acquired training data, generating a learning model that outputs information regarding the connection strength of the connecting portion when the distribution of each wavelength component contained in the light from the connecting portion is input.

11. A method for manufacturing a balloon catheter, comprising the steps of: connecting a shaft and a balloon for a balloon catheter, and / or a continuous shaft; irradiating the connecting portion of the shaft and the balloon, and / or the connecting portion of the continuous shaft, with near-infrared light by a light-emitting portion, acquiring spectroscopic data of the light from the connecting portion; determining information relating to the connection strength of the connecting portion based on the acquired spectroscopic data; and determining whether the determined information relating to the connection strength is equal to or greater than a reference value.

Images