Coating layer inspection device and coating layer inspection method

Abstract

To provide a coat layer inspection device and a coat layer inspection method with which characteristics of various defects in a coat layer of an analyte can be compositely detected and the defects can be determined easily with high accuracy.SOLUTION: A coat layer inspection device 10 detects defects in a coat layer 102 formed on a surface of an analyte, and has an imaging unit 30 that picks up an image of the coat layer 102 to generate image data, and a processing unit 42 that performs arithmetic processing. The processing unit 42 has an image acquisition unit 50 that can read a plurality of pieces of defective image data divided into a plurality of different types of defects and determination object image data, a determination device creation unit 52 that compositely extracts feature quantities from the plurality of pieces of defective image data to create a determination device, an abnormality degree detection unit 53 that detects a degree of abnormality numeralized from the determination device, and a defect determination unit 55 that compares the degree of abnormality detected from the determination object image data by using the determination device and the abnormality degree detection unit 53 with a threshold value to determine whether the coat layer 102 has defects.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a coating layer inspection apparatus and a coating layer inspection method for inspecting defects in a coating layer formed on the surface of a test object. [Background technology]

[0002] In recent years, balloon catheters have been used to treat lesions (stenosis) that have developed in biological lumens. A balloon catheter usually includes a long shaft and a balloon that is provided on the distal side of the shaft and is expandable in the radial direction. The deflated balloon is allowed to travel through a narrow biological lumen to a target location in the body, and then expanded to widen the lesion.

[0003] However, forcibly expanding the lesion may cause excessive proliferation of smooth muscle cells, resulting in new stenosis (restenosis) at the lesion. For this reason, drug-eluting balloons (DEBs), in which the surface of the balloon is coated with a drug to suppress stenosis, have recently been used. When the drug-eluting balloon expands, it releases the drug coated on its surface into the lesion, thereby suppressing restenosis.

[0004] From the viewpoint of drug release and tissue migration, it is preferable that the drug coating layer is provided uniformly on the surface of the balloon in a desired amount. Therefore, it is desirable to be able to determine the state of the drug coating layer that has been formed.

[0005] For example, Patent Document 1 describes an apparatus that automatically reviews the threshold for determining whether or not a test subject has passed or failed each time an image of the test subject is inspected, and verifies, through the review, whether or not there have been any erroneous judgments made on the pass / fail judgments of the image so far. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Patent No. 5865707 specification Summary of the Invention [Problem to be solved by the invention]

[0007] There are various types of defects in the drug coating layer. For example, there are defects in which the crystal form of the drug varies depending on the site, the drug is not sufficiently loaded on the balloon, large crystal agglomerates are present, crystal agglomerates are present over a wide area, defects in the coating layer are present, and foreign matter that should not be present is present. Among these, to determine defects such as the type in which the drug is not sufficiently loaded on the balloon or the type in which crystal agglomerates are present over a wide area, it is necessary to detect the degree of abnormality of the drug or crystal agglomerates rather than simply determining the presence or absence of the drug or crystal agglomerates.

[0008] However, with the device and method described in Patent Document 1, it is difficult to effectively detect the degree of abnormality.

[0009] The present invention has been made to solve the above-mentioned problems, and aims to provide a coating layer inspection device and a coating layer inspection method that can comprehensively detect the characteristics of various defects in the coating layer of a test object and easily determine the defects with high accuracy. [Means for solving the problem]

[0010] A coating layer inspection device according to the present invention for achieving the above object is a coating layer inspection device that detects defects in a coating layer containing a plurality of drug crystals formed on the surface of a subject, which is a balloon that can be inserted into a biological lumen, and includes an imaging unit that images the coating layer to create image data, an irradiation unit that irradiates the subject with light from a direction intersecting with the imaging direction of the imaging unit to form a shadow due to unevenness on the surface of the subject, and a processing unit that performs arithmetic processing, the processing unit including an image acquisition unit that is capable of reading a plurality of defect image data that are classified into a plurality of different defect types and judgment target image data, a judger creation unit that extracts feature amounts from the plurality of defect image data and creates a judger, an abnormality detection unit that detects a degree of abnormality that has been quantified by the judger, and a defect judgment unit that judges whether or not there is a defect in the coating layer by comparing the degree of abnormality detected from the judgment target image data using the judger and the abnormality detection unit with a threshold value. The plurality of different defect types are at least two of the following types: Type 1, in which the crystal form of the drug varies depending on the part of the balloon; Type 2, in which the drug is not sufficiently loaded on the balloon; Type 3, in which large crystal aggregates are present; Type 4, in which crystal aggregates are present over a wide area; Type 5, in which defects are present in the coating layer; and Type 6, in which foreign matter is present. do.

[0011] A coating layer inspection method according to the present invention for achieving the above object is a coating layer inspection method for detecting defects in a coating layer containing a plurality of drug crystals formed on the surface of a specimen that is a balloon that can be inserted into a biological lumen, the method including the steps of: reading in a plurality of defect image data that can be divided into a plurality of different defect types as teacher data; extracting feature amounts from the plurality of defect image data using a neural network to create a classifier; quantifying and detecting the degree of abnormality from the created classifier using a support vector machine; setting a threshold value for the degree of abnormality that can comprehensively detect the features of the plurality of defects; the step of irradiating the object with light from a light source to form a shadow on the surface of the object due to the unevenness of the coating layer, and creating judgment target image data by the imaging unit; the step of reading the judgment target image data; the step of extracting features from the judgment target image data and detecting a degree of abnormality of the judgment target image data by the created judger and support vector machine; the step of comparing the degree of abnormality of the judgment target image data with the threshold value to judge whether the degree of abnormality is equal to or greater than the threshold value or not, and if the judgment is positive, judging that there is a defect in the coating layer of the judgment target image data, and if the judgment is negative, judging that there is no defect in the coating layer of the judgment target image data. The plurality of different defect types are at least two of the following types: Type 1, in which the crystal form of the drug varies depending on the part of the balloon; Type 2, in which the drug is not sufficiently loaded on the balloon; Type 3, in which large crystal aggregates are present; Type 4, in which crystal aggregates are present over a wide area; Type 5, in which defects are present in the coating layer; and Type 6, in which foreign matter is present. do. Effect of the Invention

[0012] The coating layer inspection device configured as described above compares the degree of abnormality detected by the judger and the abnormality detection unit with a threshold value to determine the degree of the defect, thereby making it possible to detect the characteristics of various defects in the coating layer of the test subject in a comprehensive manner and to easily determine the defects with high accuracy.

[0013] The coating layer inspection method configured as described above compares the degree of abnormality detected by the classifier and the support vector machine with a threshold value to determine the degree of the defect, thereby making it possible to detect the characteristics of various defects in the coating layer of the subject in a comprehensive manner and to easily determine the defects with high accuracy.

[0014] The test object may be a balloon that can be inserted into a biological lumen, and the coating layer may contain a drug. In this way, the coating layer inspection device and the coating layer inspection method can easily determine various types of defects in the coating layer that contains a drug on the surface of the balloon with high accuracy. 。 [Brief description of the drawings]

[0015] [Figure 1] 1 is a block diagram showing a coating layer inspection device according to an embodiment of the present invention; [Diagram 2] FIG. 2 is a front view showing a balloon as a test object. [Diagram 3] 10 is a flowchart showing a process flow in a processing unit. [Figure 4] Schematic diagrams showing the types of defects in the coating layer of a balloon: (A) is a type in which the crystal form of the drug varies depending on the location, (B) is a type in which the drug is not sufficiently loaded onto the balloon, (C) is a type in which large crystal clumps are present, (D) is a type in which crystal clumps are present over a wide area, (E) is a type in which there is a defect in the coating layer, and (F) is a type in which a foreign object is present. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Note that the dimensional ratios in the drawings may be exaggerated for convenience of explanation and may differ from the actual ratios. In addition, in this specification and the drawings, components having substantially the same functions are denoted by the same reference numerals, and duplicated explanations will be omitted.

[0017] A coating layer inspection device 10 according to an embodiment of the present invention, as shown in Figures 1 and 2, is used to determine whether or not there is a defect in a coating layer 102 disposed on the surface of a balloon 101 of a drug-eluting balloon catheter 100.

[0018] The surface of the balloon 101 is smooth and non-porous, but may be porous. The balloon 101 is formed of a material that has a certain degree of flexibility and a certain degree of hardness so that the balloon 101 expands when it reaches a blood vessel or tissue, and the drug is released from the coating layer 102 on the surface. Specifically, at least the surface of the balloon 101 on which the coating layer 102 is provided is preferably made of a resin. Examples of materials that can be used for at least the surface of the balloon 101 include polyolefins such as polyethylene, polypropylene, polybutene, ethylene-propylene copolymer, ethylene-vinyl acetate copolymer, ionomer, or a mixture of two or more of these, soft polyvinyl chloride resin, polyamide, polyamide elastomer, nylon elastomer, polyester, polyester elastomer, polyurethane, thermoplastic resin such as fluororesin, silicone rubber, latex rubber, etc.

[0019] The coating layer 102 is formed containing a drug. The coating layer 102 may contain an additive. The drug is not particularly limited, but is preferably a water-insoluble drug. For example, the coating layer 102 is formed of a plurality of minute drug crystals. The drug may be in the form of crystal, non-crystalline (amorphous), or a mixture thereof. The crystal or non-crystalline may be arranged in the coating layer 102 so as to have regularity, or may be arranged irregularly. The crystal form is not particularly limited, and may be, for example, a columnar form elongated in one direction, a plate-like form, an irregular block form, or the like.

[0020] The surface of the coating layer 102 has smooth regions and / or uneven regions. The coating layer 102 has various kinds of colors and color shading, which are generated by factors such as the type of material and the presence or absence of light refraction or reflection. The range in which these are present may be the entire coating layer 102 or a localized part. The coating layer 102 may have these characteristics continuously or intermittently.

[0021] Some examples of preferred water insoluble drugs include immunosuppressants, e.g., cyclosporines including cyclosporine, immunostimulants such as rapamycin, anti-cancer drugs such as paclitaxel, antiviral or antibacterial drugs, antineoplastic agents, analgesics and anti-inflammatory drugs, antibiotics, antiepileptic drugs, anxiolytic drugs, antiparalytic drugs, antagonists, neuron blocking drugs, anticholinergic and cholinergic drugs, antimuscarinic and muscarinic drugs, antiadrenergic drugs, antiarrhythmic drugs, antihypertensive drugs, hormonal drugs, and nutritional drugs.

[0022] The water-insoluble drug is preferably at least one selected from the group consisting of rapamycin, paclitaxel, docetaxel, and everolimus. In this specification, rapamycin, paclitaxel, docetaxel, and everolimus include their analogs and / or their derivatives as long as they have the same pharmacological effect. For example, paclitaxel and docetaxel are in an analogue relationship. Rapamycin and everolimus are in a derivative relationship. Among these, paclitaxel is more preferred.

[0023] Next, a description will be given of the coating layer inspection device 10. The coating layer inspection device 10 includes an illumination device 20 that illuminates a balloon 101, which is an object to be inspected, an imaging section 30 that captures an image of the illuminated object to create image data and transmit the data as an electrical signal, a processing device 40 that receives the image data from the imaging section 30, a display device 70 that displays the determination results by the processing device 40, and an input device 80 that can input information to the processing device 40.

[0024] The illumination device 20 illuminates the subject. It is desirable for the illumination device 20 to uniformly illuminate the determination target area A on the surface of the subject. The illumination of the illumination device 20 can be determined if the vicinity of the surface of the subject is 100 lx or more, but is preferably 1,000 to 200,000 lx, and more preferably 20,000 to 40,000 lx, since the surface of the subject can be uniformly irradiated. The type of light source is not important, and may be an incandescent bulb, a fluorescent lamp, a halogen lamp, an HID lamp, an LED lamp, an organic LE lamp, or an HID halogen lamp as long as the above-mentioned illumination can be ensured. The HID halogen lamp is a general term for high-pressure mercury lamps and the like, and has a higher brightness than an incandescent lamp, and is also used for lighting in automobiles and trains, for example.

[0025] The imaging unit 30 captures an image of the subject illuminated by the illumination device 20, creates a digital image, and transmits it to the processing device 40. The imaging unit 30 includes, for example, a lens for forming an image of the reflected light from the subject, and a CCD camera for converting the formed image of the subject into an electrical signal. Since the greater the number of pixels, the clearer the boundary between light and dark colors becomes, and the accuracy of judgment improves, the number of pixels of the camera is 200,000 or more, but is preferably 400,000 to 10 million pixels, and more preferably 1 million to 5 million pixels. The shutter speed is 1 / 15 or more, but is preferably 1 / 20 to 1 / 100,000, and more preferably 1 / 30 to 1 / 100.

[0026] The display device 70 is a device that receives and displays the signals output from the processing device 40, and has, for example, a monitor. The display device 70 displays the images captured by the imaging unit 30, the images that have been subjected to image processing, and the determination result as to whether or not the object has a defect.

[0027] The input device 80 is a device capable of inputting information to the processing device 40, and is, for example, a keyboard, a mouse, a connector capable of connecting a storage medium, another computer, etc. The input device 80 is used by a user to perform various operations.

[0028] The processing device 40 includes an interface 41 for receiving an electric signal from the imaging unit 30 as image data, and a processing unit 42 for performing arithmetic processing. The processing device 40 is, for example, a computer.

[0029] The processing unit 42 includes, as its physical components, a memory circuit and an arithmetic circuit. The memory circuit stores programs and various parameters. The arithmetic circuit is, for example, a CPU (Central Processing Unit), which reads the programs and various parameters from the memory circuit and performs preprocessing on the image, such as arithmetic processing, binarization, contrast adjustment, and noise control.

[0030] The processing unit 42 has, as its functional configuration, an image acquisition unit 50, an area specification unit 51, a judger creation unit 52, an abnormality detection unit 53, a threshold setting unit 54, a defect determination unit 55, and a result output unit 56. The image acquisition unit 50, the area specification unit 51, the judger creation unit 52, the abnormality detection unit 53, the threshold setting unit 54, the defect determination unit 55, and the result output unit 56 will be described in detail together with the description of the processing method described later.

[0031] Next, a coating layer inspection method using the coating layer inspection device 10 according to the embodiment will be described with reference to a flow chart of the processing unit 42 shown in FIG.

[0032] The imaging unit 30 captures an image of the subject illuminated by the illumination device 20, and transmits the resulting digital image as an electrical signal to the processing device 40. The electrical signal of the digital image is transmitted to an interface 41 of the processing device 40. The interface 41 captures the electrical signal from the imaging unit 30 as image data. The imaging unit 30 is used to capture a plurality of defect image data that serve as training data for indicating the coating layer 102 having defects. The training data is used for machine learning by AI (artificial intelligence). The imaging unit 30 is also used to capture determination target image data that is to be used to determine the presence or absence of defects.

[0033] Next, the image acquisition unit 50 of the processing unit 42 reads the plurality of pieces of defect image data from the imaging unit 30 via the interface 41 (step S10). The plurality of pieces of defect image data may be input from a storage medium connected to the input device 80, another computer, or the like. Therefore, the defect image data is not limited to that captured by the imaging unit 30.

[0034] The types of defects in the coating layer 102 include, for example, type 1 in which the crystal form of the drug differs depending on the region as shown in FIG. 4(A), type 2 in which the drug is not sufficiently loaded on the balloon as shown in FIG. 4(B), type 3 in which a large crystal mass 103 exists as shown in FIG. 4(C), type 4 in which the crystal mass 103 exists over a wide area as shown in FIG. 4(D), type 5 in which a defect 104 exists in the coating layer as shown in FIG. 4(E), and type 6 in which a foreign body 105 that should not be present exists as shown in FIG. 4(F). As shown in FIG. 4(B), the coating layer 102 may be defective and the coating layer 102 may be defective, resulting in an exposed portion 106 in which the surface of the balloon 101 itself is exposed. The above-mentioned types of defects may be combined in a complex manner. For example, when the exposed portion 105 occurs, the crystal mass 103 may be present or the crystal mass 103 may not be present.

[0035] The greater the number of defect image data sheets used to create a classifier, the higher the accuracy of the classifier. Generally, at least several hundred to a thousand pieces of defect image data are required to create a classifier. Although a classifier can be created even with only about 50 pieces of image data for each defect type, the accuracy of reading in the features can be improved by using images in which the features of each defect are clearly visible. The defect image data may be colored or colorless. Moreover, the image data to be judged may be colored or colorless.

[0036] In this embodiment, the data required for creating a classifier are image data related to the shading of the color of the drug (e.g., white) and image data related to the size of the crystal mass 103. From two perspectives, the intensity of the shading of the white color and its proportion, and the size of the crystal mass 103 and its proportion, the defect image data of the coating layer 102 corresponding to the defect is read as teacher data. The read-in defect image data preferably includes image data of the upper and lower limits of each defect. Since machine learning of the degree of a defect is difficult, at least about 20 pieces of image data are required for each defect.

[0037] The defects such as the defect 104 and the foreign matter 105 in the coating layer 102 are not necessarily required for creating a classifier. Machine learning is easy for the defect 104 and the foreign matter 105 because the presence or absence is judged rather than the degree. Therefore, for the defect 104 and the foreign matter 105, defect image data with prominent features is read as training data. Since machine learning is easy for the defect 104 and the foreign matter 105, at least about five pieces of defect image data are required for each defect.

[0038] Next, the processing unit 42 uses the region specifying unit 51 to specify the determination target region A in which the coating layer 102 is to be formed from the determination target image data (step S11). The determination target image data includes the determination target region A. When the position of the balloon 101 relative to the imaging unit 30 is substantially determined, the determination target region A can be fixedly set within the field of view of the imaging unit 30. Note that the method of specifying the determination target region A is not particularly limited. For example, the determination target region A may be separated from the background by grayscaling each pixel of the determination target image data by brightness and comparing it with an appropriate threshold value. Alternatively, the determination target region A may be manually specified by displaying the determination target image data on the display device 60.

[0039] Next, the processing unit 42 extracts features from the defect image data and creates a classifier by the classifier creation unit 52 using a neural network (step S12). The features for creating the classifier are not particularly limited, but are a 4096-dimensional feature vector. The neural network used is not particularly limited, and may be, for example, Alexnet, VGG-16, or VGG-19.

[0040] Next, the processing unit 42 detects the degree of abnormality by quantifying it from the created classifier using the abnormality detection unit 53 (step S13). In order to quantify the degree of abnormality from the classifier, a support vector machine (SVM) is used. When there are the above-mentioned first to sixth type defects, the classifier is created so that the degree of abnormality detected by the support vector machine from the read feature amount is detected as high. In the second type defect in which the drug is not sufficiently loaded on the balloon 101 and the fourth type defect in which the crystal mass 103 is present over a wide area, not only the presence or absence of an abnormality but also the degree of the abnormality is related to the presence or absence of a defect. That is, in the second type, it is necessary to specify how little drug is loaded on the balloon 101 to determine that it is a defect, and in the fourth type, how widely the crystal mass 103 is present to determine that it is a defect. In order to be able to determine such defects using a threshold value, it is preferable to create a classifier so that the degree of abnormality is detected as high as the degree of abnormality (the degree to which the drug is not loaded or the degree of spread of the crystal mass 103) increases.

[0041] When creating a classifier, the training data is decomposed into, for example, 4096-dimensional feature vectors, and the distribution of vectors characteristic of defects is learned. The closer a vector is to the center of the distribution, the more prominent its defect characteristics are, and the greater its influence on the degree of abnormality. This increases the degree of abnormality, which is a single numerical value calculated from the defect image data. This makes it possible to comprehensively detect the characteristics of various defects in the coating layer 102 of the test object, and to determine the degree of the defect.

[0042] Next, the processing unit 42 sets a threshold value capable of detecting the above-mentioned first to sixth types of defects by the threshold value setting unit 54 (step S14). The threshold value setting unit 54 receives a threshold value input by the user from the input device 80, and sets the value as the threshold value. In particular, if the classifier is designed so that the degree of abnormality is detected higher as the degree of abnormality of the above-mentioned second and fourth types of defects increases, the degree of abnormality required to determine a defect can be set with high accuracy by adjusting the threshold value. The threshold value may be automatically calculated and set by the threshold value setting unit 54 so as to detect the above-mentioned first to sixth types of defects. Completion of steps S10 to S14 completes the setting before determining whether or not the determination target image data has a defect. The threshold value is not particularly limited because it is set appropriately, but may be set to a value in the range of 1 to 100, further 20 to 80, further 30 to 60, or further 35 to 40, for example.

[0043] When determining whether or not there is a defect in the determination target image data, the processing unit 42 causes the image acquisition unit 50 to read the determination target image data from the imaging unit 30 via the interface 41 (step S15).

[0044] Next, the processing unit 42 extracts a feature amount from the determination target image data, and detects the degree of abnormality using the classifier created by the classifier creating unit 52 and the abnormality degree detecting unit 53 (step S16).

[0045] When the image is judged by the created judger, the judgement target image data is decomposed into, for example, 4096-dimensional feature vectors, vectors characteristic of defects are detected, and the degree of abnormality is detected. Therefore, the characteristics of various defects in the coating layer 102 of the subject can be detected in a composite manner. Composite detection of defects means detecting the characteristics of the type of defect (for example, the six defects in Fig. 4(A) to (F)) from the captured image, detecting feature vectors (for example, 4096 feature vectors), comparing them with the vector distribution of the defect image data when the judger was created, weighting all the vectors, and quantifying the degree of failure as the degree of abnormality and ranking the values. "Weighting vectors" means weighting vectors among the 4096 feature vectors that are characteristic of defects. A vector that is more characteristic of a defect is weighted so that the numerical value of the degree of abnormality of the image in which the vector is detected is higher. "Quantifying the degree of failure as the degree of abnormality" means quantifying not only the abnormal parts of one image, but the degree of failure of the entire image. "Ranking the numbers" means that each image is quantified according to its degree of abnormality and then ranked.

[0046] Next, the processing unit 42 causes the defect determination unit 55 to compare the detected degree of abnormality with the threshold value set by the threshold setting unit 54 (step S17). The defect determination unit 55 determines whether the degree of abnormality is equal to or greater than the threshold value. If the determination is positive, the defect determination unit 55 determines that any of the above-mentioned first to sixth types of defects exists, and determines that the coating layer 102 has a defect (step S18). Next, the processing unit 42 causes the result output unit 56 to display the result on the display device 70 (step S20), and ends the processing.

[0047] If the determination is negative, the defect determination unit 55 determines that the coating layer 102 does not have any of the above-mentioned first to sixth types of defects (step S19). Next, the processing unit 42 causes the result output unit 56 to display the result on the display device 70 (step S20), and ends the processing.

[0048] As described above, the coating layer inspection apparatus 10 of this embodiment is a coating layer inspection apparatus 10 that detects defects in the coating layer 102 formed on the surface of a test object, and has an imaging unit 30 that images the coating layer 102 and creates image data, and a processing unit 42 that performs arithmetic processing. The processing unit 42 has an image acquisition unit 50 that can read multiple defect image data and image data to be judged, a judger creation unit 52 that extracts composite features from the multiple defect image data and creates a judger, an abnormality detection unit 53 that detects a quantified degree of abnormality from the judger, and a defect judgment unit 55 that compares the degree of abnormality detected from the image data to be judged using the judger and anomaly detection unit 53 with a threshold value to judge whether or not there is a defect in the coating layer 102.

[0049] The coating layer inspection device 10 configured as described above compares the degree of abnormality detected by the judger and the abnormality detection unit 53 with a threshold value to judge defects, and therefore can detect the characteristics of various defects in the coating layer 102 of the specimen in a composite manner (for example, a type in which the crystal form of the drug varies depending on the site, a type in which the drug is not sufficiently loaded on the balloon 101, a type in which a large crystal mass 103 is present, a type in which the crystal mass 103 is present over a wide area, a type in which there is a defect in the coating layer 102, and / or a type in which a foreign object that should not be present is present, etc.), and can judge defects with high accuracy and easily. In particular, if the judger is designed so that the greater the degree of abnormality of the defect, the higher the degree of abnormality is detected, then the threshold value can be adjusted to set with high accuracy the level of abnormality required to be judged as a defect.

[0050] Furthermore, the coating layer inspection method in this embodiment is a coating layer inspection method for detecting defects in the coating layer 102 formed on the surface of a test object, and includes the steps of: step S10 of reading in a plurality of defect image data, which can be divided into a plurality of different defect types, as teacher data; step S12 of extracting features from the plurality of defect image data using a neural network to create a classifier; step S13 of quantifying and detecting the degree of abnormality from the created classifier using a support vector machine; step S14 of setting a threshold value of the degree of abnormality capable of detecting the features of a plurality of defects in a composite manner; step S15 of reading in image data to be judged; step S16 of extracting features from the image data to be judged and detecting the degree of abnormality of the image data to be judged using the created classifier and support vector machine; and steps S17 to S18 of comparing the degree of abnormality of the image data to be judged with a threshold value to determine whether the degree of abnormality is equal to or greater than the threshold value or not, and determining that there is a defect in the coating layer 102 of the image data to be judged if the determination is positive, and determining that there is no defect in the coating layer 102 of the image data to be judged if the determination is negative.

[0051] The coating layer inspection method configured as described above compares the degree of abnormality detected by the classifier and the support vector machine with a threshold value to determine defects, and therefore can detect a variety of defect features in a composite manner in the coating layer 102 of the subject, and can easily determine defects with high accuracy. In particular, if the classifier is designed so that the greater the degree of abnormality of a defect, the higher the degree of abnormality that is detected, then by adjusting the threshold value, it is possible to set with high accuracy the level of abnormality required to determine a defect.

[0052] The specimen is a balloon 101 that can be inserted into a biological lumen, and the coating layer 102 contains a drug. As a result, the coating layer inspection device 10 and the coating layer inspection method can comprehensively detect the characteristics of various defects in the coating layer 102 that contains a drug on the surface of the balloon 101, and can easily determine the defects with high accuracy.

[0053] The present invention is not limited to the above-mentioned embodiment, and various modifications can be made by those skilled in the art within the technical concept of the present invention. For example, the specimen inspected by the coating layer inspection device 10 is not limited to the balloon 101. The specimen may be, for example, a medical device such as the balloon 101, a stent, a catheter, a guidewire, or an endoscope, or may be a non-medical device. The coating layer 102 provided on the specimen may be, for example, a layer formed of a drug, a layer formed of a low-friction material (for example, a fluororesin) for improving abrasion resistance, a layer formed of a hydrophilic polymer, or a layer formed of a polymer for preventing thrombus adhesion. More specifically, for example, the specimen may be a catheter having a hydrophilic coating as the coating layer 102, a guidewire having a polymer coating for preventing thrombus adhesion as the coating layer 102, a guidewire or an endoscope having a PTFE coating or PEEK coating for improving abrasion resistance as the coating layer 102, or the like.

[0054] Even if the coating layer 102 is transparent, the above-mentioned inspection may be possible by devising the irradiation of light onto the coating layer 102 by the lighting device 20, or by carrying out pre-processing on the coating layer 102 or the captured image. In the case where the surface of the coating layer 102 has an uneven surface, the coating layer 102 may be irradiated with light from a direction intersecting the direction of imaging by the imaging unit 30 (for example, a perpendicular direction). This allows shadows of different sizes and patterns depending on the shape of the coating layer 102 to be captured in the image, making it possible to carry out the above-mentioned inspection method. The number of illuminations does not need to be one. By irradiating the coating layer 102 from different directions, shadows of different sizes and patterns depending on the shape of the coating layer 102 are captured in the grayscale image, making it possible to carry out the above-mentioned inspection method using the grayscale image. When multiple illuminations are used, the types of illuminations used may be the same or different.

[0055] Pre-processing can be a destructive method such as applying a coloring process to the coating layer 102 to make the transparent coating layer 102 visible, or a non-destructive method such as applying a binarization process, contrast adjustment, noise control, etc. to the captured image to make it visible. [Explanation of symbols]

[0056] 10 Coating layer inspection device 30 Imaging unit 42 Processing section 50 Image acquisition unit 51 Area identification part 52 Decision tool creation section 53 Anomaly detection unit 54 Threshold setting unit 55 Defect Judgment Section 56 Result output section 101 Balloon (Subject) 102 Coating layer A Judgment area

Claims

1. A coating layer inspection device for detecting defects in a coating layer including a plurality of drug crystals formed on a surface of a subject, the coating layer being a balloon that can be inserted into a biological lumen, comprising: an imaging unit that images the coating layer and creates image data; an illumination device that illuminates the subject with light from a direction intersecting an imaging direction of the imaging unit to form a shadow due to unevenness on a surface of the subject; A processing unit that performs calculation processing, The processing unit includes an image acquisition unit capable of reading a plurality of pieces of defect image data classified into a plurality of different defect types and a judgment target image data; a classifier creation unit that extracts feature amounts from the plurality of defect image data and creates a classifier; an abnormality detection unit that detects a numerically quantified abnormality degree from the determiner; a defect determination unit that compares the degree of abnormality detected from the determination target image data using the determiner and the abnormality detection unit with a threshold value to determine whether or not the coating layer has a defect; The coating layer inspection device has at least two of the following types of defects: Type 1, in which the crystal form of the drug varies depending on the part of the balloon; Type 2, in which the drug is not sufficiently loaded on the balloon; Type 3, in which large crystal clumps are present; Type 4, in which crystal clumps are present over a wide area; Type 5, in which there is a defect in the coating layer; and Type 6, in which foreign matter is present.

2. A coating layer inspection method for detecting defects in a coating layer containing a plurality of drug crystals formed on a surface of a subject, the coating layer being a balloon that can be inserted into a biological lumen, comprising: reading a plurality of defect image data that can be classified into a plurality of different defect types as training data; A step of extracting feature amounts from the plurality of defect image data using a neural network to create a classifier; A step of detecting an anomaly degree by quantifying the anomaly degree from the created classifier using a support vector machine; setting a threshold value of the degree of abnormality that can comprehensively detect the characteristics of the plurality of defects; a step of irradiating the subject with light from a direction intersecting an imaging direction of an imaging unit that images the coating layer and creates image data, forming a shadow on the surface of the subject due to unevenness of the coating layer, and creating judgment target image data by the imaging unit; reading the determination target image data; extracting features from the determination target image data, and detecting an abnormality level of the determination target image data by the created determiner and the support vector machine; a step of comparing the degree of abnormality of the judgment target image data with the threshold value to determine whether the degree of abnormality is equal to or greater than the threshold value, and determining that the coating layer of the judgment target image data has a defect when the determination is positive, and determining that the coating layer of the judgment target image data has no defect when the determination is negative, The coating layer inspection method includes at least two of the following types of defects: Type 1, in which the crystal form of the drug varies depending on the part of the balloon; Type 2, in which the drug is not sufficiently loaded on the balloon; Type 3, in which large crystal clumps are present; Type 4, in which crystal clumps are present over a wide area; Type 5, in which there is a defect in the coating layer; and Type 6, in which foreign matter is present.

Images