Method and device for identifying and positioning patch

Abstract

The present application provides a method and device for identifying positioning patches, the method comprising: inputting a medical image into a pre-trained segmentation model, determining a positioning patch segmentation region and positioning patch coding, a signal generator segmentation region and signal generator coding, and a connecting line segmentation region and connecting line coding; determining at least one intersection coding corresponding to at least one intersection region based on the positioning patch coding, the signal generator coding, and the connecting line coding; screening multiple connected domains in at least one intersection region in sequence according to a preset order to identify the positioning patches included in the medical image, wherein the preset order is determined based on at least one intersection coding. The technical solution of the present application can quickly and efficiently identify positioning patches by screening connected domains according to a preset order determined by the intersection coding.

Description

Technical Field

[0001] The present application relates to the field of deep learning technology, and in particular to a method and device for identifying and positioning patches. Background Art

[0002] Automatic puncture algorithms can assist doctors in performing puncture operations, thereby greatly improving the speed and accuracy of puncture operations. The basic principle of using automatic puncture algorithms is to use positioning patches to locate the body, so that the puncture needle can be inserted into the body cavity to extract or inject drugs. It can be seen that in the technology of applying automatic puncture algorithms, the position of the positioning patch (also called positioning patch or body surface positioning patch, etc.) must be determined first.

[0003] In view of this, how to quickly and accurately obtain the position of the positioning patch has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a method and device for identifying a positioning patch, which can quickly and efficiently identify a positioning patch.

[0005] In a first aspect, an embodiment of the present application provides a method for identifying positioning patches, the method comprising: inputting a medical image into a pre-trained segmentation model, determining a positioning patch segmentation area and a positioning patch code, a signal generator segmentation area and a signal generator code, and a connecting line segmentation area and a connecting line code; based on the positioning patch code, the signal generator code and the connecting line code, determining at least one intersection code corresponding to at least one intersection area; and screening multiple connected domains in at least one intersection area in sequence according to a preset order to identify the positioning patch included in the medical image, wherein the preset order is determined based on at least one intersection code.

[0006] In certain embodiments of the present application, multiple connected domains in at least one intersection area are screened in sequence according to a preset order to identify a positioning patch included in a medical image, including: determining the largest connected domain in each intersection area one by one according to a preset order, wherein the multiple connected domains include the largest connected domain; when the largest connected domain includes a signal generator segmentation area, using the largest connected domain as the positioning patch.

[0007] In certain embodiments of the present application, when the number of positioning patches is two or more, the method further includes: determining whether the number of positioning patches reaches a preset number; when the number of positioning patches does not reach the preset number, determining whether there is a positioning patch in the next intersection area according to a preset order.

[0008] In certain embodiments of the present application, at least one intersection code corresponding to at least one intersection area is determined based on the positioning piece code, the signal generator code and the connecting line code, including: determining a first intersection area and a first intersection code based on the signal generator code, the positioning piece code and the connecting line code; determining a second intersection area and a second intersection code based on the signal generator code and the positioning piece code; determining a third intersection area and a third intersection code based on the signal generator code and the connecting line code; determining a fourth intersection area and a fourth intersection code based on the positioning piece code and the connecting line code, wherein the preset order is the first intersection area, the second intersection area, the third intersection area and the fourth intersection area.

[0009] In certain embodiments of the present application, before inputting the medical image into a pre-trained segmentation model and determining the positioning film segmentation area and positioning film coding, the signal generator segmentation area and signal generator coding, and the connecting line segmentation area and connecting line coding, it also includes: cropping the medical image with label information input into the initial network model by at least one cropping method to obtain regional blocks, wherein the medical image with label information includes an annotated area, the annotated area includes a positioning film annotated area, a signal generator annotated area and a connecting line annotated area, and the signal generator annotated area is located within the positioning film annotated area; filling the regional blocks onto the human body surface to form training data; and training the initial network model according to the training data to obtain a segmentation model.

[0010] In certain embodiments of the present application, a medical image with label information input into an initial network model is cropped by at least one cropping method to obtain regional blocks, including: cropping the medical image with label information by at least one cropping method according to a preset ratio to obtain regional blocks.

[0011] In certain embodiments of the present application, at least one cropping method includes a first cropping method of cropping at the center of the marked area, a second cropping method of cropping at the center of the signal generator marked area, and a third cropping method of cropping at twice the distance from the center of the signal generator marked area to the origin of the preset coordinate axis, wherein the preset ratio is 5 for the first cropping method, 3 for the second cropping method, and 2 for the third cropping method.

[0012] In certain embodiments of the present application, the method further includes: increasing or decreasing the HU value of the area within the positioning piece marked area except the signal generator marked area to a preset value, so that the initial network model can learn the connecting line area and the positioning piece area except the signal generator area.

[0013] In certain embodiments of the present application, regional blocks are filled into the human body surface to form training data, including: filling regional blocks into multiple positions of the human body surface, and deleting part of the regional blocks that exceed the human body surface to obtain training data.

[0014] In a second aspect, an embodiment of the present application provides a device for identifying positioning patches, the device comprising: a first determination module, used to input a medical image into a pre-trained segmentation model, determine a positioning patch segmentation area and a positioning patch code, a signal generator segmentation area and a signal generator code, and a connecting line segmentation area and a connecting line code; a second determination module, used to determine at least one intersection code corresponding to at least one intersection area based on the positioning patch code, the signal generator code and the connecting line code; an identification module, used to sequentially screen multiple connected domains in at least one intersection area according to a preset order to identify the positioning patch included in the medical image, wherein the preset order is determined based on at least one intersection code.

[0015] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, the storage medium storing a computer program, and the computer program is used to execute the method for identifying and positioning patches described in the first aspect above.

[0016] In a fourth aspect, an embodiment of the present application provides an electronic device, comprising: a processor; and a memory for storing processor executable instructions, wherein the processor is used to execute the method for identifying and positioning the patch described in the first aspect above.

[0017] The embodiment of the present application provides a method and device for identifying positioning patches, by encoding the intersection area (i.e., intersection coding), and screening the connected domains in the intersection area according to a preset order determined by the intersection coding, and outputting the connected domains that meet the conditions in the intersection area (i.e., the identified positioning patches), so that the technical solution of the present application can quickly and efficiently identify the positioning patches, and at the same time improve the speed of identifying the positioning patches while reducing the false detection rate of the largest connected domain. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of a method for identifying and positioning a patch provided by an exemplary embodiment of the present application.

[0019] Figure 2 It is a schematic diagram of the structure of a positioning patch provided by an exemplary embodiment of the present application.

[0020] Figure 3 It is a flowchart of a method for identifying and positioning a patch provided by another exemplary embodiment of the present application.

[0021] Figure 4It is a flowchart of a method for identifying and positioning a patch provided by another exemplary embodiment of the present application.

[0022] Figure 5 It is a flowchart of a method for identifying and positioning a patch provided by yet another exemplary embodiment of the present application.

[0023] Figure 6 It is a schematic diagram of the structure of a device for identifying and positioning a patch provided by an exemplary embodiment of the present application.

[0024] Figure 7 It is a block diagram of an electronic device for identifying a positioning patch provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0026] At present, when assisting doctors in puncture surgery, it is necessary to place a positioning patch on the human body surface to locate the relative position between the human body surface and medical images (such as computer tomography). In order to obtain the specific position of the positioning patch, the following two methods are mainly used.

[0027] Method 1: Use the method of rough positioning of the positioning patch. First, find the fuzzy position of the positioning patch in the medical image of the human body (i.e. the rough positioning position). Then, enter the next stage based on this rough positioning position and perform precise positioning of the positioning patch in the trained model. Method 2: Mainly use the deep learning segmentation method to directly segment the positioning patch, and then select the largest connected domains as the positioning patch output.

[0028] However, when the positioning patches are identified using the above two methods, since the density of the positioning patches is similar to the density of the segmented scenes, it will lead to inaccurate positioning patch segmentation results, high false positive rate of maximum connected domain extraction, and cumbersome identification process.

[0029] In order to solve the above technical problems, the present application proposes the following method for identifying and positioning patches.

[0030] Figure 1 It is a flowchart of a method for identifying and positioning a patch provided by an exemplary embodiment of the present application. Figure 1 The method is performed by a computing device, such as a server. Figure 1 As shown, the method for identifying and locating a patch includes the following contents.

[0031] 110: Input the medical image into a pre-trained segmentation model to determine the localizer segmentation region and localizer coding, the signal generator segmentation region and signal generator coding, and the connection line segmentation region and connection line coding.

[0032] Specifically, see Figure 2 The positioning patch may include a positioning patch area 201, a signal generator area 202, and a connection line area 203. The signal generator area 202 may be in a "human" shape, located inside the positioning patch area 201, and connected to the connection line area 203. The positioning patch area 201 may be circular or square, and the embodiment of the present application does not specifically limit the specific structure of the positioning patch.

[0033] The server can input the medical image input by the user into the pre-trained segmentation model to obtain the positioning film segmentation area, the signal generator segmentation area and the connecting line segmentation area. Then the server or manually encodes the positioning film segmentation area, the signal generator segmentation area and the connecting line segmentation area to obtain the positioning film code, the signal generator code and the connecting line code. Among them, the positioning film code, the signal generator code and the connecting line code are different from each other and are all unique codes, that is, one area category corresponds to one code, and the embodiment of the present application does not limit the specific code set.

[0034] For example, the positioning piece is coded as 1, the signal generator is coded as 2, and the connecting line is coded as 4.

[0035] It should be noted that the medical image may be a computed tomography (CT) image or a magnetic resonance imaging (MRI) image, and the embodiments of the present application do not specifically limit the specific type of the medical image.

[0036] Preferably, the embodiment of the present application uses CT images with fast scanning time and clear images as medical images.

[0037] It should also be noted that the segmentation model is used to segment and classify the positioning patch. The segmentation model needs to output binary results of four channels. The four channels are independent channels. The four channels are used to predict the segmentation results and the binary classification results of the three categories of positioning patch, signal generator and connecting line.

[0038] It should also be noted that due to the large number of parameters of the segmentation model, the prediction data (i.e., medical images) can be cut into blocks for prediction. And since it is easy to cut the object to be detected (i.e., the positioning patch in the medical image) into different blocks during cutting, resulting in poor detection effect, the step size of the segmentation model fusion should be smaller than the step size of the cutting.

[0039] 120: Determine at least one intersection code corresponding to at least one intersection area based on the positioning piece code, the signal generator code, and the connection line code.

[0040] Specifically, the server may add the positioning piece code, the signal generator code and the connecting line code in pairs, and add the positioning piece code, the signal generator code and the connecting line code in three, to determine at least one intersection code corresponding to at least one intersection area, wherein the intersection code corresponding to at least one intersection area is unique.

[0041] In one embodiment, the signal generator code, the positioning piece code and the connecting line code are added to determine the first intersection area and the first intersection code. The signal generator code and the positioning piece code are added to determine the second intersection area and the second intersection code. The signal generator code and the connecting line code are added to determine the third intersection area and the third intersection code. The positioning piece code and the connecting line code are added to determine the fourth intersection area and the fourth intersection code.

[0042] For example, the positioning piece is coded 1, the signal generator is coded 2, and the connecting line is coded 4. At this time, the first intersection code is 7, the second intersection code is 3, the third intersection code is 6, and the fourth intersection code is 5.

[0043] 130: Filter multiple connected domains in at least one intersection region in sequence according to a preset order to identify a positioning patch included in the medical image.

[0044] In one embodiment, the predetermined order is determined based on at least one intersection code.

[0045] Specifically, since there is a signal generator area with a high HU value (Hounsfield unit, HU) inside the positioning patch, and the material of the signal generator area is different from that of the external positioning patch area, the segmentation model can be used to identify the positioning patch. Therefore, the preset order can be to sort the intersection area containing the signal generator segmentation area at the front, the one containing the positioning patch segmentation area second, and the one containing the connecting line segmentation area at the end.

[0046] In one embodiment, the preset order may be the first intersection area, the second intersection area, the third intersection area, and the fourth intersection area.

[0047] The server may determine the maximum connected domain in each intersection area one by one according to a preset order. Each intersection area may be any area of ​​the first intersection area, the second intersection area, the third intersection area, and the fourth intersection area, and each intersection area may include multiple connected domains, and the multiple connected domains include the maximum connected domain. When the maximum connected domain includes the signal generator segmentation area, the server may use the maximum connected domain as a positioning patch. When the maximum connected domain does not include the signal generator segmentation area, the server may determine whether the maximum connected domain in the next intersection area is a positioning patch according to a preset order.

[0048] For example, when the server detects that the first intersection area does not include the positioning patch, the server may detect the maximum connected domain in the second intersection area according to a preset order to determine whether the second intersection area includes the positioning patch.

[0049] The number of positioning patches may be one or more, and the embodiment of the present application does not specifically limit the number of positioning patches.

[0050] In one embodiment, when the number of positioning patches is two or more, the server may determine whether the number of positioning patches currently determined has reached a preset number. When the number of positioning patches currently determined has not reached a preset number, the server may determine whether there is a positioning patch in the next intersection area according to a preset order. When the number of positioning patches currently determined has reached a preset number, the server may stop detecting the intersection area.

[0051] It can be seen from this that the embodiment of the present application encodes the intersection area (i.e., intersection coding), and screens the connected domains in the intersection area according to a preset order determined by the intersection coding, and outputs the connected domains that meet the conditions in the intersection area (i.e., the identified positioning patches), so that the technical solution of the present application can quickly and efficiently identify the positioning patches, and while reducing the false detection rate of the largest connected domain, it improves the speed of identifying the positioning patches.

[0052] Figure 3 It is a flowchart of a method for identifying and positioning a patch provided by another exemplary embodiment of the present application. Figure 3 The embodiment is Figure 1 The examples of the embodiments are not repeated here, and the differences are described here. Figure 3 As shown, the method for identifying and locating a patch includes the following contents.

[0053] 310: Determine the maximum connected domain in each intersection area one by one according to a preset order, wherein the multiple connected domains include the maximum connected domain.

[0054] Specifically, the server may determine the maximum connected domain in each intersection area one by one according to a preset order. Each intersection area may be any area of ​​the first intersection area, the second intersection area, the third intersection area, and the fourth intersection area. Each intersection area may include multiple connected domains. The multiple connected domains may include the maximum connected domain.

[0055] 320: When the maximum connected domain includes the segmented area of ​​the signal generator, the maximum connected domain is used as a positioning patch.

[0056] Specifically, when the server detects that the maximum connected domain includes the segmented area of ​​the signal generator, the server may use the maximum connected domain including the segmented area of ​​the signal generator as the positioning patch.

[0057] It should be noted that the server can also delete the connected domains within a preset distance from the maximum connected domain, filter out the connected domains with a distance less than the preset distance, and only retain the larger connected domains in the intersection area. The preset distance can be 1000 pixels, and the embodiment of the present application does not specifically limit the preset distance.

[0058] It can be seen that the embodiment of the present application reduces the false detection rate of the largest connected domain by screening the largest connected domains in the intersection areas one by one and using the largest connected domain including the segmented area of ​​the signal generator as the positioning patch.

[0059] Figure 4 It is a flowchart of a method for identifying and positioning a patch provided by another exemplary embodiment of the present application. Figure 4 The embodiment is Figure 1 The examples of the embodiments are not repeated here, and the differences are described here. Figure 4 As shown, the method for identifying and locating a patch includes the following contents.

[0060] 410: Determine whether the number of positioning patches reaches a preset number.

[0061] Specifically, the number of positioning patches can be two or more, that is, the number of positioning patches can be multiple, and the embodiment of the present application does not specifically limit the number of positioning patches. The server can detect the number of identified positioning patches, that is, the server can determine whether the number of currently identified positioning patches reaches a preset number.

[0062] 420: When the number of positioning tiles does not reach the preset number, determine whether there is a positioning tile in the next intersection area according to a preset order.

[0063] Specifically, when the number of identified positioning tiles does not reach the preset number, the server may determine whether there is a positioning tile in the next intersection area according to the preset order. The next intersection area may be the next intersection area after the current intersection area determined according to the preset order. For example, if the current intersection area is the first intersection area, then according to the preset order, it can be known that the next intersection area is the second intersection area.

[0064] In another embodiment, when the number of identified positioning tiles reaches a preset number, the server may stop detecting the intersection area.

[0065] It should be noted that when multiple positioning patches are identified according to a preset order, the server can first filter out duplicate positioning patches that are close in distance (for example, positioning patches with a distance of less than 1000 pixels). If there are still more than the preset number of positioning patches after filtering, the server can select the connected domains with the largest preset number of areas among the positioning patches as the positioning patches for this identification.

[0066] Preferably, the preset number is 3.

[0067] It can be seen that the embodiment of the present application screens the positioning patches in a preset order, so that the embodiment of the present application can efficiently and reasonably screen out a specified number of positioning patches.

[0068] In one embodiment of the present application, based on the positioning piece code, the signal generator code and the connecting line code, at least one intersection code corresponding to at least one intersection area is determined, including: based on the signal generator code, the positioning piece code and the connecting line code, a first intersection area and a first intersection code are determined; based on the signal generator code and the positioning piece code, a second intersection area and a second intersection code are determined; based on the signal generator code and the connecting line code, a third intersection area and a third intersection code are determined; based on the positioning piece code and the connecting line code, a fourth intersection area and a fourth intersection code are determined, wherein the preset order is the first intersection area, the second intersection area, the third intersection area and the fourth intersection area.

[0069] Specifically, the server may add the signal generator code, the positioning piece code, and the connecting line code to determine the first intersection area and the first intersection code. The first intersection area is the intersection of the positioning piece segmentation area, the signal generator segmentation area, and the connecting line segmentation area. The first intersection code is the sum of the signal generator code, the positioning piece code, and the connecting line code.

[0070] The server may add the signal generator code and the positioning piece code to determine a second intersection area and a second intersection code, wherein the second intersection area is the intersection of the signal generator segmentation area and the positioning piece segmentation area, and the second intersection code is the sum of the signal generator code and the positioning piece code.

[0071] The server may add the signal generator code and the connecting line code to determine a third intersection area and a third intersection code, wherein the third intersection area is the intersection of the signal generator segmentation area and the connecting line segmentation area, and the third intersection code is the sum of the signal generator code and the connecting line code.

[0072] The server may add the positioning piece code and the connecting line code to determine a fourth intersection area and a fourth intersection code. The fourth intersection area is the intersection of the positioning piece segmentation area and the connecting line segmentation area. The fourth intersection area is the sum of the positioning piece code and the connecting line code. In addition, the first intersection code, the second intersection code, the third intersection code, and the fourth intersection code are all unique codes and are different from each other.

[0073] For example, the positioning piece is coded as 1, the signal generator is coded as 2, and the connecting line is coded as 4. At this time, the first intersection is coded as 7, the second intersection is coded as 3, the third intersection is coded as 6, and the fourth intersection is coded as 5.

[0074] In one embodiment, the preset order may be the first intersection area, the second intersection area, the third intersection area, and the fourth intersection area.

[0075] It can be seen that the embodiment of the present application sets a preset screening order so that a specified number (eg, one or more) of positioning patches can be screened out efficiently and reasonably subsequently.

[0076] Figure 5 It is a flowchart of a method for identifying and positioning a patch provided by yet another exemplary embodiment of the present application. Figure 5 The embodiment is Figure 1 The examples of the embodiments are not repeated here, and the differences are described here. Figure 5 As shown, the method for identifying and locating a patch includes the following contents.

[0077] 510: Crop the medical image with label information input into the initial network model by at least one cropping method to obtain regional blocks.

[0078] In one embodiment, the medical image with label information may include a marked area, and the marked area includes a positioning film marked area, a signal generator marked area, and a connecting line marked area, and the signal generator marked area is located in the positioning film marked area.

[0079] It should be noted that since there are few positioning patches with label information, the positioning patches with label information need to be pasted on other data (such as human body surface) for enhanced training, so the positioning patches need to be cropped first.

[0080] Specifically, the initial network model can crop the medical image with label information input by the user through at least one cropping method to obtain regional blocks. Among them, the cropping method can be 2 or 3, etc., and the embodiment of the present application does not specifically limit this. The regional blocks can be obtained based on at least one cropping method, and the sizes of the regional blocks cropped by different cropping methods can be the same or different. The embodiment of the present application does not specifically limit the size of the regional blocks.

[0081] Preferably, the embodiment of the present application adopts three cutting methods for cutting.

[0082] In one embodiment, the three cropping methods can crop the medical image according to a preset ratio.

[0083] Medical images may be annotated with label information. The label information may be annotated by machine or by manual labeling. The present application does not specifically limit this. The area on the medical image annotated with label information may be used as an annotated area. That is, the annotated area may include a positioning film annotated area, a signal generator annotated area, and a connecting line annotated area. Figure 2 It can be seen from the positioning patch structure that the positioning patch marked area can include the signal generator marked area.

[0084] 520: Fill the area cuts onto the human body surface to form training data.

[0085] Specifically, the initial network model can fill the regional blocks obtained based on at least one cropping method into multiple positions on the human body surface to generate training data.

[0086] In one embodiment, due to the different sizes of the cut blocks, the situation of exceeding the human body boundary may occur during training. Therefore, the initial network model can delete the cut blocks of the partial area exceeding the human body boundary to avoid various cross-border occurrences.

[0087] 530: Train the initial network model according to the training data to obtain a segmentation model.

[0088] Specifically, the server may apply the training data to train the initial network model to obtain the desired segmentation model.

[0089] 540: Input the medical image into a pre-trained segmentation model to determine the localizer segmentation region and localizer coding, the signal generator segmentation region and signal generator coding, and the connection line segmentation region and connection line coding.

[0090] 550: Determine at least one intersection code corresponding to at least one intersection area based on the positioning piece code, the signal generator code, and the connection line code.

[0091] 560: Filter multiple connected domains in at least one intersection region in sequence according to a preset order to identify a positioning patch included in the medical image.

[0092] In one embodiment, the predetermined order is determined based on at least one intersection code.

[0093] It can be seen that the embodiment of the present application ensures that the training data of the positioning patch in various situations can be generated by filling the cropped area blocks in various parts of the human body surface, thereby enriching the model's learning of the positioning patch. In addition, the subsequent positioning patch sorting and screening method can also solve the problem that the density of the positioning patch is close to the density of the scene to be segmented, and at the same time solve the problem of high false detection rate of the maximum connected domain extraction.

[0094] In one embodiment of the present application, a medical image with label information input into an initial network model is cropped by at least one cropping method to obtain regional blocks, including: cropping the medical image with label information according to a preset ratio by at least one cropping method to obtain regional blocks.

[0095] Specifically, the server may crop the medical image in at least one cropping method according to a preset ratio to obtain regional blocks, wherein the sizes of the regional blocks corresponding to different cropping methods may be the same or different, which is not specifically limited in the embodiments of the present application.

[0096] In one embodiment, the number of cropping methods is 3. The three cropping methods are a first cropping method for cropping at the center of the marked area, a second cropping method for cropping at the center of the signal generator marked area, and a third cropping method for cropping at twice the distance from the center of the signal generator marked area to the origin of the preset coordinate axis. Among them, the preset coordinate axis can be determined according to the LPS (Left, Posterior, Superior) coordinate axis of the medical image. The origin of the preset coordinate axis can be the origin of the LPS coordinate system of the signal generator marking area. In addition, the preset ratio can be 5 for the first cropping method, 3 for the second cropping method, and 2 for the third cropping method.

[0097] It can be seen that the embodiment of the present application adopts three cropping strategies to perform random sampling in proportion, so that the initial network model can learn the overall and local information of the positioning patch, and also avoids the problem of using a single cropping method, which makes the model learning unable to take into account both the overall and local aspects.

[0098] In one embodiment of the present application, at least one cropping method includes a first cropping method of cropping at the center of the marked area, a second cropping method of cropping at the center of the signal generator marked area, and a third cropping method of cropping at twice the distance from the center of the signal generator marked area to the origin of the preset coordinate axis, wherein the preset ratio is 5 for the first cropping method, 3 for the second cropping method, and 2 for the third cropping method.

[0099] Specifically, the first cropping method can be to use the center of the labeled area as the midpoint and perform cropping within a preset range. It should be noted that although the first cropping method can enable the initial network model to better learn the overall structure of the positioning patch and obtain better connecting line segmentation areas. However, due to the long connecting lines of the positioning patch, the area cut is larger, which makes the signal generator labeled area in some positioning patches, in the subsequent process of sticking to the human body surface to construct training data, there is a risk of the signal generator labeled area being truncated due to the excessive length of the area cut, thereby affecting the model's learning effect on the connecting line labeled area.

[0100] The second cropping method can be to use the center of the signal generator labeled area as the midpoint and perform cropping within a preset range to obtain a regional cut. The second cropping method allows the initial network model to better learn the signal generator labeled area, but will have a weaker learning of the connection line labeled area.

[0101] The third cropping method can be to crop the area with twice the distance from the midpoint of the signal generator marked area to the origin of the preset coordinate axis to obtain the area cut. Among them, the preset coordinate axis can be determined according to the LPS (Left, Posterior, Superior) coordinate axis of the medical image. The origin of the preset coordinate axis can be the origin of the LPS coordinate system of the signal generator marked area. The third cropping method can enable the initial network model to better learn the signal generator marked area, but because the area cut is small, the initial network model learns less about the connecting line marked area, which in turn causes the trained segmentation model to be unable to segment the connecting line segmentation area well.

[0102] In one embodiment, the preset ratio may be that the first cropping method accounts for 5, the second cropping method accounts for 3, and the third cropping method accounts for 2.

[0103] It should be noted that, since different cropping methods have their own advantages and disadvantages, the embodiment of the present application adopts three cropping strategies to perform cropping in a random sampling manner, thereby making the initial network model learning more diverse.

[0104] It can be seen that the embodiment of the present application adopts three types of proportional cropping, so that the model can learn the overall and local information of the positioning patch, thereby improving the segmentation accuracy of the model.

[0105] In one embodiment of the present application, the method also includes: increasing the HU value of the area within the positioning piece marked area except the signal generator marked area, or decreasing the preset value, so that the initial network model can learn the connecting line area and the positioning piece area except the signal generator area.

[0106] Specifically, the server may use random HU value jittering for the area inside the positioning patch except the area marked by the signal generator, so that the initial network model can further learn the HU values ​​of the connecting line marked area and the area inside the positioning patch except the area marked by the signal generator.

[0107] The initial network model can adjust the HU value of the area other than the signal generator marked area in the positioning piece marked area to a preset value, so that the initial network model can learn the connection line area and the positioning piece excluding the signal generator area. Alternatively, the initial network model can adjust the HU value of the area other than the signal generator marked area in the positioning piece marked area to a preset value, so that the initial network model can learn the connection line area and the positioning piece excluding the signal generator area. The embodiment of the present application does not specifically limit the preset value to be adjusted.

[0108] For example, when the HU value of the area in the positioning film annotation area other than the signal generator annotation area is 100, the initial network model can adjust its HU value from 100 to 90 and then perform learning.

[0109] It should be noted that since the training data of the positioning patch is usually small and there is more than one type of positioning patch, the density of the positioning patch is different. In addition, the density of the positioning patch in the test set is equivalent to the unknown body membrane of the experiment. Therefore, if the model is too sensitive to the HU value, the segmentation effect of the segmentation model will be poor, and it will be difficult to identify new positioning patches. In addition, the segmentation effect of the segmentation model will also vary greatly with the volatility of the HU value inside the positioning patch, resulting in insufficient robustness of the segmentation model.

[0110] It can be seen that the embodiment of the present application improves the model segmentation accuracy and makes the model more robust by adopting the method of random HU value jitter.

[0111] In one embodiment of the present application, regional blocks are filled into the human body surface to form training data, including: filling regional blocks into multiple positions of the human body surface, and deleting part of the regional blocks that exceed the human body surface to obtain training data.

[0112] Specifically, the server can fill the regional slices at multiple positions on the human body surface so that the initial network model can learn the situation of the positioning patch at any position on the human body surface. At the same time, the server can also delete some of the regional slices that exceed the human body surface to facilitate the acquisition of training data and avoid various cross-border occurrences.

[0113] It should be noted that the above-mentioned cutting and deleting of the partial area beyond the boundary of the human body surface can be understood as a normalized operation for obtaining training data.

[0114] It can be seen that the present application forms training data by filling the regional blocks onto the human body surface, thus providing a basis for subsequent training of the initial network model.

[0115] Figure 6 FIG. 6 is a schematic diagram of a device 600 for identifying and positioning a patch according to an exemplary embodiment of the present application. Figure 6 As shown, the device 600 for identifying and positioning a patch includes: a first acquisition module 610 , a filling module 620 , a second acquisition module 630 , a first determination module 640 , a second determination module 650 and an identification module 660 .

[0116] The first determination module 640 is used to input the medical image into a pre-trained segmentation model to determine the positioning piece segmentation area and positioning piece code, the signal generator segmentation area and signal generator code, and the connecting line segmentation area and connecting line code; the second determination module 650 is used to determine at least one intersection code corresponding to at least one intersection area based on the positioning piece code, the signal generator code and the connecting line code; the identification module 660 is used to sequentially screen multiple connected domains in at least one intersection area according to a preset order to identify the positioning patch included in the medical image, wherein the preset order is determined based on at least one intersection code.

[0117] An embodiment of the present application provides a device for identifying positioning patches, which encodes the intersection area (i.e., intersection coding), and screens the connected domains in the intersection area according to a preset order determined by the intersection coding, and outputs the connected domains that meet the conditions in the intersection area (i.e., the identified positioning patches), so that the technical solution of the present application can quickly and efficiently identify the positioning patches, and while reducing the false detection rate of the largest connected domain, improves the speed of identifying the positioning patches.

[0118] According to one embodiment of the present application, the identification module 660 is used to determine the maximum connected domain in each intersection area one by one according to a preset order, wherein multiple connected domains include the maximum connected domain; when the maximum connected domain includes the signal generator segmentation area, the maximum connected domain is used as a positioning patch.

[0119] According to an embodiment of the present application, when the number of positioning patches is two or more, the identification module 660 is used to determine whether the number of positioning patches reaches a preset number; when the number of positioning patches does not reach the preset number, determine whether there is a positioning patch in the next intersection area according to a preset order.

[0120] According to an embodiment of the present application, the second determination module 650 is used to determine the first intersection area and the first intersection code based on the signal generator code, the positioning piece code and the connecting line code; determine the second intersection area and the second intersection code based on the signal generator code and the positioning piece code; determine the third intersection area and the third intersection code based on the signal generator code and the connecting line code; determine the fourth intersection area and the fourth intersection code based on the positioning piece code and the connecting line code, wherein the preset order is the first intersection area, the second intersection area, the third intersection area and the fourth intersection area.

[0121] According to an embodiment of the present application, the first acquisition module 610 is used to crop the medical image with label information input into the initial network model through at least one cropping method to obtain regional blocks, wherein the medical image with label information includes an annotated area, the annotated area includes a positioning film annotated area, a signal generator annotated area and a connecting line annotated area, and the signal generator annotated area is located within the positioning film annotated area; the filling module 620 is used to fill the regional blocks into the human body surface to form training data; the second acquisition module 630 is used to train the initial network model according to the training data to obtain a segmentation model.

[0122] According to an embodiment of the present application, the first acquisition module 610 is used to crop the medical image with label information according to a preset ratio using at least one cropping method to obtain a regional block.

[0123] According to one embodiment of the present application, at least one cropping method includes a first cropping method of cropping at the center of the marked area, a second cropping method of cropping at the center of the signal generator marked area, and a third cropping method of cropping at twice the distance from the center of the signal generator marked area to the origin of the preset coordinate axis, wherein the preset ratio is 5 for the first cropping method, 3 for the second cropping method, and 2 for the third cropping method.

[0124] According to one embodiment of the present application, the first acquisition module 610 is used to increase or decrease the HU value of the area within the positioning piece marked area except the signal generator marked area to a preset value, so that the initial network model can learn the connecting line area and the positioning piece except the signal generator area.

[0125] According to an embodiment of the present application, the filling module 620 is used to fill the regional slices to multiple positions on the human body surface, and delete part of the regional slices that exceed the human body surface to obtain training data.

[0126] It should be understood that the specific working processes and functions of the first acquisition module 610, the filling module 620, the second acquisition module 630, the first determination module 640, the second determination module 650 and the identification module 660 in the above embodiment can refer to the above Figures 1 to 5 The description of the method for identifying and positioning the patch provided in the embodiment will not be repeated here to avoid repetition.

[0127] Figure 7 is a block diagram of an electronic device 700 for identifying a positioning patch provided by an exemplary embodiment of the present application.

[0128] Reference Figure 7 The electronic device 700 includes a processing component 710, which further includes one or more processors, and a memory resource represented by a memory 720 for storing instructions executable by the processing component 710, such as an application. The application stored in the memory 720 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 710 is configured to execute instructions to perform the above-mentioned method of identifying and positioning patches.

[0129] The electronic device 700 may also include a power supply component configured to perform power management of the electronic device 700, a wired or wireless network interface configured to connect the electronic device 700 to a network, and an input / output (I / O) interface. The electronic device 700 may be operated based on an operating system stored in the memory 720, such as Windows Server 2000. TM , Mac OSX TM , Unix TM , Linux TM , FreeBSD TM or similar.

[0130] A non-temporary computer-readable storage medium, when the instructions in the storage medium are executed by the processor of the above-mentioned electronic device 700, enables the above-mentioned electronic device 700 to execute a method for identifying a positioning patch, including: inputting a medical image into a pre-trained segmentation model, determining a positioning patch segmentation area and a positioning patch code, a signal generator segmentation area and a signal generator code, and a connecting line segmentation area and a connecting line code; based on the positioning patch code, the signal generator code and the connecting line code, determining at least one intersection code corresponding to at least one intersection area; and sequentially screening multiple connected domains in at least one intersection area according to a preset order to identify the positioning patch included in the medical image, wherein the preset order is determined based on at least one intersection code.

[0131] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, and will not be described in detail here.

[0132] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0133] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0134] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0135] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0136] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0137] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program check codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0138] It should be noted that, in the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0139] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for identifying and positioning a patch, characterized in that: include: Inputting the medical image into a pre-trained segmentation model, determining the localizer segmentation region and localizer coding, the signal generator segmentation region and signal generator coding, and the connection line segmentation region and connection line coding; Determine at least one intersection code corresponding to at least one intersection area based on the positioning piece code, the signal generator code and the connecting line code; screening a plurality of connected domains in the at least one intersection region in sequence according to a preset order to identify the positioning patch included in the medical image, wherein the preset order is determined based on the at least one intersection code; The determining, based on the positioning piece code, the signal generator code and the connecting line code, at least one intersection code corresponding to at least one intersection area comprises: Adding the positioning piece code, the signal generator code and the connecting line code in pairs, and adding the positioning piece code, the signal generator code and the connecting line code in three, to determine the at least one intersection code corresponding to the at least one intersection area; The intersection code corresponding to the at least one intersection area is unique.

2. The method according to claim 1, characterized in that The step of sequentially screening the plurality of connected domains in the at least one intersection region according to a preset order to identify the positioning patch included in the medical image includes: Determine the largest connected domain in each intersection area one by one according to the preset order, wherein the multiple connected domains include the largest connected domain; When the maximum connected domain includes the signal generator segmentation area, the maximum connected domain is used as the positioning patch.

3. The method according to claim 1, characterized in that When the number of the positioning patches is two or more, the method further comprises: Determining whether the number of the positioning patches reaches a preset number; When the number of the positioning tiles does not reach the preset number, it is determined whether the positioning tile exists in the next intersection area according to the preset order.

4. The method according to claim 1, characterized in that: The determining, based on the positioning piece code, the signal generator code and the connecting line code, at least one intersection code corresponding to at least one intersection area comprises: Determine a first intersection area and a first intersection code based on the signal generator code, the positioning piece code and the connecting line code; Determine a second intersection area and a second intersection code based on the signal generator code and the positioning piece code; Determine a third intersection area and a third intersection code based on the signal generator code and the connection line code; Based on the positioning piece code and the connecting line code, a fourth intersection area and a fourth intersection code are determined, Among them, the preset order is the first intersection area, the second intersection area, the third intersection area and the fourth intersection area.

5. The method according to any one of claims 1 to 4, characterized in that In the method of inputting the medical image into the pre-trained segmentation model, before determining the localizer segmentation region and localizer coding, the signal generator segmentation region and signal generator coding, and the connecting line segmentation region and connecting line coding, the method further includes: The medical image with label information input into the initial network model is cropped by at least two cropping methods to obtain regional blocks, wherein the medical image with label information includes a marked area, and the marked area includes a positioning piece marked area, a signal generator marked area and a connecting line marked area, and the signal generator marked area is located in the positioning piece marked area; Cutting the area into blocks and filling them onto the human body surface to form training data; The initial network model is trained according to the training data to obtain the segmentation model.

6. The method according to claim 5, characterized in that The medical image with label information input into the initial network model is cropped by at least two cropping methods to obtain regional blocks, including: The medical image with label information is cropped using the at least two cropping methods according to a preset ratio to obtain the regional cut.

7. The method according to claim 6, characterized in that The at least two cropping methods include a first cropping method of cropping at the center of the marked area, a second cropping method of cropping at the center of the signal generator marked area, and a third cropping method of cropping at twice the distance from the center of the signal generator marked area to the origin of the preset coordinate axis. Among them, the preset ratio is that the first cropping method accounts for 5, the second cropping method accounts for 3, and the third cropping method accounts for 2.

8. The method according to claim 5, characterized in that The method further comprises: The HU value of the area in the positioning piece marked area except the area marked by the signal generator is increased, or the preset value is decreased, so that the initial network model can learn the connecting line area and the area in the positioning piece marked area except the area marked by the signal generator.

9. The method according to claim 5, characterized in that The step of filling the area cuts into blocks on the human body surface to form training data includes: The regional slices are filled into a plurality of positions on the human body surface, and a portion of the regional slices beyond the human body surface is deleted to obtain the training data.

10. A device for identifying and positioning a patch, characterized in that: include: A first determination module is used to input the medical image into a pre-trained segmentation model to determine the localizer segmentation region and localizer coding, the signal generator segmentation region and signal generator coding, and the connecting line segmentation region and connecting line coding; A second determination module, configured to determine at least one intersection code corresponding to at least one intersection area based on the positioning piece code, the signal generator code and the connecting line code; an identification module, configured to sequentially screen a plurality of connected domains in the at least one intersection region according to a preset order to identify a positioning patch included in the medical image, wherein the preset order is determined based on the at least one intersection code; The second determination module is used to: add the positioning piece code, the signal generator code and the connecting line code in pairs, and add the positioning piece code, the signal generator code and the connecting line code in three, so as to determine the at least one intersection code corresponding to the at least one intersection area; The intersection code corresponding to the at least one intersection area is unique.

11. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to execute the method for identifying and positioning a patch according to any one of claims 1 to 9.

12. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor, Wherein, the processor is used to execute the method for identifying and locating the patch as described in any one of claims 1 to 9 above.

Images