CN114004807A

CN114004807A - Method and device for identifying positioning patch

Info

Publication number: CN114004807A
Application number: CN202111276593.0A
Authority: CN
Inventors: 郝增号; 刘恩佑; 陈宽; 王少康
Original assignee: Infervision Medical Technology Co Ltd
Current assignee: Wuhan Longdianjing Intelligent Technology Co ltd
Priority date: 2021-10-29
Filing date: 2021-10-29
Publication date: 2022-02-01

Abstract

The application provides a method and a device for identifying a positioning patch, wherein the method comprises the following steps: inputting the medical image into a pre-trained segmentation model, and determining a positioning sheet segmentation region and a positioning sheet code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code; determining at least one intersection code corresponding to the at least one intersection region based on the locating piece code, the signal generator code and the connecting line code; and sequentially screening a plurality of connected domains in the at least one intersection region according to a preset sequence to identify the positioning patches included in the medical image, wherein the preset sequence is determined based on the at least one intersection code. The technical scheme of this application screens the connected domain through the order of predetermineeing according to the intersection code is confirmed, can discern the site paster fast, efficient.

Description

Method and device for identifying positioning patch

Technical Field

The application relates to the technical field of deep learning, in particular to a method and a device for identifying a positioning patch.

Background

The automatic puncture algorithm can assist a doctor in performing a puncture operation, so that the speed and the accuracy of the puncture operation are greatly improved. The basic principle of the automatic puncture algorithm is that the positioning patch is used for positioning, and a puncture needle is used for puncturing a body cavity to extract or inject medicines and the like. It can be seen that in techniques that employ automated lancing algorithms, it is first necessary to determine the location of a site (also referred to as a topogram or body surface topogram, etc.).

In view of this, how to quickly and accurately acquire the position of the site becomes a technical problem to be solved urgently.

Disclosure of Invention

In view of this, embodiments of the present application provide a method and an apparatus for identifying a site, which can quickly and efficiently identify the site.

In a first aspect, embodiments of the present application provide a method of identifying a site, the method including: inputting the medical image into a pre-trained segmentation model, and determining a positioning sheet segmentation region and a positioning sheet code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code; determining at least one intersection code corresponding to the at least one intersection region based on the locating piece code, the signal generator code and the connecting line code; and sequentially screening a plurality of connected domains in the at least one intersection region according to a preset sequence to identify the positioning patches included in the medical image, wherein the preset sequence is determined based on the at least one intersection code.

In some embodiments of the present application, sequentially filtering a plurality of connected components in at least one intersection region according to a predetermined sequence to identify a site included in a medical image, includes: determining the maximum connected domains in each intersection region one by one according to a preset sequence, wherein the plurality of connected domains comprise the maximum connected domains; when the signal generator division area is included in the maximum connected component area, the maximum connected component area is used as the positioning patch.

In some embodiments of the present application, where the number of site locations is two or more, the method further comprises: judging whether the number of the positioning patches reaches a preset number or not; and when the number of the positioning patches does not reach the preset number, determining whether the positioning patches exist in the next intersection area or not according to the preset sequence.

In some embodiments of the present application, determining at least one intersection code corresponding to the at least one intersection region based on the topogram code, the signal generator code, and the connection line code includes: determining a first intersection region and a first intersection code based on the signal generator code, the spacer code and the connecting line code; determining a second intersection region and a second intersection code based on the signal generator code and the spacer code; determining a third intersection region and a third intersection code based on the signal generator code and the connecting line code; and determining a fourth intersection region and a fourth intersection code based on the locating piece code and the connecting line code, wherein the preset sequence is a first intersection region, a second intersection region, a third intersection region and the fourth intersection region in sequence.

In some embodiments of the present application, before inputting the medical image into the pre-trained segmentation model, determining the spacer segmentation region and the spacer code, the signal generator segmentation region and the signal generator code, and the connection line segmentation region and the connection line code, the method further includes: cutting the medical image with the label information input into the initial network model in at least one cutting mode to obtain an area cut block, wherein the medical image with the label information comprises a marking area, the marking area comprises a positioning sheet marking area, a signal generator marking area and a connecting line marking area, and the signal generator marking area is positioned in the positioning sheet marking area; filling the region cut blocks into the body surface of a human body to form training data; and training the initial network model according to the training data to obtain a segmentation model.

In some embodiments of the present application, the cropping the medical image with label information input into the initial network model by at least one cropping method to obtain the region blocks comprises: and cutting the medical image with the label information according to at least one cutting mode in a preset proportion to obtain the region blocks.

In some embodiments of the present application, the at least one cutting manner includes a first cutting manner in which cutting is performed by using the center of the labeled region, a second cutting manner in which cutting is performed by using the center of the labeled region of the signal generator, and a third cutting manner in which cutting is performed by using 2 times the distance from the center of the labeled region of the signal generator to the origin of the preset coordinate axis, where the preset ratio is that the ratio of the first cutting manner is 5, the ratio of the second cutting manner is 3, and the ratio of the third cutting manner is 2.

In certain embodiments of the present application, the method further comprises: the HU value of the region in the positioning sheet labeling region except the signal generator labeling region is increased or decreased to preset a numerical value, so that the initial network model can learn the connecting line region and the positioning sheet except the signal generator region.

In some embodiments of the present application, the region cut is filled into a body surface of a human body to form training data, including: and filling the region blocks into a plurality of positions on the body surface of the human body, and deleting partial region blocks exceeding the region blocks on the body surface of the human body to acquire training data.

In a second aspect, embodiments of the present application provide a device for identifying a site, the device comprising: the first determining module is used for inputting the medical image into a pre-trained segmentation model, and determining a positioning sheet segmentation region and a positioning sheet code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code; the second determining module is used for determining at least one intersection code corresponding to at least one intersection region based on the locating piece code, the signal generator code and the connecting line code; the identification module is used for sequentially screening a plurality of connected domains in the at least one intersection region according to a preset sequence so as to identify the positioning patches included in the medical image, wherein the preset sequence is determined based on the at least one intersection code.

In a third aspect, embodiments of the present application provide a computer-readable storage medium storing a computer program for executing the method for identifying a site described in the first aspect.

In a fourth aspect, an embodiment of the present application provides an electronic device, including: a processor; a memory for storing processor-executable instructions, wherein the processor is configured to perform the method of identifying a site of the first aspect.

The embodiment of the application provides a method and a device for identifying a positioning patch, wherein an intersection region is coded (namely, the intersection code), connected domains in the intersection region are screened according to a preset sequence determined by the intersection code, and the connected domains meeting conditions in the intersection region (namely, the identified positioning patch) are output, so that the positioning patch can be identified quickly and efficiently by the technical scheme of the application, and the speed for identifying the positioning patch is improved while the false detection rate of the maximum connected domain is reduced.

Drawings

FIG. 1 is a flow chart illustrating a method for identifying a site provided by an exemplary embodiment of the present application.

FIG. 2 is a schematic structural view of a site provided in accordance with an exemplary embodiment of the present application.

Fig. 3 is a flow chart illustrating a method for identifying a site provided by another exemplary embodiment of the present application.

Fig. 4 is a flowchart illustrating a method for identifying a site provided by another exemplary embodiment of the present application.

Fig. 5 is a flowchart illustrating a method for identifying a site provided by yet another exemplary embodiment of the present application.

FIG. 6 is a schematic diagram of an apparatus for identifying a site provided in an exemplary embodiment of the present application.

FIG. 7 is a block diagram of an electronic device for identifying a site provided by an exemplary embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

Currently, when assisting a doctor in performing a puncture operation, a positioning patch needs to be placed on the body surface of a human body for positioning the relative position of the body surface of the human body and a medical image (such as computed tomography). Generally to obtain a particular location of a site, two approaches are used.

The first method is as follows: the mode of coarse positioning of the positioning patch is adopted. First, the blurred position (i.e., the position of the coarse position) of the site is found in the medical image of the human body. Then, the next stage is entered according to the position of the rough positioning, and the precise positioning of the positioning patch is carried out in the trained model. The second method comprises the following steps: the method mainly adopts a deep learning segmentation method to directly segment the positioning patch, and then selects a plurality of largest connected domains as the positioning patch to be output.

However, when the two methods are adopted to identify the positioning patches, the density of the positioning patches is close to the density of the segmentation scene, so that the segmentation result of the positioning patches is inaccurate, the extraction false detection rate of the maximum connected domain is high, the identification process is complicated, and the like.

In order to solve the above technical problem, the present application proposes a method of identifying a site as follows.

FIG. 1 is a flow chart illustrating a method for identifying a site provided by an exemplary embodiment of the present application. The method of fig. 1 is performed by a computing device, e.g., a server. As shown in FIG. 1, the method of identifying a site includes the following.

110: the medical image is input into a segmentation model trained in advance, and a localizer segmentation region and a localizer code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code are determined.

In particular, referring to fig. 2, the site can include a site area 201, a signal generator area 202, and a connection line area 203. The signal generator region 202 may be shaped like a Chinese character 'ren', is located inside the spacer region 201, and is connected to the connection line 203. The area 201 of the positioning sheet may be circular or square, and the embodiment of the present application does not specifically limit the specific structure of the positioning sheet.

The server can input the medical image input by the user into the pre-trained segmentation model to obtain the positioning sheet segmentation area, the signal generator segmentation area and the connecting line segmentation area. And then the server or the manual work encodes the positioning sheet partition area, the signal generator partition area and the connecting line partition area to obtain the positioning sheet code, the signal generator code and the connecting line code. The locating piece codes, the signal generator codes and the connecting line codes are different from one another and are unique codes, namely one region type corresponds to one code.

For example, the topogram is coded 1, the signal generator is coded 2, and the connecting line is coded 4.

It should be noted that the medical image may be a Computed Tomography (CT) image. The medical image may also be a Magnetic Resonance Imaging (MRI) image, and the specific type of the medical image is not particularly limited in the embodiments of the present application.

Preferably, the embodiment of the present application adopts a CT image with a fast scanning time and a clear image as the medical image.

It should be noted that, the segmentation and classification of the site patches are performed by using a segmentation model, and the segmentation model needs to output binary results of four channels. Four of which are independently separate channels. And the four channels respectively carry out prediction segmentation results and binary classification results of three categories of a positioning sheet, a signal generator and a connecting line.

It should be noted that, because the parameter quantity of the segmentation model is large, the prediction data (i.e. the medical image) can be used for block prediction during prediction, and because the object to be detected (i.e. the site in the medical image) is easy to be cut into different blocks during block cutting, the detection effect is poor, and therefore the step size of the segmentation model fusion is smaller than the step size of the blocks.

120: and determining at least one intersection code corresponding to the at least one intersection region based on the locating piece code, the signal generator code and the connecting line code.

Specifically, the server may add the spacer code, the signal generator code, and the connection line code two by two, and add the spacer code, the signal generator code, and the connection line code to determine at least one intersection code corresponding to at least one intersection region. And the intersection code corresponding to at least one intersection region is unique.

In one embodiment, the signal generator code, the spacer code, and the connecting line code are added to determine a first intersection region and a first intersection code. And adding the signal generator code and the locating piece code to determine a second intersection area and a second intersection code. And adding the signal generator code and the connecting line code to determine a third intersection area and a third intersection code. And adding the locating piece codes and the connecting line codes to determine a fourth intersection area and a fourth intersection code.

For example, the topogram code 1, the signal generator code 2 and the connecting line code 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.

130: and sequentially screening a plurality of connected domains in at least one intersection region according to a preset sequence so as to identify the positioning patches included in the medical image.

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

Specifically, since a signal generator region having a high HU (Hounsfield unit) exists inside the site, and the material of the signal generator region is different from that of the outside site region, the site can be identified by the segmentation model. Therefore, the preset sequence can be that the intersection region containing the signal generator segmentation region is sequenced at the forefront, the second time containing the positioning sheet segmentation region, and the last time containing the connecting line segmentation region.

In an embodiment, the preset order may be a first intersection region, a second intersection region, a third intersection region, and a fourth intersection region in sequence.

The server may sequentially determine the largest connected domain in each intersection region one by one according to a preset sequence. Wherein each intersection region may be any one of a first intersection region, a second intersection region, a third intersection region, and a fourth intersection region, each intersection region may include a plurality of connected domains, and the plurality of connected domains includes a largest connected domain. When the signal generator split area is included in the maximum connected domain, the server may treat the maximum connected domain as a site. When the maximum connected domain does not include the signal generator partition region, the server may determine whether the maximum connected domain in the next intersection region is the site according to a preset sequence.

For example, when the server detects that the first intersection region does not include a patch, the server may detect the largest connected domain in the second intersection region according to a preset order, and determine whether the second intersection region includes a patch.

The number of the site can be one or more, and the number of the site is not particularly limited in the embodiments of the present application.

In one embodiment, when the number of site tiles is two or more, the server can determine whether the number of currently determined site tiles reaches a preset number. When the number of the currently determined site patches does not reach the preset number, the server may determine whether a site patch exists in the next intersection area according to the preset sequence. The server may stop detecting the intersection region when the number of currently determined site patches reaches a preset number.

Therefore, the intersection region is encoded (namely, the intersection encoding) and the connected domains in the intersection region are screened according to the preset sequence determined by the intersection encoding, and the connected domains meeting the conditions (namely, the identified positioning patches) in the intersection region are output, so that the positioning patches can be identified quickly and efficiently by the technical scheme of the application, and the speed of identifying the positioning patches is improved while the false detection rate of the maximum connected domain is reduced.

Fig. 3 is a flow chart illustrating a method for identifying a site provided by another exemplary embodiment of the present application. The embodiment of fig. 3 is an example of the embodiment of fig. 1, and the same parts are not repeated herein, and the differences are mainly described herein. As shown in FIG. 3, the method of identifying a site includes the following.

310: and determining the maximum connected domains in each intersection region one by one according to a preset sequence, wherein the plurality of connected domains comprise the maximum connected domains.

Specifically, the server may determine the largest connected domain in each intersection region one by one according to a preset order. Wherein each intersection region may be any one of a first intersection region, a second intersection region, a third intersection region, and a fourth intersection region. Each intersection region may include a plurality of connected domains. The plurality of connected domains may include a largest connected domain.

320: when the signal generator division area is included in the maximum connected component area, the maximum connected component area is used as the positioning patch.

Specifically, when the server detects that the signal generator division area is included in the maximum connected domain, the server may use the maximum connected domain including the signal generator division area as the site.

It should be noted that the server may also delete the connected domain within a preset distance from the largest connected domain, filter out the connected domain whose distance is less than the preset distance, and only reserve the larger connected domain in the intersection region. The preset distance may be 1000 pixels, and the preset distance is not specifically limited in this embodiment of the application.

Therefore, the maximum connected domains in the intersection region are screened one by one, and the maximum connected domains including the signal generator segmentation region are used as the positioning patches, so that the false detection rate of the maximum connected domains is reduced.

Fig. 4 is a flowchart illustrating a method for identifying a site provided by another exemplary embodiment of the present application. The embodiment of fig. 4 is an example of the embodiment of fig. 1, and the same parts are not repeated herein, and the differences are mainly described here. As shown in FIG. 4, the method of identifying a site includes the following.

410: and judging whether the number of the positioning patches reaches a preset number.

Specifically, the number of the site can be two or more, that is, the number of the site can be plural, and the number of the site is not particularly limited in the embodiments of the present application. The server can detect the number of identified site patches, i.e., the server can determine whether the number of currently identified site patches reaches a preset number.

420: and when the number of the positioning patches does not reach the preset number, determining whether the positioning patches exist in the next intersection area or not according to the preset sequence.

Specifically, when the number of identified patches does not reach the preset number, the server may determine whether a next intersection area has a patch according to a preset order. The next intersection region may be a subsequent intersection region of the current intersection region determined according to the preset order. For example, if the current intersection region is the first intersection region, the next intersection region is the second intersection region according to the preset sequence.

In another embodiment, the server may stop detecting the intersection region when the number of identified site patches reaches a preset number.

It should be noted that when multiple patches are identified according to the predetermined sequence, the server may filter out duplicate patches that are closer together (e.g., patches that are less than 1000 pixels apart). If the number of the positioning patches more than the preset number still exist after filtering, the server can select the preset number of connected domains before the area ranking in the positioning patches as the positioning patches identified this time.

Preferably, the preset number is 3.

Therefore, the positioning patches are screened through the preset sequence, so that the positioning patches with the specified quantity can be screened efficiently and reasonably.

In an embodiment of the present application, determining at least one intersection code corresponding to at least one intersection region based on the topogram code, the signal generator code, and the connection line code includes: determining a first intersection region and a first intersection code based on the signal generator code, the spacer code and the connecting line code; determining a second intersection region and a second intersection code based on the signal generator code and the spacer code; determining a third intersection region and a third intersection code based on the signal generator code and the connecting line code; and determining a fourth intersection region and a fourth intersection code based on the locating piece code and the connecting line code, wherein the preset sequence is a first intersection region, a second intersection region, a third intersection region and the fourth intersection region in sequence.

Specifically, the server may add the signal generator code, the spacer code, and the connection line code to determine a first intersection region and a first intersection code. The first intersection region is the intersection of the locating piece dividing region, the signal generator dividing region and the connecting line dividing region. The first intersection code is the sum of the signal generator code, the spacer code and the connecting line code.

The server may add the signal generator code and the spacer code to determine a second intersection region and a second intersection code. And the second intersection region is the intersection of the signal generator segmentation region and the locating plate segmentation region. The second intersection code is the sum of the signal generator code and the localizer code.

The server may add both the signal generator code and the connection line code to determine a third intersection region and a third intersection code. And the third intersection area is the intersection of the signal generator segmentation area and the connecting line segmentation area. The third intersection code is the sum of the signal generator code and the connecting line code.

The server may add the topogram code and the connection line code to determine a fourth intersection region and a fourth intersection code. And the fourth intersection area is the intersection of the positioning sheet segmentation area and the connecting line segmentation area. The fourth intersection area is the sum of the spacer code and the connecting line code. And the first intersection code, the second intersection code, the third intersection code and the fourth intersection code are unique codes and are different from each other.

For example, the topogram 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.

Therefore, by setting the preset screening sequence, the embodiment of the application can efficiently and reasonably screen out the specified number (for example, one or more) of the positioning patches subsequently.

Fig. 5 is a flowchart illustrating a method for identifying a site provided by yet another exemplary embodiment of the present application. The embodiment of fig. 5 is an example of the embodiment of fig. 1, and the same parts are not repeated herein, and the differences are mainly described herein. As shown in FIG. 5, the method of identifying a site includes the following.

510: and cutting the medical image with the label information input into the initial network model by at least one cutting mode to obtain the region blocks.

In one embodiment, the medical image with the label information may include a labeling area, the labeling area includes a positioning sheet labeling area, a signal generator labeling area and a connecting line labeling area, and the signal generator labeling area is located in the positioning sheet labeling area.

Since the number of patches having tag information is small, and it is necessary to perform enhancement training by attaching the patches to which the tag information is attached to other data (for example, the body surface of a human body), it is necessary to cut out the patches first.

Specifically, the initial network model may crop the medical image with the label information input by the user through at least one cropping mode to obtain the region blocks. The cutting method may be 2 or 3, and the like, and this is not particularly limited in the embodiment of the present application. The area blocks may be obtained based on at least one cutting mode, and the size of the area blocks cut by different cutting modes may be the same or different, and the size of the area blocks is not specifically limited in the embodiment of the present application.

Preferably, the embodiment of the application adopts 3 cutting modes for cutting.

In an embodiment, the three cropping modes may crop the medical image according to a preset ratio.

The medical image may be labeled with label information. The label information can be labeled by a machine or manually, which is not limited in this embodiment. The region marked with the label information on the medical image can be used as a marking region. That is, the labeling area may include a navigation chip labeling area, a signal generator labeling area, and a connection line labeling area. Also, referring to the site structure of FIG. 2, the site labeling area may include a signal generator labeling area.

520: and filling the region cut blocks into the body surface of the human body to form training data.

Specifically, the initial network model may fill the region patches obtained based on the at least one clipping manner to a plurality of locations on the body surface of the human body to generate training data.

In one embodiment, due to the different sizes of the cut pieces, the cut pieces may exceed the boundary of the human body during training. Therefore, the initial network model can delete the partial region blocks which exceed the boundary of the human body so as to avoid various boundary overruns.

530: and training the initial network model according to the training data to obtain a segmentation model.

In particular, the server may apply the training data to train the initial network model to obtain the desired segmentation model.

540: the medical image is input into a segmentation model trained in advance, and a localizer segmentation region and a localizer code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code are determined.

550: and determining at least one intersection code corresponding to the at least one intersection region based on the locating piece code, the signal generator code and the connecting line code.

560: and sequentially screening a plurality of connected domains in at least one intersection region according to a preset sequence so as to identify the positioning patches included in the medical image.

Therefore, the training data of the positioning patch under various conditions can be generated by filling the cut region blocks in each part of the body surface of the human body, and the learning of the positioning patch by the model is enriched. And the subsequent positioning patch sorting and screening method can solve the problem that the density of the positioning patch is close to that of the scene to be segmented, and can solve the problem of high false detection rate of extraction of the maximum connected domain.

In an embodiment of the present application, the cropping, performed by at least one cropping method, the medical image with label information input into the initial network model to obtain the region tile includes: and cutting the medical image with the label information according to at least one cutting mode in a preset proportion to obtain the region blocks.

Specifically, the server may crop the medical image according to a preset ratio by using at least one cropping mode to obtain the region blocks. The sizes of the area blocks corresponding to different cutting modes may be the same or different, and this is not specifically limited in this embodiment of the application.

In one embodiment, the number of clipping modes is 3. The three cutting modes are respectively a first cutting mode for cutting by using the center of the labeling area, a second cutting mode for cutting by using the center of the labeling area of the signal generator and a third cutting mode for cutting by using 2 times of the distance from the center of the labeling area of the signal generator to the origin of the preset coordinate axis. The preset coordinate axis may be determined according to a Left (LPS) coordinate axis of the medical image. The preset coordinate axis origin may be the origin of the LPS coordinate system of the signal generator marking region. The preset ratio may be 5 for the first clipping mode, 3 for the second clipping mode, and 2 for the third clipping mode.

Therefore, the random sampling is carried out in proportion by adopting the three cutting strategies in the embodiment of the application, so that the initial network model can learn the whole and local information of the positioning patch, and the problem that the whole and local information cannot be considered in the learning of the model by adopting a single cutting mode is also avoided.

In an embodiment of the application, the at least one cutting mode includes a first cutting mode in which cutting is performed by using a center of the labeled area, a second cutting mode in which cutting is performed by using the center of the labeled area of the signal generator, and a third cutting mode in which cutting is performed by using 2 times of a distance from the center of the labeled area of the signal generator to an origin of a preset coordinate axis, where a preset ratio is that a ratio of the first cutting mode is 5, a ratio of the second cutting mode is 3, and a ratio of the third cutting mode is 2.

Specifically, the first clipping manner may be to clip the preset range with the center of the labeled area as a midpoint. The first clipping method can enable the initial network model to better learn the overall structure of the site, and can obtain better connecting line segmentation areas. However, the connecting line of the positioning patch is long, so that the area cut is large, and further, the labeling area of the signal generator in part of the positioning patch is cut off due to the fact that the area cut is too long in the process of subsequently pasting the signal generator on the body surface to construct training data, and the learning effect of the model on the connecting line labeling area is further influenced.

The second cutting method may be to cut the preset range by using the center of the labeled area of the signal generator as a midpoint, so as to obtain the area cut block. The second cutting mode enables the initial network model to better learn the signal generator labeling area, but can weakly learn the connecting line labeling area.

The third cutting mode can be cutting by 2 times of the distance from the midpoint of the signal generator labeling area to the origin of the preset coordinate axis, so as to obtain an area cutting block. The preset coordinate axis may be determined according to a Left (LPS) coordinate axis of the medical image. The preset coordinate axis origin may be the origin of the LPS coordinate system of the signal generator marking region. The third cutting mode can enable the initial network model to better learn the signal generator labeling area, but the area cutting block is smaller, so that the learning of the initial network model to the connecting line labeling area is less, and the trained segmentation model cannot better segment the connecting line segmentation area.

In an embodiment, the preset ratio may be that the first clipping mode ratio is 5, the second clipping mode ratio is 3, and the third clipping mode ratio is 2.

It should be noted that, because different clipping manners each have advantages and disadvantages, in the embodiment of the present application, three clipping strategies are used to perform clipping in a random sampling manner, so that the initial network model learning situation is richer.

Therefore, the three types of cutting are adopted for cutting according to the proportion in the embodiment of the application, so that the model can learn the whole and local information of the positioning patch, and the segmentation precision of the model is improved.

In an embodiment of the present application, the method further includes: the HU value of the region in the positioning sheet labeling region except the signal generator labeling region is increased or decreased to preset a numerical value, so that the initial network model can learn the connecting line region and the positioning sheet except the signal generator region.

Specifically, the server may adopt a random HU value dithering mode for the regions of the internal spacer labeling region of the spacer other than the signal generator labeling region, so that the initial network model further learns the HU values of the regions of the internal spacer labeling region of the spacer other than the signal generator labeling region.

Initial network model can be with the HU value heightening of the region in the spacer mark region except that signal generator marks the region and predetermine the numerical value to initial network model learns in the region of connecting wire and spacer except that signal generator region. Or, initial network model can reduce the HU value in the region except that signal generator mark region in the spacer mark region and predetermine the numerical value to initial network model learns in connecting wire region and spacer except that signal generator region. The preset value of the adjustment is not particularly limited in the embodiment of the present application.

For example, when the HU value of the region other than the signal generator labeling region in the positioning sheet labeling region is 100, the initial network model may adjust the HU value from 100 to 90 and perform learning.

It should be noted that the site densities vary because the training data for the site is typically less and there are more than one site. And the site density in the test set is quite unknown relative to the experimental body membrane. Therefore, if the model is too sensitive to HU values, the segmentation of the segmented model will not work well, making it difficult to identify new site patches. In addition, the segmentation effect of the segmentation model also varies greatly with the fluctuation of the HU values inside the site, so that the segmentation model is not robust enough.

Therefore, the embodiment of the application improves the model segmentation precision by adopting the mode of random HU value dithering, and enables the model to have stronger robustness.

In an embodiment of the present application, the region is cut and filled into the body surface of the human body to form training data, including: and filling the region blocks into a plurality of positions on the body surface of the human body, and deleting partial region blocks exceeding the region blocks on the body surface of the human body to acquire training data.

In particular, the server may fill the region patches into multiple locations on the body surface so that the initial network model can learn the situation of the site at any location on the body surface. Meanwhile, the server can delete partial region blocks in the region blocks exceeding the body surface of the human body so as to obtain training data and avoid various boundary overruns.

It should be noted that, the deletion of the partial region cut beyond the boundary of the body surface of the human body can be understood as a normalized operation for obtaining the training data.

Therefore, the training data is formed by filling the region blocks into the body surface of the human body, and a foundation is provided for the subsequent training of the initial network model.

Fig. 6 is a schematic diagram of an apparatus 600 for identifying a site provided in an exemplary embodiment of the present application. As shown in FIG. 6, the device 600 for identifying a site 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.

The first determining module 640 is configured to input the medical image into a pre-trained segmentation model, and determine a spacer segmentation region and a spacer code, a signal generator segmentation region and a signal generator code, and a connection line segmentation region and a connection line code; the second determining module 650 is configured to determine at least one intersection code corresponding to the at least one intersection region based on the topogram code, the signal generator code, and the connection line code; the identifying module 660 is configured to sequentially filter the plurality of connected components in the at least one intersection region according to a preset sequence to identify the site included in the medical image, where the preset sequence is determined based on the at least one intersection code.

The embodiment of the application provides a device for identifying a positioning patch, which encodes an intersection region (namely, the intersection encoding), screens a connected domain in the intersection region according to a preset sequence determined by the intersection encoding, and outputs the connected domain (namely, the identified positioning patch) meeting conditions in the intersection region, so that the positioning patch can be identified quickly and efficiently by the technical scheme of the application, and the speed for identifying the positioning patch is improved while the false detection rate of the maximum connected domain is reduced.

According to an embodiment of the present application, the identifying module 660 is configured to determine the largest connected domain in each intersection region one by one according to a preset sequence, where the plurality of connected domains include the largest connected domain; when the signal generator division area is included in the maximum connected component area, the maximum connected component area is used as the positioning patch.

According to an embodiment of the application, when the number of the site patches is two or more, the identification module 660 is configured to determine whether the number of the site patches reaches a preset number; and when the number of the positioning patches does not reach the preset number, determining whether the positioning patches exist in the next intersection area or not according to the preset sequence.

According to an embodiment of the present application, the second determining module 650 is configured to determine the first intersection region and the first intersection code based on the signal generator code, the spacer code, and the connecting line code; determining a second intersection region and a second intersection code based on the signal generator code and the spacer code; determining a third intersection region and a third intersection code based on the signal generator code and the connecting line code; and determining a fourth intersection region and a fourth intersection code based on the locating piece code and the connecting line code, wherein the preset sequence is a first intersection region, a second intersection region, a third intersection region and the fourth intersection region in sequence.

According to an embodiment of the present application, the first obtaining module 610 is configured to cut a medical image with label information input into an initial network model by at least one cutting method to obtain a region cut block, where the medical image with label information includes a labeling region, the labeling region includes a spacer labeling region, a signal generator labeling region and a connecting line labeling region, and the signal generator labeling region is located in the spacer labeling region; the filling module 620 is used for filling the region cut blocks into the body surface of the human body to form training data; the second obtaining module 630 is configured to train the initial network model according to the training data to obtain a segmentation model.

According to an embodiment of the present application, the first obtaining module 610 is configured to cut the medical image with the label information according to a preset ratio by using at least one cutting mode, so as to obtain the region cut block.

According to an embodiment of the application, the at least one cutting mode includes a first cutting mode for cutting by using the center of the labeled area, a second cutting mode for cutting by using the center of the labeled area of the signal generator, and a third cutting mode for cutting by using 2 times of the distance from the center of the labeled area of the signal generator to the origin of the preset coordinate axis, wherein the preset ratio is that the ratio of the first cutting mode is 5, the ratio of the second cutting mode is 3, and the ratio of the third cutting mode is 2.

According to an embodiment of the present application, the first obtaining module 610 is configured to increase or decrease the HU value of the region in the spacer labeling region except for the signal generator labeling region to preset a value, so that the initial network model learns the connection line region and the spacer except for the signal generator region.

According to an embodiment of the present application, the filling module 620 is configured to fill the region slices into a plurality of positions on the body surface of the human body, and delete partial region slices that exceed the region slices on the body surface of the human body, so as to obtain the training data.

It should be understood that, for specific working processes and functions of the first obtaining module 610, the filling module 620, the second obtaining module 630, the first determining module 640, the second determining module 650, and the identifying module 660 in the foregoing embodiments, reference may be made to the description in the method for identifying a site provided in the foregoing embodiments of fig. 1 to 5, and details are not repeated here in order to avoid repetition.

Fig. 7 is a block diagram of an electronic device 700 for identifying a site provided by an exemplary embodiment of the present application.

Referring to fig. 7, electronic device 700 includes a processing component 710 that further includes one or more processors, and memory resources, represented by memory 720, for storing instructions, such as applications, that are executable by processing component 710. The application programs stored in memory 720 may include one or more modules that each correspond to a set of instructions. Further, the processing component 710 is configured to execute instructions to perform the above-described method of identifying a site.

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. May operate based on an operating system stored in memory 720Electronic device 700, such as Windows Server^TM，Mac OS X^TM，Unix^TM，Linux^TM，FreeBSD^TMOr the like.

A non-transitory computer readable storage medium having instructions therein which, when executed by a processor of the electronic device 700, enable the electronic device 700 to perform a method of identifying a site, comprising: inputting the medical image into a pre-trained segmentation model, and determining a positioning sheet segmentation region and a positioning sheet code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code; determining at least one intersection code corresponding to the at least one intersection region based on the locating piece code, the signal generator code and the connecting line code; and sequentially screening a plurality of connected domains in the at least one intersection region according to a preset sequence to identify the positioning patches included in the medical image, wherein the preset sequence is determined based on the at least one intersection code.

All the above optional technical solutions can be combined arbitrarily to form optional embodiments of the present application, and are not described herein again.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program check codes, such as a U disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk.

It should be noted that, in the description of the present application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. In addition, in the description of the present application, "a plurality" means two or more unless otherwise specified.

The above description is only exemplary of the present application and should not be taken as limiting the present application, as any modifications, equivalents and the like that are within the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A method of identifying a site, comprising:

inputting the medical image into a pre-trained segmentation model, and determining a positioning sheet segmentation region and a positioning sheet code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code;

determining at least one intersection code corresponding to at least one intersection region based on the locating piece code, the signal generator code and the connecting line code;

sequentially screening a plurality of connected domains in the at least one intersection region according to a preset sequence to identify the site included in the medical image, wherein the preset sequence is determined based on the at least one intersection code.

2. The method of claim 1, wherein the sequentially filtering the plurality of connected components in the at least one intersection region according to a predetermined order to identify the site included in the medical image comprises:

determining the largest connected domain in each intersection region one by one according to the preset sequence, wherein the plurality of connected domains comprise the largest connected domain;

when the signal generator division area is included in the maximum connected component area, the maximum connected component area is taken as the positioning patch.

3. The method of claim 1, wherein when the number of site is two or more, the method further comprises:

judging whether the number of the positioning patches reaches a preset number or not;

and when the number of the positioning patches does not reach the preset number, determining whether the positioning patches exist in the next intersection area or not according to the preset sequence.

4. The method of claim 1, wherein determining at least one intersection code corresponding to at least one intersection region based on the topogram code, the signal generator code, and the connection line code comprises:

determining a first intersection region and a first intersection code based on the signal generator code, the spacer code, and the connecting line code;

determining a second intersection region and a second intersection encoding based on the signal generator encoding and the localizer encoding;

determining a third intersection region and a third intersection code based on the signal generator code and the connecting line code;

determining a fourth intersection region and a fourth intersection encoding based on the spacer encoding and the connecting line encoding,

wherein the preset sequence is sequentially the first intersection region, the second intersection region, the third intersection region and the fourth intersection region.

5. The method according to any one of claims 1 to 4, wherein before the inputting the medical image into the pre-trained segmentation model, determining the spacer segmentation region and the spacer encoding, the signal generator segmentation region and the signal generator encoding, and the connecting line segmentation region and the connecting line encoding, further comprises:

cutting a medical image with label information input into an initial network model in at least one cutting mode to obtain a region cut block, wherein the medical image with label information comprises a labeling region, the labeling region comprises a positioning sheet labeling region, a signal generator labeling region and a connecting line labeling region, and the signal generator labeling region is located in the positioning sheet labeling region;

filling the region cut blocks into the body surface of a human body to form training data;

and training the initial network model according to the training data to obtain the segmentation model.

6. The method according to claim 5, wherein the cropping the medical image with the label information input into the initial network model by at least one cropping method to obtain the region blocks comprises:

and cutting the medical image with the label information according to the at least one cutting mode in a preset proportion to obtain the region cut block.

7. The method of claim 6, wherein the at least one cropping mode comprises a first cropping mode cropping at the center of the labeled region, a second cropping mode cropping at the center of the signal generator labeled region, and a third cropping mode cropping at 2 times the distance from the center of the signal generator labeled region to the origin of the predetermined coordinate axis,

the preset proportion is that the first cutting mode proportion is 5, the second cutting mode proportion is 3, and the third cutting mode proportion is 2.

8. The method of claim 5, further comprising:

will except in the spacer mark region the HU value in the region outside the signal generator mark region is increaseed, or is increaseed and predetermine the numerical value, so that initial network model learns in the region of connecting wire and spacer except that signal generator region.

9. The method of claim 5, wherein said filling said region cut into a body surface to form training data comprises:

and filling the region blocks into a plurality of positions of the human body surface, and deleting partial region blocks exceeding the region blocks of the human body surface to obtain the training data.

10. An apparatus for identifying a site, comprising:

the first determining module is used for inputting the medical image into a pre-trained segmentation model, and determining a positioning sheet segmentation region and a positioning sheet code, a signal generator segmentation region and a signal generator code, and a connecting line segmentation region and a connecting line code;

the second determination module is used for determining at least one intersection code corresponding to at least one intersection region based on the locating piece code, the signal generator code and the connecting line code;

an identifying module, configured to sequentially screen a plurality of connected domains in the at least one intersection region according to a preset order to identify the site included in the medical image, where the preset order is determined based on the at least one intersection code.

11. A computer-readable storage medium, wherein the storage medium stores a computer program for executing the method of identifying a site of any of claims 1-9.

12. An electronic device, comprising:

a processor;

a memory for storing the processor-executable instructions,

wherein the processor is configured to perform the method of identifying a site of any of claims 1-9.