Method, device and nonvolatile storage medium for identifying split integrity

Abstract

The application discloses a kind of intercleft integrity identification method, device and nonvolatile storage medium.Therein, the method comprises: using target intercleft identification model to process the image features contained in the scan image of target physiological tissue, obtain first intercleft identification result, wherein, first intercleft identification result includes the first intercleft image region corresponding to the intercleft of target physiological tissue in scan image;Image segmentation is carried out to scan image, and the internal boundary identification result of target physiological tissue is determined according to the segmentation image obtained after segmentation;According to first intercleft identification result and internal boundary identification result, determine the integrity index of the intercleft of target physiological tissue.The application solves the technical problem that the accuracy of integrity judgment result is low caused by the experience of doctor to determine the integrity of the intercleft of target physiological tissue in the related art.

Description

Technical Field

[0001] This application relates to the field of medical technology, and more specifically, to a method, apparatus, and non-volatile storage medium for identifying interstitial crack integrity. Background Technology

[0002] Currently, before performing surgery on a patient's physiological tissues (such as various organs), it is necessary to identify the interstitial fissures in the patient's tissues and assess their integrity to determine whether a specific surgery on the patient's lungs is feasible. However, in related technologies, the method commonly used for identifying and determining the integrity of interstitial fissures in patients is based on the doctor's experience in judging the scan images of physiological tissues, resulting in a low accuracy rate in assessing the integrity of interstitial fissures.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a method, apparatus, and non-volatile storage medium for identifying the integrity of interstitial fissures, in order to at least solve the technical problem of low accuracy in integrity judgment results caused by relying on doctors' experience to determine the integrity of interstitial fissures in target physiological tissues in related technologies.

[0005] According to one aspect of the embodiments of this application, a method for identifying the integrity of interstitial fissures is provided, comprising: processing image features contained in a scanned image of a target physiological tissue using a target interstitial fissure identification model to obtain a first interstitial fissure identification result, wherein the first interstitial fissure identification result includes a first interstitial fissure image region in the scanned image corresponding to an interstitial fissure in the target physiological tissue; performing image segmentation on the scanned image and determining an internal boundary identification result of the target physiological tissue based on the segmented image obtained after segmentation, wherein the internal boundary identification result includes a boundary image region in the segmented image corresponding to each internal boundary of the target physiological tissue, and the internal boundary is the boundary between each part of the target physiological tissue; determining an integrity index of the interstitial fissures of the target physiological tissue based on the first interstitial fissure identification result and the internal boundary identification result, wherein the integrity index is used to reflect the health status of the target physiological tissue.

[0006] Optionally, the segmented image includes image regions corresponding to various parts of the target physiological tissue, and the voxels in the same image region have the same label value, while the voxels in different image regions have different label values; the step of determining the internal boundary recognition result of the target physiological tissue based on the segmented image obtained after segmentation includes: determining the internal boundary recognition result based on the label value of each voxel in the segmented image.

[0007] Optionally, the step of determining the internal boundary recognition result of the target physiological tissue based on the label value of each voxel in the segmented image includes: determining the label value of the voxels in the neighborhood corresponding to each voxel; determining the boundary image region in the segmented image based on the label value of the voxels in the neighborhood corresponding to each voxel, wherein, in the neighborhood corresponding to any voxel in the boundary image region, there exists a voxel with a label value different from any voxel.

[0008] Optionally, each boundary image region corresponds to a different interfission category; the step of determining the integrity index of the target physiological tissue based on the first interfission identification result and the internal boundary identification result includes: determining the target distance field corresponding to any boundary image region, wherein the distance between any voxel in the target distance field and any boundary image region is not greater than a preset distance; based on the target distance field corresponding to each boundary image region, filtering the voxels contained in the first interfission image region to obtain interfission voxels, and determining the interfission category corresponding to any interfission voxel based on the distance between any interfission voxel and each boundary image region; determining the second interfission identification result based on the interfission category corresponding to the interfission voxel, wherein the second interfission identification result includes multiple second interfission image regions, and the multiple second interfission image regions correspond one-to-one with multiple interfission categories; and determining the integrity index of the interfission of the target physiological tissue based on the second interfission identification result and the boundary identification result.

[0009] Optionally, the integrity index of the interfissure of the target physiological tissue includes the integrity index of the interfissure in each of the multiple interfissure categories; the step of determining the integrity index of the target physiological tissue based on the second interfissure identification result and the internal boundary identification result includes: determining the second interfissure image region corresponding to any interfissure category in the second interfissure identification result; comparing the second interfissure image region and the boundary image region corresponding to the same interfissure category respectively to determine the integrity index of the interfissure corresponding to each interfissure category.

[0010] Optionally, the step of comparing the second interstic image region and the boundary image region corresponding to the same interstic category to determine the interstic integrity index for each interstic category includes: determining the number of first voxels in each second interstic image region; determining the number of second voxels in the corresponding boundary image region; and determining the ratio of the number of first voxels to the number of second voxels as the interstic integrity index.

[0011] Optionally, the step of comparing the second interstitial image region and the boundary image region corresponding to the same interstitial category to determine the interstitial integrity index for each interstitial category includes: determining the first volume of each second interstitial image region; determining the second volume of the corresponding boundary image region; and determining the ratio of the first volume to the second volume as the interstitial integrity index.

[0012] Optionally, the inter-target split recognition model is trained in the following way: A training dataset is obtained, which includes multiple training images, comprising a first type of training image and a second type of training image. The first type of training image consists of scan images corresponding to target physiological tissues with healthy health conditions, and the second type of training image consists of scan images corresponding to target physiological tissues with abnormal health conditions. The HU values ​​in each training image are normalized and divided into multiple sub-training images with a preset resolution. The inter-target split recognition model is then used to process the sub-training images, and the model parameters of the inter-target split recognition model are adjusted based on the processing results and a preset loss function to obtain the inter-target split recognition model. The preset loss function includes a weighted cross loss function, a Dice loss function, and an edge loss function.

[0013] Optionally, the target split recognition model includes an encoding layer, a decoding layer, and an attention module. The encoding layer includes multiple encoding modules, the decoding layer includes multiple decoding modules, and the encoding modules and their corresponding decoding modules are connected through the attention module. The attention module is used to improve the image fusion weight of the region of interest in the image features output by the encoding module, and input the image features adjusted based on the image fusion weight into the decoding module.

[0014] According to another aspect of the embodiments of this application, a lung interfissure integrity recognition system is also provided, including an image acquisition device and a processor. The image acquisition device is used to scan the lungs of a target patient to obtain a scanned image of the lungs of the target patient. The processor is used to process the image features contained in the scanned image using a target interfissure recognition model to obtain a first lung interfissure recognition result, wherein the first lung interfissure recognition result includes a first lung interfissure image region in the scanned image corresponding to the lung interfissure. The processor segments the scanned image and determines the lung lobe boundary recognition result based on the segmented image obtained after segmentation, wherein the lung lobe boundary recognition result includes a boundary image region in the segmented image corresponding to the boundary of each lung lobe, and the lung lobe boundary is the boundary between different lung lobes in the lung. The processor determines the lung interfissure integrity index based on the lung interfissure recognition result and the internal boundary recognition result, wherein the interfissure integrity index is used to reflect the health status of the lungs.

[0015] According to another aspect of the embodiments of this application, a fissure integrity identification device is also provided, comprising: a first processing module, configured to process the image features contained in a scanned image of a target physiological tissue using a target fissure identification model to obtain a first fissure identification result, wherein the first fissure identification result includes a first fissure image region in the scanned image corresponding to the fissure of the target physiological tissue; a second processing module, configured to perform image segmentation on the scanned image and determine the internal boundary identification result of the target physiological tissue based on the segmented image obtained after segmentation, wherein the internal boundary identification result includes boundary image regions in the segmented image corresponding to each internal boundary of the target physiological tissue, and the internal boundary is the boundary between each part of the target physiological tissue; and a third processing module, configured to determine the fissure integrity index of the target physiological tissue based on the first fissure identification result and the internal boundary identification result, wherein the integrity index is used to reflect the health status of the target physiological tissue.

[0016] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, wherein a program is stored in the non-volatile storage medium, wherein the program controls the device where the non-volatile storage medium is located to execute the inter-crack integrity identification method when it runs.

[0017] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, the processor being configured to run a program stored in the memory, wherein a method for identifying inter-crack integrity is executed when the program is running.

[0018] In this embodiment, a target fissure recognition model is used to extract and fuse multiple image features of different resolutions from the scanned image of the target physiological tissue to obtain a first fissure recognition result. The first fissure recognition result includes a first fissure image region in the scanned image corresponding to the fissure of the target physiological tissue. The scanned image is then segmented to obtain a segmented image. Each segmented image includes image regions corresponding to different parts of the target physiological tissue, with voxels within the same image region having the same label value, while different image regions have different label values. Based on the label values ​​of each voxel in the segmented image, the internal boundary recognition result of the target physiological tissue is determined. This internal boundary recognition result includes the boundary in the segmented image corresponding to the internal boundary of the target physiological tissue. The image region has internal boundaries that define the boundaries between different parts of the target physiological tissue. Based on the first fissure identification result and the internal boundary identification result, the integrity index of the target physiological tissue is determined. This integrity index reflects the health status of the target physiological tissue. The first fissure identification result is obtained using a target fissure identification model, and the internal boundary identification result is obtained using image segmentation. This achieves the goal of determining the integrity index of the target physiological tissue fissure through the first fissure identification result and the internal boundary result, thus realizing the technical effect of automatically determining the integrity index of the target physiological tissue fissure. This solves the technical problem of low accuracy in integrity judgment results caused by relying on doctors' experience to determine the integrity of the target physiological tissue fissure in related technologies. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0020] Figure 1 This is a hardware structure block diagram of a computer terminal (or mobile device) according to an embodiment of this application;

[0021] Figure 2 This is a flowchart illustrating a method for identifying interstitial cracks according to an embodiment of this application;

[0022] Figure 3 This is a schematic diagram of an interfissure in the lung according to an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of a scanned image and a scanned image after window width and window level adjustment according to an embodiment of this application;

[0024] Figure 5 This is a schematic diagram of the structure of a target crack identification model according to an embodiment of this application;

[0025] Figure 6This is a schematic diagram of the structure of a residual module according to an embodiment of this application;

[0026] Figure 7 This is a schematic diagram of the structure of an attention module according to an embodiment of this application;

[0027] Figure 8 This is a schematic diagram of a segmented image according to an embodiment of this application;

[0028] Figure 9 This is a schematic diagram of a six-field embodiment according to an embodiment of this application;

[0029] Figure 10 This is a schematic diagram of a lung lobe boundary according to an embodiment of this application;

[0030] Figure 11 This is a schematic diagram of a distance field according to an embodiment of this application;

[0031] Figure 12 This is a schematic diagram of an interstitial image region according to an embodiment of this application;

[0032] Figure 13 This is a schematic diagram of a second interstitial crack identification result according to an embodiment of this application;

[0033] Figure 14 This is a flowchart illustrating an integrity assessment process according to an embodiment of this application;

[0034] Figure 15 This is a schematic diagram of the structure of a crack integrity identification device according to an embodiment of this application;

[0035] Figure 16 This is a schematic diagram of the structure of a lung fissure integrity identification system according to an embodiment of this application. Detailed Implementation

[0036] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0037] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0038] Currently, before performing surgery on a patient's target physiological tissue, it is necessary to identify the interstitial fissures within the tissue and assess their integrity to determine whether the specific surgery can be performed. However, in related technologies, the common method for identifying and determining the integrity of interstitial fissures involves physicians relying on experience to judge the scan images of the target physiological tissue. For example, physicians manually mark the interstitial fissures on the scan images of the target physiological tissue. Since the scan images of the target physiological tissue are three-dimensional and contain many images, marking the interstitial fissures on the scan images is very time-consuming for physicians. Furthermore, due to individual differences and instrument drift, the interstitial fissures are often not clearly visible on CT images, making it difficult to guarantee the accuracy and consistency of the results.

[0039] It should be noted that the aforementioned target physiological tissues may include organs such as the liver, lungs, and heart. By dividing these organs into multiple parts, and based on the boundaries between these parts and the characteristics of each part, the health status of the organs can be more accurately determined, thus facilitating subsequent organ-specific biopsies or surgeries. For ease of explanation, the lungs are used as an example in the embodiments of this application to further illustrate the solution provided in this application. It is understood that the method for analyzing the interstitial fissure integrity of the lungs provided in the embodiments of this application is also applicable to other organs, and will not be elaborated upon here.

[0040] Related technologies also provide some interpulmonary fissure segmentation methods based on anatomical knowledge or based on grayscale and shape information. However, these methods have poor recognition performance when there are lesions in the lungs or the interpulmonary fissures are unclear, resulting in the recognition results being unusable for analyzing the integrity of the interpulmonary fissures.

[0041] To address the aforementioned issues, this application provides relevant solutions, which are detailed below.

[0042] According to an embodiment of this application, a method embodiment for identifying interstitial crack integrity is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] The methods and embodiments provided in this application can be executed on mobile terminals, computer terminals, or similar computing devices. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a method for identifying interstitial crack integrity is shown. Figure 1 As shown, the computer terminal 10 (or mobile device 10) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0044] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0045] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the inter-split integrity identification method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the inter-split integrity identification method of the application described above. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0046] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0047] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0048] Under the above operating environment, embodiments of this application provide a method for identifying interstitial crack integrity, such as... Figure 2 As shown, the method includes the following steps:

[0049] Step S202: The image features contained in the scanned image of the target physiological tissue are processed by the target interfission recognition model to obtain the first interfission recognition result, wherein the first interfission recognition result includes the first interfission image region in the scanned image corresponding to the interfission of the target physiological tissue;

[0050] In some embodiments of this application, such as Figure 3 As shown, the target physiological tissue includes the lungs, and the interfissures within the target physiological tissue include the left oblique fissure, the right oblique fissure, and the right horizontal fissure. From Figure 3As can be seen from the diagram, the lungs include the right upper lobe, right middle lobe, right lower lobe, left upper lobe, and left lower lobe. The septum between the right upper lobe and the right middle lobe is the right horizontal fissure, the septum between the right middle lobe and the right lower lobe is the right oblique fissure, and the septum between the left upper lobe and the left lower lobe is the left oblique fissure.

[0051] In the technical solution provided in step S202, the target split recognition model can be trained in the following way: A training dataset is obtained, wherein the training dataset includes multiple training images, including a first type of training image and a second type of training image. The first type of training image is a scan image corresponding to a target physiological tissue with a healthy health condition, and the second type of training image is a scan image corresponding to a target physiological tissue with an abnormal health condition. Both the first and second type of training images are labeled with the sample split image region corresponding to the split of the target physiological tissue contained therein; the HU value in each training image is normalized and split into multiple pre-... A sub-training image with a specified resolution is used. The sub-training image is processed using the target split recognition model to be trained, and a processing result is obtained. The processing result includes the test split image region corresponding to the split contained in the sub-training image, which is output by the target split recognition model to be trained. The model parameters of the target split recognition model to be trained are adjusted according to the processing result and the preset loss function until the processing result and the value of the preset loss function meet the preset requirements, and the target split recognition model is obtained. The preset loss function includes the weighted cross loss function, the Dice loss function and the edge loss function. The preset requirements can be set according to actual needs, and there is no limitation on this in the embodiments of this application.

[0052] In some embodiments of this application, the target interstitial fissure recognition model can adopt a 3D Unet architecture, with an attention module added to this architecture. Specifically, the target interstitial fissure recognition model includes an encoding layer, a decoding layer, and an attention module. The encoding layer includes multiple encoding modules, and the decoding layer includes multiple decoding modules. The encoding modules and their corresponding decoding modules are connected through the attention module. The attention module is used to increase the image fusion weight of the region of interest in the image features output by the encoding module, and input the image features adjusted based on the image fusion weight to the decoding module. Thus, a scanned image of the target physiological tissue is input into the encoding layer, where the encoding module extracts image features of different dimensions from the scanned image. The encoding module can input the extracted image features into its connected attention module, which increases the image fusion weight of the region of interest in the input image features. The image features input into the attention module are then processed using the image fusion weight, and the processed image features are input into the decoding module connected to the attention module. The target interstitial fissure recognition model can obtain the corresponding first interstitial fissure recognition result based on the image features output by its last encoding module.

[0053] As an optional implementation, to reduce the learning difficulty, it is necessary to ensure that there is overlap between image blocks during cropping. Specifically, for any one of the multiple sub-training images, there exists a sub-training image in the multiple sub-training images with an overlap degree not less than a preset overlap threshold, which can be set by the user.

[0054] By preprocessing the training dataset as described above, the complete training images are cropped into image patches of the same size, and there is information overlap between the image patches. This not only reduces the memory usage of the training device's graphics card and lowers the configuration requirements, but also reduces the training difficulty of the model and improves training efficiency. Specifically, before inputting the training images into the target split recognition model to be trained, in order to reduce the learning difficulty, the training images can be preprocessed, including window width and window level adjustment and image slicing. Window width and window level adjustment refers to processing the HU values ​​of each voxel in the training image. Specifically, a first HU value threshold and a second HU value threshold can be set, with the second HU value threshold being greater than the first HU value threshold. For example, the second HU value threshold is 100, and the first HU value threshold is -1000. The HU values ​​of regions in the training image with HU values ​​greater than the second HU value threshold or less than the first HU value threshold are set to 0. Then, the Min_Max method is used to normalize the HU values ​​of voxels in the remaining regions of the training image to 0-1. The final training image is shown below. Figure 4 As shown, where Figure 4 The image on the left is the original image to be trained. Figure 4 The image on the right is the training image after window width and window level adjustment. The Min_Max method mentioned above refers to determining the maximum and minimum HU values ​​in the remaining region, as well as the baseline difference between the maximum and minimum HU values, and then determining the HU value after normalization for any voxel as the ratio between the difference between the HU value before normalization and the minimum HU value and the aforementioned baseline difference.

[0055] To reduce the GPU memory usage on the device running the interpulmonary fissure recognition model, the normalized training images can be cropped into identically sized images, such as 96×160×160 pixels. During cropping, to ensure information overlap between different images, a second target image can be cropped from the training images according to a 50% overlap rule. Specifically, the 50% overlap rule means that the overlap between the second target image taken from the training images and the most recently cropped image is 50%. Furthermore, to ensure training efficiency, the final second target image used for training must have a ratio of interpulmonary fissure pixels to all interpulmonary fissure pixels in the training images greater than a preset ratio, such as 1 / 50. This increases the pixel ratio between interpulmonary fissures and non-interpulmonary fissures, promoting better learning of interpulmonary fissure features by the training model.

[0056] As an optional implementation, in order to improve the training effect, this application provides a loss function that includes a weighted cross loss function, a Dice loss function, and an edge loss function.

[0057] Specifically, since the interstitial fissures in the target physiological tissue are usually sheet-like, some embodiments of this application provide a loss function that includes a weighted cross-entropy loss function, a Dice loss function, and an edge loss function, as shown in the following formula:

[0058] ,

[0059] in It is an adjustable weighting coefficient that can be set by the user; for example, it can be set to... .

[0060] In the loss function provided in this application embodiment, since the cross-loss function portion of the loss function evaluates the class prediction of each voxel separately and then averages the loss of all voxels, in cases of class imbalance, the training process is easily dominated by the class with a large number of voxels. This means that even if the loss function decreases during training, the final model still cannot output accurate recognition results. To solve the above problem, this application embodiment provides a method of assigning different weights to the losses of positive and negative samples to balance the influence of classes with different voxel counts on the training results. Specifically, the weighted cross-loss function can be expressed as the following formula:

[0061] ,

[0062] in It is the gold standard, which is the pre-labeled sample image region mentioned above. This is the network prediction result, which is the test interstitial image region output by the interstitial fragmentation recognition model to be trained, as mentioned above. It is the weight, which can be set by yourself, for example, it can be set to 0.9.

[0063] It should be noted that the Dice loss function in the loss function provided in this application embodiment can represent the measure of overlap between two samples, with a Dice coefficient of 1 indicating complete overlap. The advantage of this loss function is that it does not need to consider the imbalance of positive and negative samples, but only focuses on the degree of overlap between the pre-labeled sample image region and the test interstic image region output by the target interstic recognition model to be trained. That is, the weighted cross-loss function is expressed as follows:

[0064]

[0065] in Representative set and Common elements between Representative set Number of elements in the middle Representative set The number of elements in the data.

[0066] The edge loss function is designed specifically for data edge features. It strengthens the influence of edge features in the target segmentation recognition model, allowing the model to pay more attention to edge features in the training data during training, thus enabling the trained model to output more accurate segmentation results. Specifically, the edge loss function is obtained through Sobel edge detection. and The edges are identified, and the distance between the edges is calculated. The smaller the distance, the closer the segmentation results are. That is, the edge loss function is expressed as: ,in Represents edge detection, This represents the mean square error.

[0067] To fully utilize the image features at different levels in the scanned images and improve the recognition results of interpulmonary fissures, this application embodiment improved the 3D Unet network architecture when constructing the physiological tissue recognition model, resulting in the following... Figure 5 The target inter-crack identification model is shown. From... Figure 5As can be seen, the overall structure of the target split recognition model includes an encoding layer (also called the encoding path) and a decoding layer (also called the decoding path), with the encoding and decoding paths connected via skip connections. Specifically, the encoding path includes multiple encoding modules (also called encoders) used to downsample the image input to the encoder, obtain the feature map of the input image, and input the obtained feature map into the next encoder or decoder; the decoding path includes multiple decoding modules (also called decoders) used to upsample the image input to the decoder, and fuse the upsampled input image with the feature map sent by the encoder to obtain the output image, which is then output to the next level decoder or directly output as the final recognition result. It can be seen that each encoder in the multiple encoders can obtain the feature map of the image at a specific resolution through downsampling, and the resolution of the image feature map corresponding to different encoders is different. Furthermore, from... Figure 5 As can be seen, after downsampling, the encoder sends the obtained feature map to the corresponding decoder via skip connections. The decoder then upsamples the image input to it before fusing it with the image input to the encoder. To improve training performance, such as... Figure 5 As shown, an attention module is added to the target split recognition model. The attention module is used to increase the weight of the region of interest (specifically the region corresponding to the split in this embodiment) in the feature map sent directly from the encoder during the image fusion process.

[0068] It should be noted that each of the multiple coding layers includes a first residual module, and each of the multiple decoding layers includes a second residual module, an attention module, and a feature concatenation module. In the target interspindle recognition model provided in this embodiment, the decoding layer can utilize the attention module to adjust the output features of the coding layer, providing more effective high-resolution edge features. Feature concatenation fuses the high-resolution edge features with the output features of the upsampling layer, compensating for the loss of some edge features during the downsampling process. The aforementioned target interspindle recognition model also includes a convolutional layer with a kernel size of 1×1×1, used to adjust the channels in the image output by the decoding path, and applying a 0-1 probability map (i.e., the probability value of belonging to the interspindle voxel) using a sigmoid function. Finally, threshold segmentation (a threshold greater than or equal to a preset threshold indicates an interspindle voxel, and a threshold less than the preset threshold indicates a non-interspindle voxel; the preset threshold can be 0.5 or other custom values ​​between 0 and 1) yields the interspindle segmentation result, which is the first interspindle recognition result.

[0069] from Figure 5As can be seen, in the encoding path, when an image is input to the encoding layer, the number of channels in the image increases, and the size of the resulting feature map changes. Furthermore, except for the lowest-level encoding layer, the remaining encoding layers, after downsampling the input image, send the resulting feature maps to the attention modules in the corresponding decoding layers. The attention modules then enhance the weights of the regions of interest in the feature maps during the image fusion stage. Different encoding layers are connected through max-pooling layers, meaning that the feature maps obtained after downsampling are max-pooled before being passed to the next encoding layer. Different decoding layers, as well as the lowest-level encoding layer and the decoding layer, are connected through upsampling modules. This means that the decoding layer upsamples the images passed from other decoding layers or the lowest-level encoding layer. The feature concatenation module in the decoding layer then concatenates the upsampled feature maps with the feature maps processed by the attention modules. Figure 5 In this context, H, W, and D represent the height, width, and depth of the image, respectively, while 1, 16, 32, 64, 128, and 256 represent the number of images.

[0070] The structures of the first residual module and the second residual module are as follows: Figure 6 As shown, it consists of two convolutional modules, each containing a convolutional layer (CONV), a batch normalization (BN) layer, and a ReLU (activation) layer. The result obtained after processing by the two convolutional modules is concatenated with the input to obtain the final output.

[0071] The attention module provided in the embodiments of this application is as follows: Figure 7 As shown, Figure 7 Input in It is the feature map of the encoding part, input This is the feature map of the decoding part. From Figure 7 As can be seen from this, the attention module will... and The operation is performed in parallel, specifically a convolution operation with a 1×1×1 kernel. pass We get B. pass To obtain A, thus making the number of channels in A and B the same, where and Both use 1×1×1 convolution kernels. Then, the A+B operation (the "+" operation adds the values ​​of the same voxel in A and B) is performed to obtain C. Since g and... The resolutions of A and B are not the same, so the resolutions of A and B are also different. Therefore, before processing g through Wg, g needs to be upsampled to make the resolutions of g and xl consistent, thus making the resolutions of A and B consistent. After obtaining C, a ReLU layer is needed to activate C and filter out noise in C to obtain D. Then, a 1×1×1 convolution operation is performed on D. The combination of 0 and Sigmoid results in a 0-1 probability map, which is the attention coefficient. This can also be called attention weight. Finally, attention coefficient. Multiply get The final result It is a feature map that highlights the salient features of key local regions while suppressing irrelevant regions, and can effectively increase the proportion of key regions during image fusion.

[0072] The attention module provided in this application embodiment can increase the weight of feature information of local regions of interest in the feature map during the image fusion stage, while suppressing the fusion weight of regions of no interest (specifically in this application embodiment, the region of interest is the interpulmonary fissure region, and the region of no interest is the non-interpulmonary fissure region).

[0073] Step S204: Perform image segmentation on the scanned image, and determine the internal boundary recognition result of the target physiological tissue based on the segmented image obtained after segmentation. The internal boundary recognition result includes the boundary image region in the segmented image corresponding to each internal boundary of the target physiological tissue. The internal boundary is the boundary voxel between each part of the target physiological tissue.

[0074] In the technical solution provided in step S204, when performing image segmentation on the scanned image, the 3DUNET architecture image segmentation model can be directly used to segment the scanned image, thereby obtaining the following... Figure 8 The segmented image shown. Figure 8 This is a schematic diagram of a cross-section of a segmented scanned image. Figure 8 As can be seen, the label values ​​in the image regions corresponding to different lung lobes are different, and the label values ​​of each voxel in the same image region are the same. Figure 8 The shades of gray in the image can represent different label values.

[0075] Step S206: Determine the integrity index of the interstitial fissure of the target physiological tissue based on the first interstitial fissure identification result and the internal boundary identification result. The integrity index is used to reflect the health status of the target physiological tissue in the voxel boundary image region.

[0076] In the technical solution provided in step S206, since different image regions have different label values, the voxels belonging to the boundary region can be determined based on the label values ​​of each voxel in the segmented image, thereby obtaining the boundary image region. Specifically, the label values ​​of voxels in the neighborhood corresponding to each voxel can be determined first; then, based on the label values ​​of voxels in the six-neighborhood corresponding to each voxel, the boundary image region is determined in the segmented image. In the boundary image region, in the six-neighborhood corresponding to any voxel, there exists a voxel with a label value different from any voxel. The neighborhood of a voxel refers to the spatial region composed of all voxels whose distance from that voxel is within a preset distance, such as the six-neighborhood composed of voxels whose distance from any of the aforementioned voxels is zero.

[0077] Figure 9 This is a schematic diagram of the six neighborhoods corresponding to each voxel. It can be seen that the voxels in the six neighborhoods are the voxels that are closest to each other in space, including the voxels that are closest in six directions: directly above, directly below, front, back, left, and right.

[0078] The final segmented lung lobe boundary, i.e., the boundary image region, is as follows: Figure 10 As shown, where Figure 10 The left side shows a schematic diagram of the lung lobe boundaries in the coronal interface of the lung. Figure 10 The image on the right is a schematic diagram of the lobar boundaries in a vertical interface of the lung. It can be seen that the lobar boundaries are sheet-like, and the label values ​​differ in different lobar boundary regions.

[0079] In some embodiments of this application, each boundary image region of an interfission corresponds to a different interfission category. The step of determining the integrity index of the target physiological tissue based on the first interfission identification result and the internal boundary identification result includes: determining the target distance field corresponding to any boundary image region, wherein the distance between any voxel in the target distance field and any boundary image region is not greater than a preset distance; based on the target distance field corresponding to each boundary image region, filtering the voxels contained in the first interfission image region to obtain interfission voxels, and determining the interfission category corresponding to the interfission voxels based on the distance between the interfission voxels and each boundary image region; determining the second interfission identification result based on the interfission category corresponding to the interfission voxels, wherein the second interfission identification result includes multiple second interfission image regions, and the multiple second interfission image regions correspond one-to-one with multiple interfission categories; and determining the integrity index of the interfission of the target physiological tissue based on the second interfission identification result and the boundary identification result.

[0080] Specifically, Figure 11 The distance field is obtained based on the image region corresponding to the inner boundary, where points on the same line are equidistant from the inner edge, and the thicker the line, the closer it is to the inner boundary. Figure 12This is a schematic diagram of the interstitial voxels determined based on the first interstitial voxel identification result. It can be seen that the interstitial voxels corresponding to each interstitial voxel have not yet been distinguished at this time. Figure 13 This is a schematic diagram showing the identification results of multiple second-interstitial cracks. Figure 13 It includes multiple different grayscale values ​​( Figure 13 The grayscale values ​​in the image represent label values, and each region corresponds to a type of split. It should be noted that... Figure 11 The distance field shown can also be used to denoise the first interstitial crack identification result.

[0081] As an optional implementation, the integrity index of the target physiological tissue includes interslit integrity indices for each of the multiple interslit categories, and from... Figure 10 As can be seen, the steps for determining the integrity index of the target physiological tissue based on the second interfiscation identification result and the internal boundary identification result include: determining the second interfiscation image region corresponding to any interfiscation category in the second interfiscation identification result; comparing the second interfiscation image region and the boundary image region corresponding to the same interfiscation category respectively to determine the integrity index corresponding to each interfiscation category.

[0082] In some embodiments of this application, the interstitial integrity index can be determined based on the number of voxels in each image region. The step of determining the interstitial integrity index of each second interstitial image region based on each second interstitial image region and its corresponding boundary image region includes: determining the second interstitial image region corresponding to any interstitial category in the second interstitial identification result; comparing the second interstitial image region corresponding to the same interstitial category with the boundary image region to determine the boundary image region of the interstitial integrity index corresponding to each interstitial category.

[0083] In some other embodiments of this application, the inter-crack integrity index can be determined based on the volume of each image region. The step of determining the inter-crack integrity index of each second inter-crack image region based on each second inter-crack image region and the corresponding boundary image region includes: determining the first volume of each second inter-crack image region; determining the second volume of the corresponding boundary image region; and determining the ratio of the first volume and the second volume as the inter-crack integrity index.

[0084] Specifically, the formula for calculating the interstitial crack integrity index is as follows:

[0085]

[0086] in A represents the interstitial fracture integrity index. i A represents the number or volume of voxels in the identified interstitial regions. CThis represents the number or volume of voxels in a theoretically complete split region (i.e., the boundary image region corresponding to the split region in the aforementioned internal boundary image region). Specifically, in A... i When representing the number of voxels in the second split image region This indicates the number of voxels in the corresponding boundary image region; in A i When representing the volume of the second interslit image region, This represents the volume of the corresponding boundary image region.

[0087] After obtaining the integrity index of each interfissure, the integrity index of each interfissure can be directly used as the integrity index of the target physiological tissue, or the interfissure integrity index with the lowest / highest value can be used as the integrity index of the target physiological tissue, or the average value of the interfissure integrity index can be used as the integrity index of the target physiological tissue, etc.

[0088] In summary, this application provides a method for identifying the integrity of a target physiological tissue (such as the lungs), and the overall identification process is as follows: Figure 14 As shown, it includes the following steps:

[0089] Step S1402: Acquire lung CT image data;

[0090] Step S1404: Use the interpulmonary fissure segmentation network to identify the lung CT image data and obtain the interpulmonary fissure identification result;

[0091] Step S1406: Use a lung lobe segmentation network to identify lung CT image data and obtain lung lobe boundary identification results;

[0092] Step S1408: Calculate the interpulmonary fissure integrity index based on the interpulmonary fissure identification results and the lung lobe boundary identification results, and assess lung integrity based on the interpulmonary fissure integrity index.

[0093] It should be noted that steps S1404 and S1406 can be performed simultaneously or at different times. The aforementioned interpulmonary fissure segmentation network is the target interpulmonary fissure recognition model.

[0094] By employing a target fissure recognition model, multiple image features of different resolutions are extracted and fused from scanned images of the target physiological tissue to obtain a first fissure recognition result. This first fissure recognition result includes the first fissure image region in the scanned image corresponding to the fissure of the target physiological tissue. The scanned image is then segmented to obtain a segmented image. This segmented image includes image regions corresponding to various parts of the target physiological tissue, with voxels within the same image region having the same label value, while different image regions have different label values. Based on the label values ​​of each voxel in the segmented image, the internal boundary recognition result of the target physiological tissue is determined. This internal boundary recognition result includes the boundary image region in the segmented image corresponding to the internal boundary of the target physiological tissue. The domain, whose internal boundary is the boundary between various parts of the target physiological tissue; the integrity index of the target physiological tissue is determined based on the first interfission identification result and the internal boundary identification result. The integrity index is used to reflect the health status of the target physiological tissue. The first interfission identification result is obtained by using the target interfission identification model, and the internal boundary identification result is obtained by using image segmentation. This achieves the goal of determining the integrity index of the interfission of the target physiological tissue through the first interfission identification result and the internal boundary result, thereby realizing the technical effect of automatically determining the integrity index of the interfission of the target physiological tissue. This solves the technical problem of low accuracy of integrity judgment results caused by relying on doctors' experience to determine the integrity of the interfission of the target physiological tissue in related technologies.

[0095] This application provides an interstitial crack integrity identification device. Figure 15 This is a schematic diagram of the interstitial crack integrity identification device, such as... Figure 15 As shown, the device includes: a first processing module 150, used to process the image features contained in the scanned image of the target physiological tissue using a target interstitial fissure recognition model to obtain a first interstitial fissure recognition result, wherein the first interstitial fissure recognition result includes a first interstitial fissure image region in the scanned image corresponding to the interstitial fissure of the target physiological tissue; a second processing module 152, used to perform image segmentation on the scanned image and determine the internal boundary recognition result of the target physiological tissue based on the segmented image obtained after segmentation, wherein the internal boundary recognition result includes boundary image regions in the segmented image corresponding to each internal boundary of the target physiological tissue, and the internal boundary is the boundary between each part of the target physiological tissue; and a third processing module 154, used to determine the integrity index of the interstitial fissure of the target physiological tissue based on the first interstitial fissure recognition result and the internal boundary recognition result, wherein the integrity index is used to reflect the health status of the target physiological tissue.

[0096] In some embodiments of this application, the target physiological tissue includes the lungs, and the interfissures of the target physiological tissue include the left oblique fissure, the right oblique fissure, and the right horizontal fissure.

[0097] In some embodiments of this application, the target splitting recognition model is trained in the following manner: A training dataset is obtained, comprising multiple training images, including a first type of training image and a second type of training image. The first type of training image consists of scan images corresponding to target physiological tissues with healthy health conditions, and the second type of training image consists of scan images corresponding to target physiological tissues with abnormal health conditions. The HU values ​​in each of the multiple training images are normalized and divided into multiple sub-training images with a preset resolution. The sub-training images are processed using the target splitting recognition model to be trained, and the model parameters of the target splitting recognition model to be trained are adjusted according to the processing results and a preset loss function to obtain the target splitting recognition model. The preset loss function includes a weighted cross loss function, a Dice loss function, and an edge loss function.

[0098] In some embodiments of this application, the target split recognition model includes an encoding layer, a decoding layer, and an attention module. The encoding layer includes multiple encoding modules, the decoding layer includes multiple decoding modules, and the encoding modules and their corresponding decoding modules are connected through an attention module. The attention module is used to increase the image fusion weight of the region of interest in the image features output by the encoding module, and input the image features with adjusted image fusion weights into the decoding module.

[0099] In some embodiments of this application, for any sub-training image among multiple sub-training images, there exists a sub-training image among the multiple sub-training images whose overlap with the arbitrary sub-training image is equal to a preset overlap threshold.

[0100] In some embodiments of this application, the segmented image includes image regions corresponding to various parts of the target physiological tissue, and the label values ​​of voxels within the same image region are the same, while the label values ​​corresponding to different image regions are different; the step of the second processing module 152 in determining the internal boundary recognition result of the target physiological tissue based on the segmented image obtained after segmentation includes: determining the internal boundary recognition result based on the label values ​​of each voxel in the segmented image.

[0101] In some embodiments of this application, the step of the second processing module 152 determining the internal boundary recognition result of the target physiological tissue based on the label value of each voxel in the segmented image includes: determining the label value of the voxel in the neighborhood corresponding to each voxel; determining the boundary image region in the segmented image based on the label value of the voxel in the neighborhood corresponding to each voxel, wherein, in the neighborhood corresponding to any voxel in the boundary image region, there exists a voxel with a label value different from any voxel.

[0102] In some embodiments of this application, each boundary image region corresponds to a different interslit category; the third processing module 154 determines the integrity index of the target physiological tissue based on the first interslit identification result and the internal boundary identification result, including: determining the target distance field corresponding to any boundary image region, wherein the distance between any voxel in the target distance field and any first boundary image region is not greater than a preset distance; filtering the voxels contained in the first interslit image region to obtain interslit voxels based on the target distance field corresponding to each boundary image region, and determining the interslit category corresponding to the interslit voxel based on the distance between the interslit voxel and each first boundary image region; determining the second interslit identification result based on the interslit category corresponding to the interslit voxel, wherein the second interslit identification result includes multiple second interslit image regions, and the multiple second interslit image regions correspond one-to-one with multiple interslit categories; and determining the integrity index of the target physiological tissue based on the second interslit identification result and the boundary identification result.

[0103] In some embodiments of this application, the integrity index of the target physiological tissue includes any one of the boundary image regions of each interfiscation category among multiple interfiscation categories; the third processing module 154 determines the integrity index of the target physiological tissue based on the second interfiscation identification result and the internal boundary identification result, including: determining the second interfiscation image region corresponding to any interfiscation category in the second interfiscation identification result; comparing the second interfiscation image region and the boundary image region corresponding to the same interfiscation category respectively to determine the interfiscation integrity index corresponding to each interfiscation category.

[0104] In some embodiments of this application, the third processing module 154 compares the second interstic image region and the boundary image region corresponding to the same interstic category to determine the interstic integrity index corresponding to each interstic category. The steps include: determining the number of first voxels in each second interstic image region; determining the number of second voxels in the corresponding boundary image region; and determining the ratio of the number of first voxels to the number of second voxels as the interstic integrity index.

[0105] In some embodiments of this application, the third processing module 154 compares the second interstitial image region and the boundary image region corresponding to the same interstitial category to determine the interstitial integrity index corresponding to each interstitial category. The steps include: determining the first volume of each second interstitial image region; determining the second volume of the corresponding boundary image region; and determining the ratio of the first volume to the second volume as the interstitial integrity index.

[0106] It should be noted that each module in the above-mentioned inter-crack integrity identification device can be a program module (for example, a set of program instructions that implement a certain function) or a hardware module. For the latter, it can be manifested in the following forms, but is not limited to them: each of the above modules is manifested as a processor, or the functions of each of the above modules are implemented by a processor.

[0107] According to an embodiment of this application, a system for identifying the integrity of interpulmonary fissures is provided. Figure 16 This is a schematic diagram of the system structure, such as... Figure 16 As shown, the system includes: an image acquisition device 160 and a processor 162. The image acquisition device 160 is used to scan the lungs of a target patient to obtain a scanned image of the lungs of the target patient. The processor 162 is used to process the image features contained in the scanned image using a target interlobar recognition model to obtain a first interlobar recognition result, wherein the first interlobar recognition result includes a first interlobar image region in the scanned image corresponding to the interlobar of the lung. The system also segments the scanned image and determines the lung lobe boundary recognition result based on the segmented image, wherein the lung lobe boundary recognition result includes a boundary image region in the segmented image corresponding to the lung lobe boundary, and the lung lobe boundary is the boundary between different lobes in the lung. Finally, the system determines the integrity index of the interlobar of the lung based on the first interlobar recognition result and the internal boundary recognition result, wherein the integrity index is used to reflect the health status of the lung.

[0108] According to an embodiment of this application, a diagnostic method is provided. The diagnostic method includes the following steps: processing the image features contained in a scanned image of a target physiological tissue using a target fissure recognition model to obtain a first fissure recognition result, wherein the first fissure recognition result includes a first fissure image region in the scanned image corresponding to the fissure of the target physiological tissue; segmenting the scanned image and determining the internal boundary recognition result of the target physiological tissue based on the segmented image, wherein the internal boundary recognition result includes boundary image regions in the segmented image corresponding to each internal boundary of the target physiological tissue, and the internal boundaries are the boundaries between different parts of the target physiological tissue; determining an integrity index of the target physiological tissue based on the first fissure recognition result and the internal boundary recognition result, wherein the integrity index reflects the health status of the target physiological tissue; and determining whether to perform a pre-set surgery on the target physiological tissue based on the integrity index.

[0109] According to an embodiment of this application, an embodiment of a computer program product is also provided. When the computer program product is run by an electronic device, it controls the electronic device to perform the following interstitial fissure integrity identification method: processing the image features contained in the scanned image of the target physiological tissue using a target interstitial fissure identification model to obtain a first interstitial fissure identification result, wherein the first interstitial fissure identification result includes a first interstitial fissure image region in the scanned image corresponding to the interstitial fissure of the target physiological tissue; performing image segmentation on the scanned image, and determining the internal boundary identification result of the target physiological tissue based on the segmented image obtained after segmentation, wherein the internal boundary identification result includes boundary image regions in the segmented image corresponding to each internal boundary of the target physiological tissue, and the internal boundary is the boundary between each part of the target physiological tissue; determining the integrity index of the target physiological tissue based on the first interstitial fissure identification result and the internal boundary identification result, wherein the integrity index is used to reflect the health status of the target physiological tissue.

[0110] According to an embodiment of this application, an electronic device is also provided, including: a memory and a processor. The processor is used to run a program stored in the memory, wherein, during program execution, the following interstitial fissure integrity identification method is performed: processing the image features contained in the scanned image of the target physiological tissue using a target interstitial fissure identification model to obtain a first interstitial fissure identification result, wherein the first interstitial fissure identification result includes a first interstitial fissure image region in the scanned image corresponding to the interstitial fissure of the target physiological tissue; performing image segmentation on the scanned image, and determining the internal boundary identification result of the target physiological tissue based on the segmented image obtained after segmentation, wherein the internal boundary identification result includes boundary image regions in the segmented image corresponding to each internal boundary of the target physiological tissue, and the internal boundary is the boundary between each part of the target physiological tissue; determining the integrity index of the target physiological tissue based on the first interstitial fissure identification result and the internal boundary identification result, wherein the integrity index is used to reflect the health status of the target physiological tissue.

[0111] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to related technologies, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0116] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for identifying interstitial crack integrity, characterized in that, include: The image features contained in the scanned image of the target physiological tissue are processed by the target interfiscation identification model to obtain the first interfiscation identification result, wherein the first interfiscation identification result includes the first interfiscation image region in the scanned image that corresponds to the interfiscation of the target physiological tissue; The scanned image is segmented, and the internal boundary recognition result of the target physiological tissue is determined based on the segmented image. The internal boundary recognition result includes the boundary image regions in the segmented image that correspond to each internal boundary of the target physiological tissue. The internal boundary is the boundary between each part of the target physiological tissue, and each boundary image region corresponds to a different interstitial category. Based on the first interstitial fissure identification result and the internal boundary identification result, an integrity index of the interstitial fissure of the target physiological tissue is determined, wherein the integrity index is used to reflect the health status of the target physiological tissue; Determining the integrity index of the interfiscation of the target physiological tissue based on the first interfiscation identification result and the internal boundary identification result includes: determining the target distance field corresponding to any boundary image region, wherein the distance between any voxel in the target distance field and any boundary image region is not greater than a preset distance; filtering the voxels contained in the first interfiscation image region to obtain interfiscation voxels based on the target distance field corresponding to each boundary image region, and determining the interfiscation category corresponding to any interfiscation voxel based on the distance between any interfiscation voxel and each boundary image region; determining the second interfiscation identification result based on the interfiscation category corresponding to the interfiscation voxel, wherein the second interfiscation identification result includes multiple second interfiscation image regions, and the multiple second interfiscation image regions correspond one-to-one with multiple interfiscation categories; and determining the integrity index of the interfiscation of the target physiological tissue based on the second interfiscation identification result and the boundary identification result.

2. The interstitial crack integrity identification method according to claim 1, characterized in that, The segmented image includes image regions corresponding to various parts of the target physiological tissue, and the voxels in the same image region have the same label value, while the voxels in different image regions have different label values. The step of determining the internal boundary recognition result of the target physiological tissue based on the segmented image obtained after segmentation includes: The internal boundary recognition result is determined based on the label values ​​of each voxel in the segmented image.

3. The interstitial crack integrity identification method according to claim 2, characterized in that, The step of determining the internal boundary recognition result of the target physiological tissue based on the label values ​​of each voxel in the segmented image includes: Determine the label value of the voxels in the neighborhood corresponding to each voxel; Based on the label values ​​of the voxels in the neighborhood corresponding to each voxel, the boundary image region is determined in the segmented image, wherein in the neighborhood corresponding to any voxel in the boundary image region, there exists a voxel with a label value different from that of any voxel.

4. The interstitial crack integrity identification method according to claim 1, characterized in that, The integrity index of the interstitial fissures of the target physiological tissue includes the integrity index of the interstitial fissures for each of the multiple interstitial fissure categories. The steps for determining the integrity index of the target physiological tissue based on the second interfission identification result and the internal boundary identification result include: Determine the second interstitial image region in the second interstitial fracture identification result that corresponds to any interstitial fracture category; The second interstitial image region and the boundary image region corresponding to the same interstitial fragment category are compared to determine the integrity index of the interstitial fragments corresponding to each interstitial fragment category.

5. The interstitial crack integrity identification method according to claim 4, characterized in that, The step of comparing the second interstitial image region and the boundary image region corresponding to the same interstitial fragment category to determine the integrity index of the interstitial fragments for each category includes: Determine the first number of voxels in each of the second interslit image regions; Determine the number of second voxels in the corresponding boundary image region; The ratio of the number of first voxels in the second split image region corresponding to the same split category to the number of second voxels in the boundary image region is determined, and the obtained ratio is used as the integrity index of the corresponding split.

6. The interstitial crack integrity identification method according to claim 4, characterized in that, The step of comparing the second interstitial image region and the boundary image region corresponding to the same interstitial fragment category to determine the integrity index of the interstitial fragments for each category includes: Determine the first volume of each of the second interslit image regions; Determine the second volume of the corresponding boundary image region; The ratio of the first volume of the second interstitial image region corresponding to the same interstitial fragment category to the second volume of the boundary image region is determined, and the obtained ratio is used as the integrity index of the corresponding interstitial fragment.

7. The interstitial crack integrity identification method according to claim 1, characterized in that, The target split identification model is trained in the following way: Obtain a training dataset, wherein the training dataset includes multiple training images, the training images include a first type of training images and a second type of training images, the first type of training images are scan images corresponding to the target physiological tissue in a healthy state, and the second type of training images are scan images corresponding to the target physiological tissue in an abnormal state. The HU value in each of the multiple training images is normalized and then split into multiple sub-training images with a preset resolution. The sub-training image is processed using the target split recognition model to be trained, and the model parameters of the target split recognition model to be trained are adjusted according to the processing results and the preset loss function to obtain the target split recognition model. The preset loss function includes a weighted cross loss function, a Dice loss function and an edge loss function.

8. The interstitial crack integrity identification method according to claim 1, characterized in that, The target split recognition model includes an encoding layer, a decoding layer, and an attention module, wherein... The encoding layer includes multiple encoding modules, the decoding layer includes multiple decoding modules, and the encoding modules and their corresponding decoding modules are connected through the attention module; The attention module is used to increase the image fusion weight of the region of interest in the image features output by the encoding module, and input the image features adjusted based on the image fusion weight into the decoding module.

9. A system for identifying the integrity of interpulmonary fissures, characterized in that, Includes image acquisition devices and processors, among which, The image acquisition device is used to scan the lungs of the target patient, thereby obtaining scan images of the lungs of the target patient; The processor is configured to process the image features contained in the scanned image using a target interlobar recognition model to obtain a first interlobar recognition result, wherein the first interlobar recognition result includes a first interlobar image region in the scanned image corresponding to the interlobar of the lung; segment the scanned image and determine the lobe boundary recognition result of the lung based on the segmented image, wherein the lobe boundary recognition result includes a boundary image region in the segmented image corresponding to the lobe boundary of the lung, the lobe boundary being the boundary between different lobes in the lung, and each boundary image region corresponding to a different interlobar category; determine the integrity index of the interlobar of the lung based on the first interlobar recognition result and the lobe boundary recognition result, wherein the integrity index is used to reflect the health status of the lung; and determine the integrity index of the interlobar of the lung based on the first interlobar recognition result and the lobe boundary recognition result. The determination of the interfissure integrity index of the target physiological tissue based on the internal boundary recognition result includes: determining the target distance field corresponding to any boundary image region, wherein the distance between any voxel in the target distance field and any boundary image region is not greater than a preset distance; filtering the voxels contained in the first interfissure image region to obtain interfissure voxels based on the target distance field corresponding to each boundary image region, and determining the interfissure category corresponding to any interfissure voxel based on the distance between any interfissure voxel and each boundary image region; determining the second interfissure recognition result based on the interfissure category corresponding to the interfissure voxel, wherein the second interfissure recognition result includes multiple second interfissure image regions, and the multiple second interfissure image regions correspond one-to-one with multiple interfissure categories; and determining the interfissure integrity index of the target physiological tissue based on the second interfissure recognition result and the boundary recognition result.

10. A device for identifying interstitial crack integrity, characterized in that, include: The first processing module is used to process the image features contained in the scanned image of the target physiological tissue using a target interstitial fissure recognition model to obtain a first interstitial fissure recognition result, wherein the first interstitial fissure recognition result includes a first interstitial fissure image region in the scanned image that corresponds to the interstitial fissure of the target physiological tissue; The second processing module is used to perform image segmentation on the scanned image and determine the internal boundary recognition result of the target physiological tissue based on the segmented image obtained after segmentation. The internal boundary recognition result includes the boundary image regions in the segmented image that correspond to each internal boundary of the target physiological tissue. The internal boundary is the boundary between each part of the target physiological tissue, and each boundary image region corresponds to a different interstitial category. The third processing module is used to determine the integrity index of the interstitial fissures of the target physiological tissue based on the first interstitial fissure identification result and the internal boundary identification result, wherein the integrity index is used to reflect the health status of the target physiological tissue. The third processing module is further configured to determine the target distance field corresponding to any boundary image region, wherein the distance between any voxel in the target distance field and any boundary image region is not greater than a preset distance; based on the target distance field corresponding to each boundary image region, the voxels contained in the first interslit image region are screened to obtain interslit voxels, and the interslit category corresponding to any interslit voxel is determined based on the distance between any interslit voxel and each boundary image region; a second interslit identification result is determined based on the interslit category corresponding to the interslit voxel, wherein the second interslit identification result includes multiple second interslit image regions, and the multiple second interslit image regions correspond one-to-one with multiple interslit categories; and the integrity index of the interslits of the target physiological tissue is determined based on the second interslit identification result and the boundary identification result.

11. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a program, wherein when the program is executed, it controls the device containing the non-volatile storage medium to execute the inter-crack integrity identification method according to any one of claims 1 to 8.

12. An electronic device, characterized in that, include: A memory and a processor, the processor being configured to run a program stored in the memory, wherein the program, when running, executes the inter-crack integrity identification method according to any one of claims 1 to 8.

Images