A standardized coordinate system based method and apparatus for registration of regions of interest

Abstract

The application provides a standardized coordinate system-based registration method and device for a region of interest, and relates to the technical field of medical imaging. The method comprises the following steps: acquiring chest images of a patient at different times; for each chest image, performing the following steps: determining a lung boundary box based on the chest image; establishing a standardized coordinate system with the center point of the lung boundary box as the origin; determining a target region of interest in the chest image; determining the relative coordinates of the target region of interest in the standardized coordinate system; and determining the target region of interest in other chest images that matches the target region of interest based on the relative coordinates of the target region of interest. The application can reduce the complexity of registration calculation.

Description

Technical Field

[0001] This application relates to the field of medical imaging technology, and in particular to a method and apparatus for region of interest registration based on a standardized coordinate system. Background Technology

[0002] In evaluating treatment efficacy, physicians need to compare images from different time periods to track changes in the volume, density, morphology, and other characteristics of the same region of interest in order to determine the treatment response. For example, in evaluating the efficacy of lung cancer treatment, chest CT images from the baseline period, the first follow-up examination after medication, and subsequent follow-up periods are compared to determine changes in the lesions. Therefore, registration of regions of interest in different images is necessary.

[0003] In existing technologies, rigid registration or affine transformation registration is usually used. However, these registration methods are whole-lung registrations, which have high computational complexity. Summary of the Invention

[0004] This application is made in view of at least one of the above-mentioned technical problems existing in the prior art, and the application can reduce the computational complexity of registration.

[0005] In a first aspect, embodiments of this application provide a region-of-interest registration method based on a standardized coordinate system, including:

[0006] Acquire chest images of patients at different times;

[0007] Perform the following for each of the aforementioned chest images:

[0008] Based on the chest images, the lung bounding box is determined;

[0009] A standardized coordinate system is established with the center point of the lung bounding box as the origin;

[0010] Identify the target region of interest in the chest image;

[0011] Determine the relative coordinates of the target region of interest in the normalized coordinate system;

[0012] Based on the relative coordinates of the target region of interest, other target regions of interest in chest images that match the target region of interest are determined.

[0013] Secondly, embodiments of this application provide a region-of-interest registration device based on a standardized coordinate system, comprising:

[0014] The acquisition module is configured to acquire chest images of patients at different times.

[0015] The registration module is configured to perform the following for each of the chest images: determining a lung bounding box based on the chest image; establishing a standardized coordinate system with the center point of the lung bounding box as the origin; determining a target region of interest in the chest image; determining the relative coordinates of the target region of interest in the standardized coordinate system; and determining target regions of interest in other chest images that match the target region of interest based on the relative coordinates of the target region of interest.

[0016] Thirdly, embodiments of this application provide an electronic device, one or more processors, and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform any of the methods described above.

[0017] Fourthly, embodiments of this application provide a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods described above.

[0018] This application provides a method and apparatus for region-of-interest (ROI) registration based on a standardized coordinate system. By establishing a standardized coordinate system, it eliminates the inconsistency of absolute coordinates caused by differences in image setup, respiratory phase, and scanning bed position across different time periods. Even if the absolute coordinates of the same target differ significantly in the original chest images, a unified benchmark can still be achieved through relative coordinates in the standardized coordinate system, ensuring comparability of positions across time phases. The lung bounding box is dynamically calculated based on the actual lung morphology of each chest image. The origin of the standardized coordinate system adaptively adjusts with lung deformation, ensuring that the relative coordinates of the target's ROI always reflect its anatomical relative position within the lung. This allows for precise adaptation to non-rigid deformation scenarios such as atelectasis and compensatory expansion, improving registration accuracy. The registration process does not require the complex calculations of rigid full-lung registration or deformable registration, nor does it require the large amount of training data for deep learning registration. The process is simple and computationally efficient. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating a region-of-interest registration method based on a standardized coordinate system, as provided in one embodiment of this application.

[0021] Figure 2 This is a flowchart of a region-of-interest registration method based on a standardized coordinate system, provided in another embodiment of this application;

[0022] Figure 3 This is a schematic diagram of a region of interest registration device based on a standardized coordinate system, provided in one embodiment of this application. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions of the embodiments of this application, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Existing registration methods may have the following problems:

[0025] Question 1: The coordinate systems of images from different periods are not consistent.

[0026] During baseline, initial follow-up, and subsequent scans, the origin and orientation of the coordinate system established for each scan differ due to factors such as positioning, respiratory phase, and changes in the scanning bed position. Even for the same lesion, its absolute coordinates will vary significantly in different images, making direct matching based on coordinate values ​​impossible.

[0027] Question 2: Changes in the relative position of lesions due to lung tissue deformation

[0028] The lungs are flexible organs, affected by respiratory movements, heartbeats, changes in mediastinal structure, atelectasis or compensatory expansion after tumor treatment, etc. The morphology, volume and internal anatomical structure distribution of the lung lobes will undergo non-rigid deformation at different times, and the relative position of lesions in the lung lobes will also change accordingly. Traditional coarse localization methods based on the percentage of lung apex to lung base or lung lobe division are not accurate enough.

[0029] Question 3: Ambiguity in one-to-one matching in scenarios with multiple lesions

[0030] Lung cancer patients often have multiple lesions, and during treatment, new lesions may appear, existing lesions may disappear, or lesions may merge or split. Matching based solely on the principle of closest spatial distance can easily lead to matching errors, such as mistaking two adjacent different lesions as the same lesion manifesting at different times, especially when lesions are densely distributed spatially or when their location changes after treatment.

[0031] Question 4: Inconsistent physical dimension calibration due to differences in scanning parameters

[0032] For example, the slice thickness, pixel spacing, and reconstruction matrix size of CT scans at different time phases may vary, resulting in inconsistent physical dimensions of voxels. Directly calculating the distance in the pixel coordinate system will introduce systematic errors and cannot accurately reflect the positional relationship of lesions in the real three-dimensional physical space.

[0033] Question 5: Limitations of existing registration methods

[0034] While methods based on rigid or affine registration of the whole lung can correct for positional differences, they cannot handle local non-rigid deformations of the lungs and are computationally time-consuming, making it difficult to meet the real-time needs of clinical practice.

[0035] Deformable registration methods, such as B-spline and Demons algorithm, can handle non-rigid deformations, but they are complex to tune, consume a lot of computational resources, have a high risk of registration failure, and have limited ability to track individuals with multiple lesions.

[0036] Deep learning-based registration methods require a large amount of paired training data and the model's generalization depends on the data distribution, resulting in insufficient robustness to rare lesion morphologies or extreme deformation scenarios.

[0037] In view of one or more of the above issues, such as Figure 1 As shown, this application provides a method for region-of-interest registration based on a standardized coordinate system, including:

[0038] Step 101: Obtain chest images of the patient at different times.

[0039] Chest images of patients are collected at different time points (such as the baseline period before treatment, the follow-up period after treatment, and the long-term follow-up period). The chest images can be of different types, such as chest CT (Computed Tomography) images and lung MRI (Magnetic Resonance Imaging). Subsequent examples will use CT as an example for illustration.

[0040] For example, obtaining chest CT images of lung cancer patients at different stages of treatment.

[0041] Baseline period: denoted as CT1, scan date as T1;

[0042] First follow-up examination period (1-2 months after medication): recorded as CT2, scan date as T2;

[0043] Follow-up period (3-6 months after medication): denoted as CT3, scan date as T3.

[0044] For each target scan date, images can be selected from multiple sequences (corresponding to plain scans, enhanced scans, or different reconstruction algorithms) existing in the image archiving and communication system. Chest images contain information such as voxel coordinates and voxel spacing.

[0045] Step 102: Perform the following for each chest image: Determine the lung bounding box based on the chest image.

[0046] The lung bounding box refers to the smallest rectangle (2D image) or cuboid (3D volume) that can completely enclose the left and right lung regions.

[0047] Step 103: Establish a standardized coordinate system with the center point of the lung bounding box as the origin.

[0048] Step 104: Identify the target region of interest in the chest image.

[0049] The region of interest can be a lesion, a special anatomical structure of the lung (such as bullae, bronchiectasis areas, fibrotic lesions, etc. that need to be tracked), or a target point after interventional lung treatment (such as ablation foci, particle implantation areas, etc.).

[0050] The target region of interest (ROI) is the region of interest that needs to be registered, such as a lesion located within a lung bounding box. A chest image may contain one or more ROIs; if multiple ROIs exist, subsequent steps are performed for each ROI.

[0051] Step 105: Determine the relative coordinates of the target region of interest in the standardized coordinate system.

[0052] Step 106: Based on the relative coordinates of the target region of interest, determine the target regions of interest in other chest images that match the target region of interest.

[0053] The target region of interest is usually determined by calculating Euclidean distance, overlap, etc.

[0054] This application's embodiments eliminate the inconsistency in absolute coordinates caused by differences in positioning, respiratory phase, and scanning bed position in images from different time periods by establishing a standardized coordinate system. Even if the absolute coordinates of the same target differ significantly in the original chest images, a unified benchmark can still be achieved through relative coordinates in the standardized coordinate system, ensuring comparability of positions across time phases. The lung bounding box is dynamically calculated based on the actual lung morphology of each chest image, and the origin of the standardized coordinate system adaptively adjusts with lung deformation, ensuring that the relative coordinates of the target's region of interest always reflect its anatomical relative position within the lung. This allows for precise adaptation to non-rigid deformation scenarios such as atelectasis and compensatory expansion, improving registration accuracy. The registration process does not require the complex calculations of rigid full-lung registration or deformable registration, nor does it require the large amount of training data for deep learning registration; the process is simple and computationally efficient.

[0055] In one embodiment of this application, determining the lung bounding box based on chest images includes:

[0056] Semantic segmentation is performed on chest images to generate lung masks;

[0057] The lung bounding box is determined based on the lung mask.

[0058] Lung tissue is accurately separated from the background (such as the thorax, muscles, air, mediastinum, and other non-lung structures) in chest images through semantic segmentation, generating a lung mask. The generated lung mask can be a 3D binary image of the same size as the original chest image, where the pixel value of the lung tissue region is 1, and the pixel value of the non-lung region is 0. Segmentation can be performed using a deep learning-based semantic segmentation model, such as U-Net, V-Net, or Transformer architecture.

[0059] Taking a three-dimensional lung image as an example, the three-dimensional pixel matrix of the entire lung mask is traversed, and the coordinates of all pixels with a value of 1 are selected. The extreme values ​​of these pixels on the Z-axis, Y-axis, and X-axis are determined. The extracted extreme coordinates are used as the boundaries to construct a three-dimensional lung bounding box, which is the smallest cuboid that can just accommodate all lung tissue.

[0060] Compared to traditional fixed anatomical landmark localization, the lung mask generated by semantic segmentation can accurately cover the entire lung tissue. The bounding box extracted based on the lung mask can completely encompass all lung lobes, avoiding the omission of lung areas or the inclusion of non-lung areas due to lung lobe deformation and body position differences, thus improving the localization accuracy of the bounding box. The geometric center of the lung bounding box is the origin of the standardized coordinate system, and the accuracy of the lung bounding box directly determines the reliability of the origin position. In the embodiments of this application, the bounding box obtained by semantic segmentation and mask extraction can stably reflect the anatomical center of the lung at the current stage, providing core support for subsequent elimination of body position differences and establishment of a unified coordinate benchmark across time phases.

[0061] In one embodiment of this application, semantic segmentation of a chest image is performed to generate a lung mask, including:

[0062] Based on a pre-trained 3D deep learning model, semantic segmentation is performed on chest images to generate segmentation masks for five lung lobes.

[0063] A lung mask is generated based on the segmentation mask of the five lung lobes.

[0064] This application employs a 3D deep learning model, such as 3D U-Net, pre-trained on a large amount of chest image data, which excels at capturing spatial contextual information. For the chest image, five independent lobes are accurately segmented: the left upper lobe, left lower lobe, right upper lobe, right middle lobe, and right lower lobe. Five segmentation masks of the same size as the original chest image are generated, with each segmentation mask corresponding to one lobe. Pixel values ​​in the lobe region are 1, and pixel values ​​in non-lobe regions are 0.

[0065] A logical OR operation is performed on the segmentation masks of the five lung lobes. That is, if a pixel is 1 in the segmentation mask of any lung lobe (belonging to a certain lung lobe), it is marked as 1 in the final lung mask. The generated lung mask is a three-dimensional binary image of the same size as the original chest image, which completely covers all areas of the five lung lobes, and the pixel value of non-lung areas is 0.

[0066] The pre-trained 3D deep learning model can accurately identify the anatomical structure of lung lobes. Even in complex scenarios such as lobar adhesions, localized atelectasis, and scanning artifacts, it can still stably segment all five lung lobes, significantly improving segmentation accuracy. The method of merging the segmented five lobes separately ensures both the segmentation precision of each lobe and the integrity of the whole lung mask, avoiding problems such as blurred lung lobe boundaries and missed regions caused by overall segmentation, and providing high-quality basic data for lung bounding box extraction.

[0067] In one embodiment of this application, the coordinate axes of the standardized coordinate system are the same as those of the whole-lung coordinate system of the chest image.

[0068] The coordinate axes of the whole lung coordinate system follow the original DICOM (Digital Imaging and Communications in Medicine) coordinate system definition of chest imaging, that is, the Z-axis points in the head-to-toe direction, the Y-axis points in the front-to-back direction, and the X-axis points in the left-to-right direction.

[0069] The coordinate axis orientation of the standardized coordinate system follows the orientation rules of the whole lung coordinate system without changing the orientation of any coordinate axis, and only takes the geometric center of the lung bounding box as the new origin.

[0070] Slight rotation may occur in the positioning of chest images at different time points, such as a slight lateral turn of the patient's body while lying down. However, the orientation of the whole-lung coordinate system remains consistent with the original DICOM coordinate system of the chest image. By adopting this orientation in the standardized coordinate system, even if the entire lung rotates, the relative offset of the region of interest along each coordinate axis remains stable, avoiding errors in relative coordinate calculations caused by inconsistencies in orientation. Unified coordinate axis orientation avoids the complex coordinate conversion problems caused by orientation rotation, ensuring that the relative coordinates Δz, Δy, and Δx of the lesion have clear and consistent anatomical significance.

[0071] In one embodiment of this application, determining a target region of interest in a chest image includes:

[0072] Obtain the mask of the region of interest in a chest image;

[0073] Create an empty mask of the same size as the chest image;

[0074] Fill the empty mask with the mask of the region of interest;

[0075] Based on the lung bounding box, a cropped and filled empty mask is used; the region of interest located within the lung bounding box after cropping is the target region of interest.

[0076] Chest images are automatically segmented using pre-trained deep learning lesion detection models (such as 3D Faster R-CNN and 3D U-Net), outputting a mask for the region of interest. This mask is a 3D binary image, with a pixel value of 1 for the region and a pixel value of 0 for the background. The deep learning lesion detection model can also output physical information such as the starting coordinates and voxel spacing of the region of interest in the chest image, providing a positional reference for subsequent filling into a full-size empty mask.

[0077] The created empty mask perfectly matches the original chest image in three dimensions, ensuring a complete mapping of the chest image's spatial extent. All pixels in the empty mask are initialized to 0, with no pre-defined region markers.

[0078] Based on the starting coordinates of the region of interest (ROI), its precise filling position in the empty mask is calculated, i.e., the global voxel coordinate index is determined. The region with a pixel value of 1 in the ROI mask is copied to the corresponding global coordinate position in the empty mask, ensuring that the filled region is completely consistent with the spatial position of the ROI in the original chest image. The filled empty mask is a 3D binary image of the same size as the chest image, marked with 1 only at the global coordinate position of the ROI, while the rest remains 0. The lung bounding box is used as the clipping range. This lung bounding box is the smallest 3D bounding box containing only lung tissue extracted based on the whole-lung mask. The filled empty mask is clipped according to the extreme coordinates of the lung bounding box, retaining only the pixel region within the bounding box. After clipping, the region with a pixel value of 1 in the mask is the target ROI located within the lung bounding box; if the ROI partially or completely exceeds the lung bounding box, the excess portion is discarded, retaining only the effective portion within the lung.

[0079] This embodiment of the application, based on lung bounding box cropping, retains only the region of interest within the lung, eliminating interference from non-lung regions to ensure the validity of the target region of interest and avoid errors caused by invalid regions participating in subsequent registration. The lung bounding box is dynamically adjusted according to the lung morphology of each image, and the cropped target region of interest can adapt to the non-rigid deformation of the lung, ensuring that even if the lung shrinks or expands, the target region remains a valid part within the lung, improving registration robustness. After cropping, invalid pixels in non-lung regions are eliminated, reducing the computational load of subsequent steps such as centroid calculation, coordinate transformation, and Euclidean distance matching, significantly improving computational efficiency.

[0080] In one embodiment of this application, determining the relative coordinates of the target region of interest in a standardized coordinate system includes:

[0081] Determine the pixel coordinates of the centroid and the voxel spacing of the target region of interest;

[0082] Calculate the physical coordinates of the centroid based on the pixel coordinates of the centroid and the voxel spacing;

[0083] Based on the physical coordinates of the centroid and the origin, the relative coordinates of the centroid of the target region of interest are determined.

[0084] By iterating through all pixels with a value of 1 in the mask obtained from the aforementioned cropping, their pixel indices on the Z, Y, and X axes are extracted. The mean of these three indices is then calculated to obtain the pixel coordinates of the centroid. Based on the pixel coordinates of the centroid and the voxel spacing extracted from the chest image DICOM file, the physical coordinates of the centroid are calculated. The pixel coordinates of the centroid are (z... v , y v , x v Given a voxel spacing of (dz, dy, dx), its physical coordinates (z) are... v ×dz, y v ×dy, x v ×dx). Relative coordinates P of the centroid norm = P - C i = (Δz, Δy, Δx), where, C i The physical coordinates of the origin.

[0085] This application's embodiments unify pixel size differences caused by different scanning parameters into true physical sizes by converting pixel coordinates to physical coordinates, ensuring the quantitative comparability of coordinates across time phases. By converting physical coordinates to relative coordinates, it eliminates absolute coordinate fluctuations caused by differences in patient position, scanning bed position, and respiratory phase, resolving the problem of significant absolute coordinate differences for the same lesion across different time phases. Relative coordinates can adapt to non-rigid lung deformation and positional changes, and even in complex scenarios such as lesion fusion, splitting, and slight migration, it can still accurately distinguish different lesions through relative positional differences, exhibiting superior robustness compared to traditional matching methods based on absolute coordinates.

[0086] In other words, the embodiments of this application can achieve dual standardization transformation:

[0087] Spatial standardization: Eliminating differences in CT scan parameters through physical coordinate transformation;

[0088] Positional standardization: Eliminates the effects of body position and lung deformation by using relative coordinates.

[0089] In one embodiment of this application, determining a target region of interest in other chest images that match the target region of interest based on the relative coordinates of the target region of interest includes:

[0090] Based on the relative coordinates of the centroids of the target region of interest in the current chest image and the target region of interest in the target chest image, calculate the Euclidean distance between the target region of interest in the current chest image and the target region of interest in the target chest image.

[0091] Under the condition that the Euclidean distance meets the preset matching conditions, the region of interest in the target chest image is determined to match the region of interest in the current chest image.

[0092] The Euclidean distance is calculated using a three-dimensional Euclidean distance formula. When there is only one region of interest (ROI) in the target chest image, the matching condition is that the Euclidean distance is less than a matching threshold, such as less than 100 mm. When there are at least two ROIs in the target chest image, the matching condition is that the Euclidean distance of the ROI that matches the current chest image is less than the Euclidean distance of any other ROI in the target chest image. If the Euclidean distance does not meet the preset matching condition, then there is no matching ROI.

[0093] In this embodiment, the target region of interest in the chest image closest to the current target region of interest can be selected first, and then it can be determined whether the minimum distance is less than the matching threshold. If the minimum distance is less than or equal to the matching threshold, the target region of interest in the chest image is determined to match the target region of interest in the current chest image (i.e., the cross-temporal manifestation of the same target); if the minimum distance is greater than the matching threshold, it is determined that there is no matching object (i.e., the current target has disappeared or the target chest image shows a new lesion).

[0094] In this application, for densely distributed scenarios such as multiple metastases in both lungs, the numerical differences in Euclidean distance can accurately distinguish the cross-temporal positional relationships of different lesions, avoiding matching errors based on spatial proximity. By transforming the positional matching of cross-temporal targets into objective physical distance calculation and rule-based judgment, it effectively avoids matching ambiguities involving multiple lesions, eliminates subjective errors and external interference, and ensures registration accuracy and reliability while also considering clinical real-time performance and adaptability.

[0095] like Figure 2 As shown, this application provides a method for region-of-interest registration based on a standardized coordinate system, including:

[0096] Step 201: Obtain chest images of the patient at different times.

[0097] Obtain CT images of the patient at baseline, during the first follow-up visit, and during the follow-up period.

[0098] Step 202: Perform semantic segmentation on each chest image to generate a lung mask.

[0099] Step 203: Determine the lung bounding box based on the lung mask.

[0100] Step 204: Establish a standardized coordinate system with the center point of the lung bounding box as the origin.

[0101] The center points of the lung bounding box in the three-phase CT images are (287.5, 180.0, 180.0), (285.0, 178.5, 178.5), and (289.2, 181.8, 181.8), respectively, in millimeters.

[0102] Step 205: Identify the target region of interest in the chest image.

[0103] CT scan revealed 5 lesions:

[0104] Lesion L 11 (Left upper lobe main lesion): Volume 28450 mm³

[0105] Lesion L 12 (Metastatic lesion in the upper lobe of the right lung): Volume 15230 mm³

[0106] Lesion L 13 (Micronodule in the lower lobe of the right lung): Volume 680 mm³

[0107] Lesion L 14 (Small nodule in the lower lobe of the left lung): Volume 520 mm³

[0108] Lesion L 15 (Micronodule in the middle lobe of the right lung): Volume 385 mm³

[0109] CT scan revealed 5 lesions:

[0110] Lesion L 21 Volume: 24120 mm³

[0111] Lesion L 22 Volume 12850 mm³

[0112] Lesion L 23 Volume 612 mm³

[0113] Lesion L 24 Volume 495 mm³

[0114] Lesion L 25 Volume 358 mm³

[0115] CT scan revealed 5 lesions:

[0116] Lesion L 31 Volume: 26780 mm³

[0117] Lesion L 32 Volume 14320 mm³

[0118] Lesion L 33 Volume 705 mm³

[0119] Lesion L 34 Volume 532 mm³

[0120] Lesion L 35 Volume 410 mm³

[0121] Step 206: Determine the pixel coordinates of the centroid of the target region of interest.

[0122] Step 207: Calculate the physical coordinates of the centroid based on its pixel coordinates.

[0123] Step 208: Based on the physical coordinates of the centroid and the physical coordinates of the origin, determine the relative coordinates of the centroid of the target region of interest.

[0124] With L 11For example, calculate the relative coordinates of its centroid and the relative coordinates of each lesion detected by CT2 and CT3.

[0125] Step 209: Based on the relative coordinates of the centroid of the target region of interest in the current chest image and the relative coordinates of the centroid of the target region of interest in the target chest image, calculate the Euclidean distance between the target region of interest in the current chest image and the target region of interest in the target chest image.

[0126] Step 210: Under the condition that the Euclidean distance meets the preset matching conditions, determine the target region of interest in the target chest image and match the target region of interest in the current chest image.

[0127] Calculate L 11 The Euclidean distance between the lesions detected by CT2 and each lesion is used. If the minimum Euclidean distance is less than the matching threshold, the corresponding lesion is considered to be within the range of L. 11 Matching lesions, the same applies to CT3.

[0128] Once the match is successful, the clinical comparison data can be output, as shown in Table 1.

[0129] Table 1

[0130]

[0131] like Figure 3 As shown, this application embodiment provides a region-of-interest registration device based on a standardized coordinate system, including:

[0132] The acquisition module 301 is configured to acquire chest images of patients at different times.

[0133] The registration module 302 is configured to perform the following for each chest image: based on the chest image, determine the lung bounding box; establish a standardized coordinate system with the center point of the lung bounding box as the origin; determine the target region of interest in the chest image; determine the relative coordinates of the target region of interest in the standardized coordinate system; and determine the target regions of interest in other chest images that match the target region of interest based on the relative coordinates of the target region of interest.

[0134] In one embodiment of this application, the registration module 302 is configured to perform semantic segmentation on a chest image to generate a lung mask; and to determine a lung bounding box based on the lung mask.

[0135] In one embodiment of this application, the registration module 302 is configured to perform semantic segmentation on a chest image based on a pre-trained three-dimensional deep learning model to generate a segmentation mask for five lung lobes; and to generate a lung mask based on the segmentation mask for the five lung lobes.

[0136] In one embodiment of this application, the registration module 302 is configured to: acquire a mask of the region of interest in a chest image; create an empty mask of the same size as the chest image; fill the empty mask with the mask of the region of interest; and crop the filled empty mask based on the lung bounding box; wherein the region of interest located within the lung bounding box after cropping is the target region of interest.

[0137] In one embodiment of this application, the registration module 302 is configured to determine the pixel coordinates and voxel spacing of the centroid of the target region of interest; calculate the physical coordinates of the centroid based on the pixel coordinates and voxel spacing of the centroid; and determine the relative coordinates of the centroid of the target region of interest based on the physical coordinates of the centroid and the physical coordinates of the origin.

[0138] In one embodiment of this application, the registration module 302 is configured to calculate the Euclidean distance between the target region of interest in the current chest image and the target region of interest in the target chest image based on the relative coordinates of the centroid of the target region of interest in the current chest image and the relative coordinates of the centroid of the target region of interest in the target chest image; and determine that the target region of interest in the target chest image matches the target region of interest in the current chest image if the Euclidean distance meets the preset matching conditions.

[0139] This application provides an electronic device, including:

[0140] One or more processors;

[0141] Storage device for storing one or more programs.

[0142] When one or more programs are executed by one or more processors, the one or more processors implement the methods as described in any of the above embodiments.

[0143] This application provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in any of the above embodiments.

[0144] This application provides a computer program product that, when executed by a processor, implements the method as described in any of the above embodiments.

[0145] It should be understood that in the embodiments of this application, the processor may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0146] It should also be understood that the memory mentioned in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (Read-Only Memory). Only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct memory bus RAM (DR RAM).

[0147] It should be noted that when the processor is a general-purpose processor, DSP, ASIC, FPGA, or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, the memory (storage module) is integrated into the processor.

[0148] It should be noted that the memories described herein are intended to include, but are not limited to, these and any other suitable types of memories.

[0149] In addition to the data bus, this bus may also include a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled "bus" in the diagram.

[0150] It should also be understood that the first, second, third, fourth and various numerical designations used herein are merely for descriptive convenience and are not intended to limit the scope of this application.

[0151] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0152] In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software. The steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, detailed descriptions are omitted here.

[0153] In the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0154] Those skilled in the art will recognize that the various illustrative logical blocks (ILBs) and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0155] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0156] 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 network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0157] In addition, 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.

[0158] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0159] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A region-of-interest registration method based on a standardized coordinate system, characterized in that, include: Obtain chest images of the same patient at different times; Perform the following for each of the aforementioned chest images: Based on the chest images, the lung bounding box is determined; A standardized coordinate system is established with the center point of the lung bounding box as the origin; Identify the target region of interest in the chest image; Determine the pixel coordinates of the centroid and the voxel spacing of the target region of interest; The physical coordinates of the centroid are calculated based on the pixel coordinates of the centroid and the voxel spacing. Based on the physical coordinates of the centroid and the physical coordinates of the origin, the relative coordinates of the centroid of the target region of interest are determined; Based on the relative coordinates of the target region of interest, and according to Euclidean distance and preset matching conditions, other target regions of interest in chest images that match the target region of interest are determined.

2. The method as described in claim 1, characterized in that, Based on the chest images, the lung bounding box is determined, including: Semantic segmentation is performed on the chest image to generate a lung mask; Based on the lung mask, the lung bounding box is determined.

3. The method as described in claim 2, characterized in that, Semantic segmentation of the chest image to generate a lung mask includes: Based on a pre-trained 3D deep learning model, semantic segmentation is performed on the chest image to generate segmentation masks for five lung lobes. The lung mask is generated based on the segmentation mask of the five lung lobes.

4. The method as described in claim 1, characterized in that, in, The coordinate axes of the standardized coordinate system are the same as those of the whole-lung coordinate system of the chest image.

5. The method as described in claim 1, characterized in that, Determining the target region of interest in the chest image includes: Obtain the mask of the region of interest in the chest image; Create a void mask of the same size as the chest image; Fill the empty mask with the mask of the region of interest; Based on the lung bounding box, a cropped and filled empty mask is used; wherein, the region of interest located within the lung bounding box after cropping is the target region of interest.

6. The method as described in claim 1, characterized in that, Based on the relative coordinates of the target region of interest, determine target regions of interest in other chest images that match the target region of interest, including: Based on the relative coordinates of the centroid of the target region of interest in the current chest image and the relative coordinates of the centroid of the target region of interest in the target chest image, calculate the Euclidean distance between the target region of interest in the current chest image and the target region of interest in the target chest image. If the Euclidean distance satisfies the preset matching conditions, the target region of interest in the target chest image is determined to match the target region of interest in the current chest image.

7. A region-of-interest registration device based on a standardized coordinate system, characterized in that, include: The acquisition module is configured to acquire chest images of the same patient at different times; The registration module is configured to perform the following for each of the chest images: determining the lung bounding box based on the chest image; establishing a standardized coordinate system with the center point of the lung bounding box as the origin; determining the target region of interest in the chest image; determining the pixel coordinates and voxel spacing of the centroid of the target region of interest; and calculating the physical coordinates of the centroid based on the pixel coordinates and voxel spacing. Based on the physical coordinates of the centroid and the physical coordinates of the origin, the relative coordinates of the centroid of the target region of interest are determined; Based on the relative coordinates of the target region of interest, and according to Euclidean distance and preset matching conditions, other target regions of interest in chest images that match the target region of interest are determined.

8. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

Images