An automatic positioning method and device based on head clinical images

Abstract

The present application relates to the technical field of image processing, and particularly relates to an automatic positioning method and device based on head clinical image, which performs coordinate transformation on the head anatomical structure of CVH according to spatial direction mapping relationship and spatial size mapping relationship, and obtains CVH registration head anatomical structure; the CVH registration head anatomical structure and the position mark and organ name of each organ region in the CVH registration head anatomical structure constitute a new CVH anatomical knowledge graph. The patient head clinical image is subjected to multi-planar reconstruction, and the patient head anatomical structure is obtained; there is a real-time mapping relationship between the patient head anatomical structure and the CVH registration head anatomical structure, so that the position mark and organ name of each organ region in the patient head anatomical structure and the CVH registration head anatomical structure have a one-to-one correspondence. The interested region is selected from the anatomical knowledge graph, and the corresponding position of the interested region in the patient head anatomical structure is automatically positioned, so that the recognition difficulty of the patient head clinical image is reduced.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to an automatic localization method and device based on clinical head images. Background Technology

[0002] Clinical tomographic imaging, such as CT and MRI, plays a crucial role in disease diagnosis and treatment. However, due to the abstract nature of the imaging methods and the complexity of human anatomy, accurate interpretation of clinical medical images often requires the expertise and experience of senior physicians, especially radiologists. In particular, the complex structures of the head present significant challenges, often appearing too abstract and difficult to accurately interpret for junior medical students, interdisciplinary researchers, biomedical engineers, and ordinary patients.

[0003] For interdisciplinary professionals, young physicians, and medical students, some complex lesions and subtle anatomical structures in medical imaging are difficult for non-experts to effectively locate and identify. For patients and their families, while their awareness of active participation in the diagnosis and treatment process is constantly increasing, and medical institutions are providing imaging data along with browsers via discs, current browsers still require guidance from a professional physician to help patients develop an intuitive understanding of lesions. Patients and their families cannot understand the various organ regions in the images on their own.

[0004] Therefore, in the era of intelligent information that emphasizes universality, the ability to quickly locate regions of interest in clinical head images and present them to non-senior radiology professionals in order to reduce the difficulty of recognizing clinical head images has unique value. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes an automatic localization method and device based on head clinical images, which reduces the difficulty of recognizing patient head clinical images and improves universality.

[0006] In a first aspect, the present invention provides an automatic localization method based on head clinical images.

[0007] In a first feasible approach, an automated localization method based on head clinical images includes:

[0008] Acquire clinical images of the patient's head;

[0009] Register the patient’s head clinical images with CVH to obtain spatial orientation mapping and spatial size mapping;

[0010] Based on the spatial orientation mapping relationship and the spatial size mapping relationship, the coordinate transformation of the CVH head anatomy structure is performed to obtain the CVH registered head anatomy structure; the CVH registered head anatomy structure, together with the positional identifiers and organ names of each organ region in the CVH registered head anatomy structure, forms a new CVH anatomy knowledge graph.

[0011] Multiplanar reconstruction of the patient's head clinical images was performed to obtain the patient's head anatomical structure; there is a real-time mapping relationship between the patient's head anatomical structure and the CVH-registered head anatomical structure.

[0012] Select a region of interest from the anatomical knowledge graph and automatically locate the corresponding position of the region of interest in the anatomical structure of the patient's head.

[0013] In conjunction with the first feasible approach, the second feasible approach involves registering the patient's head clinical images with CVH to obtain spatial orientation mapping and spatial size mapping, including:

[0014] The patient's head clinical images and CVH were first registered to obtain the spatial orientation mapping relationship and the first registration result;

[0015] Based on the first registration result, a second registration was performed on the patient's head clinical images and CVH to obtain the spatial size mapping relationship.

[0016] In the third feasible method, combined with the second feasible method, the patient's head clinical images and CVH are first registered to obtain the spatial orientation mapping relationship and the first registration result, including:

[0017] Three-dimensional contour reconstruction was performed on the clinical images of the patient's head to obtain the patient's three-dimensional contour structure.

[0018] The first feature point is obtained based on the patient's three-dimensional contour structure;

[0019] Obtain the second feature point of the CVH 3D contour structure;

[0020] The first feature point and the second feature point are registered to obtain the spatial orientation mapping relationship and the first registration result; the first registration result is the initial registration structure of the patient's three-dimensional contour, and the spatial orientation of the initial registration structure of the patient's three-dimensional contour is consistent with that of the CVH three-dimensional contour structure.

[0021] In a fourth possible implementation method, combined with the third, the first feature point is obtained based on the patient's three-dimensional contour structure, including:

[0022] The eye-nose triangle region is selected from the patient's three-dimensional contour structure, and the first feature point of the eye-nose triangle region is generated.

[0023] In the fifth feasible method, in conjunction with the third feasible method, a second registration is performed on the patient's head clinical images and CVH based on the first registration result to obtain the spatial size mapping relationship, including:

[0024] The first annular contour is obtained based on the patient's three-dimensional contour structure;

[0025] Obtain the second annular contour of the CVH 3D contour structure;

[0026] In the initial registration structure of the patient's three-dimensional contour, registration is performed based on the first and second annular contours to obtain the spatial size mapping relationship.

[0027] In the sixth possible implementation method, combined with the third implementation method, coordinate transformation is performed on the CVH head anatomy structure according to the spatial orientation mapping relationship and the spatial size mapping relationship to obtain the CVH registered head anatomy structure, including:

[0028] The homogeneous transformation matrix is ​​obtained based on the spatial orientation mapping relationship and the spatial size mapping relationship; the homogeneous transformation matrix is ​​used to characterize the coordinate transformation relationship between the patient's three-dimensional contour structure and the head anatomy structure of CVH.

[0029] The CVH head anatomy structure is obtained by performing coordinate transformation on the homogeneous transformation matrix.

[0030] Combining the sixth feasible method, in the seventh feasible method, the homogeneous transformation matrix is ​​obtained through the following formula:

[0031] Where matrix1 is the transformation matrix of the spatial direction mapping relationship, and matrix2... i Let be the transformation matrix for the spatial size mapping relationship, and matirx be the homogeneous transformation matrix, where i and n are positive integers.

[0032] Combining the first feasible method, the eighth feasible method includes:

[0033] After automatically locating the region of interest within the corresponding anatomical structure of the patient's head, the system marks the corresponding area with color and displays the plane to which the corresponding area belongs.

[0034] In a first aspect, the present invention provides an automatic head clinical image positioning device.

[0035] In the ninth possible embodiment, an automated head clinical image localization device includes:

[0036] The acquisition module is configured to acquire clinical images of the patient's head.

[0037] The registration module is configured to register the patient’s head clinical images and CVH to obtain spatial orientation mapping and spatial size mapping.

[0038] The coordinate transformation module is configured to perform coordinate transformation on the CVH head anatomy based on spatial direction mapping and spatial size mapping to obtain the CVH registered head anatomy. The CVH registered head anatomy, together with the location identifiers and organ names of each organ region in the CVH registered head anatomy, forms a new CVH anatomy knowledge graph.

[0039] The multiplanar reconstruction module is configured to perform multiplanar reconstruction of clinical images of the patient's head to obtain the anatomical structure of the patient's head; there is a real-time mapping relationship between the anatomical structure of the patient's head and the anatomical structure of the head registered by CVH.

[0040] The automatic localization module is configured to select a region of interest from the anatomical knowledge graph and automatically locate the corresponding position of the region of interest in the anatomical structure of the patient's head.

[0041] As can be seen from the above technical solution, the beneficial technical effects of the present invention are as follows:

[0042] Based on spatial orientation and size mapping relationships, coordinate transformation is performed on the head anatomy of CVH (CVH-registered head anatomical structure) to obtain the CVH-registered head anatomical structure. A new CVH anatomical knowledge graph is then formed by the location identifiers and organ names of each organ region within the CVH-registered head anatomical structure. Multiplanar reconstruction is then performed on the patient's head clinical images to obtain the patient's head anatomical structure. A real-time mapping relationship exists between the patient's head anatomical structure and the CVH-registered head anatomical structure; therefore, there is a one-to-one correspondence between the location identifiers and organ names of each organ region within the patient's head anatomical structure and the CVH-registered head anatomical structure. Regions of interest (ROIs) are selected from the anatomical knowledge graph, and their corresponding positions within the patient's head anatomical structure are automatically located. This allows even non-specialist physicians to identify various organ regions in the patient's head clinical images, reducing the difficulty of identification and benefiting interdisciplinary professionals, young physicians, medical students, patients, and their families in understanding lesions in patient's head clinical images, thus improving universality. Attached Figure Description

[0043] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0044] Figure 1This is a schematic diagram of an automatic localization method based on head clinical images provided in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of a method for automatic localization of head clinical images provided in an embodiment of the present invention. Detailed Implementation

[0046] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0047] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0048] Combination Figure 1 As shown, this embodiment provides an automatic localization method based on head clinical images, including:

[0049] Step S01: Obtain clinical images of the patient's head;

[0050] Step S02: Register the patient's head clinical images and CVH to obtain the spatial orientation mapping relationship and spatial size mapping relationship;

[0051] Step S03: Perform coordinate transformation on the head anatomy of CVH according to the spatial direction mapping relationship and the spatial size mapping relationship to obtain the CVH registered head anatomy; the CVH registered head anatomy, together with the positional identifiers and organ names of each organ region in the CVH registered head anatomy, form a new CVH anatomy knowledge graph.

[0052] Step S04: Perform multiplanar reconstruction on the patient's head clinical images to obtain the patient's head anatomical structure; the CVH-registered head anatomical structure and the location markers and organ names of each organ region in the CVH-registered head anatomical structure form a new CVH anatomical knowledge graph.

[0053] Step S05: Select the region of interest from the anatomical knowledge graph and automatically locate the corresponding position of the region of interest in the anatomical structure of the patient's head.

[0054] Based on spatial orientation and size mapping relationships, coordinate transformation is performed on the head anatomy of CVH (CVH-registered head anatomical structure) to obtain the CVH-registered head anatomical structure. A new CVH anatomical knowledge graph is then constructed by combining the location markers and organ names of each organ region within the CVH-registered head anatomical structure. Multiplanar reconstruction is then performed on the patient's head clinical images to obtain the patient's head anatomical structure. A real-time mapping relationship exists between the patient's head anatomical structure and the CVH-registered head anatomical structure; therefore, there is a one-to-one correspondence between the location markers and organ names of each organ region within the patient's head anatomical structure and the CVH-registered head anatomical structure. Regions of interest (ROIs) are selected from the anatomical knowledge graph, and their corresponding positions within the patient's head anatomical structure are automatically located based on the real-time mapping relationship between the patient's head anatomical structure and the CVH-registered head anatomical structure. This allows even non-specialist physicians to identify various organ regions in the patient's head clinical images, reducing the difficulty of identification and benefiting interdisciplinary professionals, young physicians, medical students, patients, and their families in understanding lesions in patient's head clinical images, thus improving universality.

[0055] Optionally, clinical images of the patient's head can be obtained via CT (Computed Tomography) or MRI (Nuclear Magnetic Resonance Imaging).

[0056] In some embodiments, CVH (Chinese Visible Human) includes a Chinese visualized human cross-sectional anatomy dataset and corresponding CT and MRI datasets. The datasets include clinical images and anatomical structures of various body parts. CVH includes an anatomical knowledge graph, which contains anatomical structures of various body parts, as well as location identifiers and organ names within those structures. The three-dimensional contour structure of the head in CVH is obtained by reconstructing the three-dimensional contour from clinical head images.

[0057] Optionally, the CVH-registered head anatomy structure replaces the old CVH head anatomy knowledge graph, forming a new CVH anatomy knowledge graph. A one-to-one mapping exists between the patient's head anatomy structure and the CVH-registered head anatomy structure, with location markers and organ names for each organ region present in the CVH-registered head anatomy structure. Therefore, after mapping each region in the patient's head anatomy structure to its corresponding region in the CVH-registered head anatomy structure, the location markers and organ names for each organ region in the CVH-registered head anatomy structure also correspond to their respective regions in the patient's head anatomy structure. Automatic localization is achieved based on this correspondence between the location markers and organ names of each organ region in the CVH-registered head anatomy structure and the corresponding regions in the patient's head anatomy structure. For example, if there is a mapping relationship between A and B, selecting A will map B to A, and the location markers and organ names related to B will also correspond to A.

[0058] Optionally, the patient's head clinical images and CVH are registered to obtain spatial orientation mapping and spatial size mapping, including: performing a first registration of the patient's head clinical images and CVH to obtain spatial orientation mapping and a first registration result; and performing a second registration of the patient's head clinical images and CVH based on the first registration result to obtain spatial size mapping.

[0059] Optionally, the Iterative Closest Point (ICP) algorithm is used to perform a first registration of the patient's head clinical image and CVH, obtaining a transformation matrix of spatial orientation mapping and a first registration result. Based on the first registration result, the ICP algorithm is then used to perform a second registration of the patient's head clinical image and CVH, obtaining a transformation matrix of spatial size mapping. In some embodiments, in conventional heterogeneous registration, due to morphological differences between different human heads, prominent features such as the nose and ears, as well as similar gradient features such as the forehead and occipital region, are prone to erroneous "best" matches during registration, i.e., local extremum problems. Therefore, the deviation is still relatively large, and the registration result is far from ideal. This application uses the result of the first registration as the basis for the second registration, performing a second progressive registration. This avoids the erroneous "best" match caused by the ICP algorithm stopping iteration after getting stuck in local extrema during the iteration process, effectively reducing deviation, improving the registration degree between the patient's head clinical image and CVH, and thus improving the accuracy of automatic localization.

[0060] Optionally, the patient's head clinical image and CVH are registered for the first time to obtain a spatial orientation mapping relationship and a first registration result, including: reconstructing the three-dimensional contour of the patient's head clinical image to obtain the patient's three-dimensional contour structure; obtaining a first feature point based on the patient's three-dimensional contour structure; obtaining a second feature point of the CVH three-dimensional contour structure; registering the first feature point and the second feature point to obtain a spatial orientation mapping relationship and a first registration result; the first registration result is the CVH initial registration three-dimensional contour structure, and the spatial orientation of the patient's three-dimensional contour structure is consistent with that of the CVH initial registration three-dimensional contour structure.

[0061] Optionally, the first and second feature points are registered using the Iterative Closest Point (ICP) algorithm. In some embodiments, the ICP algorithm is a method based on quaternions that achieves surface registration by iteratively calculating the sum of squared residuals of corresponding points between surfaces. For example, consider two sets of coordinate points in different world coordinate systems: P = {Pi, i = 0, 1, 2, ..., k} (red block) and U = {Ui, i = 0, 1, 2, ..., n} (blue block). Let P be the target set and U be the source set. Assuming that P and U roughly correspond in space (usually k ≥ n), a new set of points U′ is obtained by continuously rotating and translating the point set U, minimizing the distance between the source points of U′ and P (making U′ and P overlap as much as possible). U′ can be obtained using the rigid body geometric transformation formula U′ = RU + T, where R represents the three-dimensional rotation matrix of the transformed point set U, and T represents the translation vector of the transformed point set U. The core of this process is to use the least mean square method, which continuously corrects R and T to iteratively calculate the sum of squared residuals between corresponding points in the point sets U′ and P, and finds the minimum mean square error between U′ and Q. If the error is less than the preset limit value, the iteration ends, and the optimal registration solution is obtained.

[0062] Optionally, obtaining the first feature points based on the patient's three-dimensional contour structure includes: selecting the eye-nose triangle region from the patient's three-dimensional contour structure and generating the first feature points of the eye-nose triangle region. In some embodiments, the eye-nose triangle region is distinct and has large differences, and using it as a facial feature point for registration avoids the local extremum problem of ordinary heterologous registration, making the registration more accurate.

[0063] Optionally, a second registration is performed on the patient's head clinical image and CVH based on the first registration result to obtain a spatial size mapping relationship, including: obtaining a first annular contour based on the patient's three-dimensional contour structure; obtaining a second annular contour of the CVH three-dimensional contour structure; and registering based on the first annular contour and the second annular contour based on the patient's three-dimensional contour structure and the CVH's initially registered three-dimensional contour structure to obtain a spatial size mapping relationship.

[0064] In some embodiments, obtaining the annular contour includes reading the raw data using the C++ open-source FO-DICOM library and converting it into an 8-bit bitmap format, then performing binarization using the Ostu adaptive threshold segmentation method; filling small intracranial holes using opening operations and filling the entire intracranial region using the FloodFill algorithm; finally, using a 3*3 erosion template, performing front and back subtraction operations to extract the contour of one frame of data, and continuing to process the next frame of data until batch processing is completed, ultimately achieving the extraction of the annular contour.

[0065] In some embodiments, the first registration determines the spatial orientation mapping relationship. After the first registration, the spatial orientation transformation between the patient's 3D contour structure and the CVH 3D contour structure is made consistent. After the spatial orientation between the patient's 3D contour structure and the CVH 3D contour structure is consistent, a ring-shaped feature contour with the same data is manually selected from both the patient's 3D contour structure and the CVH 3D contour structure. Based on the ring-shaped feature contour, further spatial size registration is performed to accommodate the size differences of different human heads. This allows the CVH 3D contour structure to approach the patient's 3D contour structure after two registrations, thereby facilitating real-time mapping between the two. At the same time, manually selecting ring-shaped feature contours with the same data from both the patient's 3D contour structure and the CVH 3D contour structure reduces the risk of local extrema.

[0066] Optionally, coordinate transformation is performed on the head anatomy of the CVH according to the spatial orientation mapping relationship and the spatial size mapping relationship to obtain the CVH registered head anatomy structure, including: obtaining a homogeneous transformation matrix according to the spatial orientation mapping relationship and the spatial size mapping relationship; the homogeneous transformation matrix is ​​used to characterize the coordinate transformation relationship between the patient's three-dimensional contour structure and the head anatomy of the CVH; and coordinate transformation is performed on the head anatomy of the CVH according to the homogeneous transformation matrix to obtain the CVH registered head anatomy structure.

[0067] Optionally, the CVH head anatomy is transformed using spatial orientation and spatial size mapping relationships to make the CVH head anatomy resemble the patient's three-dimensional head contour, aligning its spatial orientation and size. A one-to-one mapping exists between the transformed CVH-registered head anatomy and the patient's three-dimensional head contour. Then, multi-planar reconstruction of the patient's three-dimensional head contour is performed and superimposed within the same scene to obtain the final multi-planar patient head anatomy. The patient's head anatomy corresponds one-to-one with the CVH-registered head anatomy; similarly, the CVH head anatomy corresponds one-to-one with the anatomical knowledge graph. Ultimately, the patient's head anatomy is mapped to both the CVH-registered head anatomy and the anatomical knowledge graph.

[0068] In some embodiments, the patient's head clinical images (such as CT / MRI and other patient image data) are used as the target image, and the CVH dataset is used as a floating image to be registered to the target image to obtain a homogeneous transformation matrix; the CVH gross anatomical structure segmentation layer is transformed synchronously with the movement of the floating image and registered with the patient's head clinical images. After three-dimensional reconstruction, a real-time mapping at any viewpoint is obtained to achieve the purpose of lesion / organ navigation. The registration results allow for rapid evaluation and correction.

[0069] Alternatively, the homogeneous transformation matrix can be obtained using the following formula:

[0070] Where matrix1 is the transformation matrix of the spatial direction mapping relationship, and matrix2... i Let be the transformation matrix for the spatial size mapping relationship, and matirx be the homogeneous transformation matrix. i and n are positive integers, and n represents the number of progressive registrations.

[0071] Optionally, during the multiplanar reconstruction of the patient's head clinical images, multiplanar reconstruction can be performed by referring to existing multiplanar reconstruction methods to obtain the patient's anatomical structure in multiple planes.

[0072] Optionally, the automatic localization method based on head clinical images includes: automatically locating the corresponding position of the region of interest in the patient's head anatomy, marking the corresponding position area with color, and displaying the plane to which the corresponding position area belongs.

[0073] Optionally, this embodiment provides an automatic localization method based on head clinical images, including: acquiring patient head clinical images; performing secondary progressive registration between the patient head clinical images and CVH (Chinese Visible Human) to obtain CVH-registered head anatomical structures; performing multiplanar reconstruction on the patient head clinical images to obtain the patient head anatomical structures; real-time mapping between the patient head anatomical structures and the anatomical knowledge graph in CVH; selecting regions of interest from the anatomical knowledge graph and automatically locating the corresponding positions of the regions of interest in the patient head anatomical structures.

[0074] result Figure 2As shown, this invention provides an automatic head clinical image localization device, including: an acquisition module 101, a registration module 102, a coordinate transformation module 103, a multi-plane reconstruction module 104, and an automatic localization module 105. The acquisition module 101 is configured to acquire a patient's head clinical image; the registration module 102 is configured to register the patient's head clinical image and the CVH (CVH-associated head anatomical structure) to obtain a spatial orientation mapping relationship and a spatial size mapping relationship; the coordinate transformation module 103 is configured to perform coordinate transformation on the CVH head anatomical structure according to the spatial orientation mapping relationship and the spatial size mapping relationship to obtain a CVH-registered head anatomical structure; the CVH-registered head anatomical structure, along with the location identifiers and organ names of each organ region in the CVH-registered head anatomical structure, forms a new CVH anatomical knowledge graph; the multi-plane reconstruction module 104 is configured to perform multi-plane reconstruction on the patient's head clinical image to obtain the patient's head anatomical structure; a real-time mapping relationship exists between the patient's head anatomical structure and the CVH-registered head anatomical structure; the automatic localization module 105 is configured to select a region of interest from the anatomical knowledge graph and automatically locate the corresponding position of the region of interest in the patient's head anatomical structure.

[0075] In some embodiments, the automatic localization method for head clinical images includes: Step S21, the user selects clinical image data, and displays a 3D contour structure model of the CVH standard atlas and an automated 3D contour structure model of the user-selected clinical image data. Step S22, the two contour models are adjusted to a convenient viewing angle for selecting the "eye-nose triangle" (no need to set the same size and orientation), the "eye-nose triangle" is selected by drawing a rectangle, and blue and red feature points are generated on the CVH (left) and CT / MRI models respectively according to the drawn area. Step S23, the first registration is performed to obtain the matrix equation of the spatial orientation mapping relationship, and the coarse registration result is displayed; the blue CVH feature points will move closer to the red CT / MRI feature points. Step S24, the annular contour is automatically extracted and a second registration is performed to obtain the final homogeneous matrix equation to guide the CVH anatomical knowledge data to move closer to the CT / MRI data. The registration result is presented intuitively by overlaying the same scene, and the overlay effect of the two sets of images is displayed.

[0076] Multiplanar Reconstruction (MPR) is performed on CT / MRI, displaying the results and listing the gross organs of the CVH for easy filtering and selection. For example, selecting "brainstem" will highlight the brainstem area in yellow within the corresponding MPR reconstruction area of ​​the CT / MRI, making it easier for the user to identify.

[0077] In some embodiments, the rigid structural correspondence features between the head region and the gross anatomical structure of the CVH from clinical image sets (CT and MRI) are collected and analyzed to complete the 3D reconstruction and feature point picking of the rigid features of the external contour. Then, a two-stage progressive registration method is adopted. First, the spatial orientation is determined by rigid transformation of facial features in the "eye-nose triangle" to complete coarse registration. Then, similarity registration using annular contour features is used to determine the spatial size, solving the local extremum problem and deformation problem that is easy to occur in heterogeneous registration. A standardized workflow including the two-stage registration method is summarized. Finally, the head images of labor are resampled and spatially transformed to obtain the patient's head anatomical structure. Based on the registration results, the coordinate field transformation is performed on the volume data field formed by the gross anatomical structure layer set of CVH to achieve the purpose of mapping the anatomical structure to the clinical image data. At the same time, the UI interface of MPR multi-plane reconstruction interaction is designed and optimized to achieve a productized application form.

[0078] In some embodiments, existing registration studies of anatomical knowledge graphs and medical images have certain requirements for the integrity and quality of the image structure, which differs from the imaging scanning standards used in clinical practice. To prevent overtreatment and control medical costs, real clinical data is often localized rather than the thin-slice high-field scans required for research, making it difficult to meet the "complete structure" and quality requirements of current mainstream registration algorithms. Solutions often require enormous computational resources. However, the automatic localization method based on head clinical images provided in this application has low computational requirements, low requirements for image scanning, and is lightweight.

[0079] In some embodiments, the automatic localization method based on head clinical images provided in this application can map the anatomical knowledge graph in CVH to a private CT / MRI dataset in real time without the need for professional explanation or special 3D organ printing annotation. It can intuitively present the current organ region and locate the region of interest in an intelligent navigation manner, achieving the purpose of unmanned intelligent assisted "explanation". It also has human-computer interaction features, serving the needs of patients and their families to actively participate in disease diagnosis and treatment, as well as the work needs of junior medical students, interdisciplinary researchers, and medical engineers, and jointly supporting the general direction of "patient-centered".

[0080] In some embodiments, the open-source components of this application include: Visual Studio 2022 Community Edition development platform, core R&D platform; VTK (Visualization Toolkit), for visual reconstruction of medical tomographic images; OpenCV, for morphological operations on medical images; and FO-DICOM, for reading and writing medical DICOM images.

[0081] Optionally, an automatic localization method based on head clinical images further includes: acquiring test data, performing time consumption tests and functional tests on the automatic localization method based on head clinical images according to the test data, and obtaining test results.

[0082] In some embodiments, Table 3 is an example table of test data. In Table 3, data 1, 2, and 3 have different positioning and different head scanning intervals, and come from three patients with very different facial features. Therefore, the applicability of the method in this paper can be evaluated to a certain extent. Each set of data was repeated twice to obtain a relatively objective evaluation of reproducibility. Unlike the current mainstream registration algorithms that require the "complete structure" of image data to reflect the diverse characteristics of real clinical image scanning sites, three representative tomographic image datasets were selected in this test. Among them, data 1 is CT image, and data 2 and 3 are MRI images with poor spatial resolution, with spatial resolutions of 512*512*121, 288*384*18, and 160*126*160, respectively, and pixel pitches of 0.43*0.43*0.70, 0.625*0.625*3.15, and 1.625*1.625*1.625, respectively. Considering that clinical imaging scans emphasize targeted imaging of lesion areas in order to prevent overtreatment and control medical costs, the scanning range of Data 1 only covers the forehead to the nasal alae, while the range of Data 2 and Data 3 is further compressed to the brow bone to the nasal alae, which can better test the robustness of this project's solution in registration.

[0083] Table 1 Example of test data

[0084]

[0085] In some embodiments, the time consumption of the automatic localization method based on head clinical images is tested using test data. This includes: performing 3D reconstruction of the CT / MRI and the registered CVH dataset in the same scene, and evaluating the registration accuracy based on the fit between the two. Since CVH is a 24-bit true-color bitmap dataset, its imaging characteristics are completely different from CT and MRI. Therefore, setting two sets of volume rendering shader ranges and transparency schemes can clearly separate them. During the test, some typical anatomical structures are selected and mapped onto CT / MRI images. The registration accuracy is evaluated based on the proportion of the covered areas. Typical anatomical structures selected are the brainstem and optic nerve, as their positional relationship in the head is representative, and their shape and size are moderate.

[0086] In some embodiments, Table 2 presents the test results. As shown in Table 2, a total of five repetitive tests were conducted on a typical PC configuration with an i5-4210, 2.60GHz processor, 2 x 8GB RAM, and Windows 7. All tests ran normally, and the time taken for each test did not exceed 40 seconds, indicating a relatively fast speed.

[0087] Table 2

[0088]

[0089]

[0090] In some embodiments, the automatic localization method based on head clinical images is functionally tested according to test data. Images from data sets one, two, and three are registered using the method of this application, resulting in three registration results. In the overlay image of the unregistered clinical image and the original spatial positioning of the CVH head in the same scene, the spatial difference between the two is clearly visible. Screenshots from two fixed viewing angles after registration show that, subjectively, the registration result is considered very ideal. This method also serves as a reference for ordinary operators to determine whether to perform more than three re-registration corrections. Based on this, a radiologist with five years of experience was invited to evaluate the observation results using a five-point scale. After three registrations for each of the three data sets, a score of nine registration results was obtained, and the score distribution is shown in Table 3 below:

[0091] Table 3

[0092]

[0093] As shown in Table 3, out of the nine registration attempts, seven were successful on the first attempt following the procedure, with satisfactory results. Two required further similar registration corrections to achieve satisfactory results. No cases achieved the basic registration objective or were below that level. Furthermore, from the perspective of professional radiologists, the registration results are generally considered to meet the general needs of the public as proposed in this project.

[0094] In some embodiments, the automatic localization method based on head clinical images is functionally tested according to test data, including: selecting brainstem and optic nerve for calibration measurements, and performing functional tests based on the measurement results. Data sets one, two, and three are each registered twice, and tomographic images of the same location from one of the registrations are randomly selected for calibration illustration. In some embodiments, tomographic images of the same location from one of the registrations are randomly selected for calibration illustration, obtaining multiple frames of complete anatomical structures' brainstem and optic nerve mappings, as well as brainstem and optic nerve mappings of anatomical structures cut along the sampling plane. Based on this, a radiologist with five years of experience is invited to evaluate the observation results using a five-point scale. After three registrations for each of the three data sets, a score of nine registrations is obtained, and the score distribution is shown in Table 4 below:

[0095] Table 4

[0096]

[0097] The table shows that the mapping accuracy for the brainstem and optic nerve is rated A in 3 out of 9 registrations, and B in 6 out of 5. For heterogeneous data, rigid and similar registrations cannot achieve perfect matching. However, most of the matches meet the project's goal of enabling rapid interpretation of clinical tomographic images for general patients, interdisciplinary researchers, biomedical engineers, and junior medical students, offering significant advantages in terms of lightweight design and processing speed. Furthermore, from the perspective of professional radiologists, the registration results are generally considered suitable for the project's general public needs.

[0098] In some embodiments, this application proposes an automatic localization method based on head clinical images. This method enables rapid regional similarity registration and spatial mapping of head region tomographic image data (such as CT and MRI) from any clinical patient with the Chinese Visual Human Dataset (CVH). Based on the existing gross anatomical knowledge graph of CVH, the specific anatomical names of each region in the clinical image data are labeled, allowing the originally abstract clinical tomographic images to be compared with the high-resolution CVH atlas. Through various retrieval methods, any anatomical region can be intuitively displayed in the form of multiplanar reconstruction (MPR) and volumetric reconstruction (VR), demonstrating a certain degree of technological innovation. This method is characterized by its simple operation, lightweight computation, and green, installation-free nature. It can intuitively present the region of interest in the current image to the user through intelligent navigation, achieving the purpose of unmanned intelligent assisted "explanation." This serves the active participation needs of patients and their families, as well as the work needs of junior medical students, interdisciplinary researchers, and biomedical engineers, demonstrating a certain degree of application innovation. The test results show that this method can achieve heterogeneous multimodal registration, which meets the functional goals proposed in this paper. It also has good adaptability and reproducibility for clinical imaging data, making it suitable for general public needs. The processing time on ordinary personal computers is less than 40 seconds, which has a great advantage of being lightweight. Furthermore, it can run on any Windows desktop system with .NET Framework 4.0 or higher, meeting the development requirements of a lightweight intelligent positioning system based on head CT / MRI.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. An automatic localization method based on head clinical images, characterized in that, include: Acquire clinical images of the patient's head; The patient's head clinical images and CVH were registered to obtain spatial orientation mapping and spatial size mapping, including: Spatial orientation is determined by rigid transformation of facial features in the eye-nose triangle region. Spatial dimensions are determined by similarity registration using ring-shaped contour features; Based on the spatial direction mapping relationship and the spatial size mapping relationship, the CVH head anatomy structure is subjected to coordinate transformation to obtain the CVH registered head anatomy structure; the CVH registered head anatomy structure, together with the positional identifiers and organ names of each organ region in the CVH registered head anatomy structure, forms a new CVH anatomy knowledge graph; Multiplanar reconstruction was performed on the clinical images of the patient's head to obtain the anatomical structure of the patient's head; there is a real-time mapping relationship between the anatomical structure of the patient's head and the CVH-registered anatomical structure of the head. The region of interest is selected from the anatomical knowledge graph, and the corresponding position of the region of interest in the patient's head anatomy is automatically located.

2. The method according to claim 1, characterized in that, The patient's head clinical images and CVH were registered to obtain spatial orientation mapping and spatial size mapping, including: The patient's head clinical images and CVH were first registered to obtain the spatial orientation mapping relationship and the first registration result; Based on the first registration result, a second registration is performed on the patient's head clinical image and CVH to obtain the spatial size mapping relationship.

3. The method according to claim 2, characterized in that, The patient's head clinical images and CVH were first registered to obtain the spatial orientation mapping relationship and the first registration result, including: Three-dimensional contour reconstruction was performed on the clinical images of the patient's head to obtain the patient's three-dimensional contour structure. The first feature point is obtained based on the patient's three-dimensional contour structure; Obtain the second feature point of the CVH 3D contour structure; The first feature point and the second feature point are registered to obtain a spatial orientation mapping relationship and a first registration result; the first registration result is the CVH initial registration three-dimensional contour structure, and the spatial orientation of the patient's three-dimensional contour structure is consistent with that of the CVH initial registration three-dimensional contour structure.

4. The method according to claim 3, characterized in that, The first feature point is obtained based on the patient's three-dimensional contour structure, including: The eye-nose triangle region is selected from the patient's three-dimensional contour structure, and the first feature point of the eye-nose triangle region is generated.

5. The method according to claim 3, characterized in that, Based on the first registration result, a second registration is performed on the patient's head clinical image and CVH to obtain a spatial size mapping relationship, including: A first annular contour is obtained based on the patient's three-dimensional contour structure; Obtain the second annular contour of the CVH 3D contour structure; In the patient's three-dimensional contour structure and the CVH initial registration three-dimensional contour structure, registration is performed based on the first annular contour and the second annular contour to obtain a spatial size mapping relationship.

6. The method according to claim 3, characterized in that, Based on the spatial orientation mapping relationship and the spatial size mapping relationship, the CVH head anatomy structure is subjected to coordinate transformation to obtain the CVH registered head anatomy structure, including: The homogeneous transformation matrix is ​​obtained based on the spatial orientation mapping relationship and the spatial size mapping relationship; the homogeneous transformation matrix is ​​used to characterize the coordinate transformation relationship between the patient's three-dimensional contour structure and the head anatomy structure of CVH. The CVH head anatomy structure is subjected to coordinate transformation based on the homogeneous transformation matrix to obtain the CVH registered head anatomy structure.

7. The method according to claim 6, characterized in that, The homogeneous transformation matrix is ​​obtained by the following formula: Where matrix1 is the transformation matrix for spatial direction mapping. The transformation matrix represents the spatial dimension mapping relationship. Let i be a homogeneous transformation matrix, where i and n are positive integers.

8. The method according to claim 1, characterized in that, include: After automatically locating the corresponding position of the region of interest in the patient's head anatomy, the corresponding position area is marked with color, and the plane to which the corresponding position area belongs is displayed.

9. An automatic head clinical imaging positioning device, characterized in that, include: The acquisition module is configured to acquire clinical images of the patient's head. The registration module is configured to register the patient's head clinical images and CVH to obtain spatial orientation mapping and spatial size mapping, including: Spatial orientation is determined by rigid transformation of facial features in the eye-nose triangle region. Spatial dimensions are determined by similarity registration using ring-shaped contour features; The coordinate transformation module is configured to perform coordinate transformation on the head anatomy of CVH according to the spatial direction mapping relationship and the spatial size mapping relationship to obtain the CVH registered head anatomy structure; the CVH registered head anatomy structure and the position identifiers and organ names of each organ region in the CVH registered head anatomy structure form a new CVH anatomy knowledge graph. The multiplanar reconstruction module is configured to perform multiplanar reconstruction on the clinical images of the patient's head to obtain the anatomical structure of the patient's head; there is a real-time mapping relationship between the anatomical structure of the patient's head and the CVH-registered anatomical structure of the head. An automatic localization module is configured to select a region of interest from the anatomical knowledge graph and automatically locate the corresponding position of the region of interest in the anatomical structure of the patient's head.

Images