Worm nerve foramen processing method and device and computing equipment

Abstract

The invention provides a processing method and device of a volute nerve hole and computing equipment. Determining target temporal bone area scanning data, and inputting the target temporal bone area scanning data into a pre-trained cochlea segmentation model to obtain a cochlea segmentation mask; adjusting a cochlea posture according to the cochlea segmentation mask, determining a cochlea size according to the cochlea posture, determining a cochlea cutting template according to the cochlea size, and cutting the initial temporal bone area scanning data to obtain a cochlea and internal auditory canal view; defining and extracting an internal auditory canal structure, and based on a cochlea axis direction corresponding to the cochlea posture, rotating the cochlea nerve foramen view and the internal auditory canal structure, the cochlea nerve foramen view being obtained by separating the cochlea and the internal auditory canal view; and determining whether the cochlear foramen is narrow or not through the rotated cochlear foramen view and the internal auditory canal structure. According to the technical scheme provided by the invention, whether the volute nerve hole is narrow or not can be quickly judged, so that the diagnosis precision and efficiency are improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of medical technology, and in particular, to a method, apparatus, and computing device for processing a cochlear nerve foramen. Background Art

[0002] The cochlear nerve is a key component of the auditory conduction pathway, responsible for converting sound signals in the cochlea into nerve impulses and transmitting them to the brain for processing. Hypoplasia of the cochlear nerve, manifested as the absence or significant thinning of cochlear nerve fibers, is an important cause of unilateral hearing loss in children, seriously affecting the patient's hearing function and quality of life. This lesion is closely related to the structural abnormalities of the auditory system and has become one of the focuses of clinical and scientific research.

[0003] Currently, clinical evaluation of the cochlear foramen mainly relies on computed tomography (CT) imaging examinations. However, the technology for determining whether the cochlear foramen is narrow is still in its infancy, and the commonly used manual measurement method is dominant. This process includes image acquisition, identification of anatomical structures, and direct measurement of distance or area on CT images. Doctors or technicians need to rely on their rich experience to identify the cochlear foramen in complex anatomical structures, manually mark its boundaries, and then measure its width to determine whether it is narrow.

[0004] However, manual measurement is highly dependent on the operator's experience and judgment. There may be large measurement deviations between different operators, affecting the consistency and accuracy of the measurement results. Summary of the invention

[0005] The embodiments of the present application provide a method, apparatus and computing device for processing a cochlear nerve foramen, so as to solve the problem of low accuracy in processing a cochlear nerve foramen in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a method for processing a cochlear nerve foramen, comprising:

[0007] Acquire initial temporal bone region scan data, and segment and obtain image data of the external semicircular canal based on the initial temporal bone region scan data;

[0008] Rotating the initial temporal bone region scan data according to the image data of the external semicircular canal to obtain target temporal bone region scan data in which the external semicircular canal is parallel to the cross section, wherein the target temporal bone region scan data includes three-dimensional inner ear data;

[0009] Inputting the target temporal bone region scan data into a pre-trained cochlear segmentation model, so as to segment the target temporal bone region scan data by the cochlear segmentation model to obtain a cochlear segmentation mask, wherein the cochlear segmentation mask includes the cochlea and the surrounding structures of the cochlea;

[0010] adjusting the cochlear posture according to the cochlear segmentation mask, determining the size of the cochlear according to the cochlear posture, determining the cochlear cropping template according to the cochlear size and cropping the initial temporal bone region scan data to obtain a cochlear and internal auditory canal view;

[0011] Defining and extracting the internal auditory canal structure, and rotating the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and the internal auditory canal structure, wherein the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views;

[0012] Through the rotated view of the cochlear foramen and the internal auditory canal structure, it can be determined whether the cochlear foramen is narrow.

[0013] Optionally, adjusting the cochlear posture according to the cochlear segmentation mask, determining the cochlear size according to the cochlear posture, determining the cochlear cropping template according to the cochlear size and cropping the initial temporal bone region scan data to obtain a cochlear and internal auditory canal view includes:

[0014] Analyzing all the voxel coordinates of the cochlea in the cochlear segmentation mask to determine the coordinates of the cochlear apex region, and adjusting the cochlear posture based on the coordinates of the cochlear apex region so that the modiolus direction is perpendicular to the horizontal plane;

[0015] In the cochlear posture, measuring the Z-axis distance from the coordinates of the cochlear apex region to the coordinates of the cochlear base region to determine the size of the cochlea, wherein the size of the cochlea at least includes the height of the cochlea;

[0016] A cochlear cropping template is determined according to the cochlear height, and the initial temporal bone region scan data is cropped based on the cochlear cropping template to obtain a view of the cochlea and internal auditory canal.

[0017] Optionally, it also includes:

[0018] Determining the cochlear voxels in the cochlear cropping template by using the cochlear cropping template and the cochlear segmentation mask;

[0019] The cochlear voxels are separated from the cochlear and internal auditory canal view to generate a cochlear neural foramen view.

[0020] Optionally, defining and extracting the internal auditory canal structure includes:

[0021] Determine the X-ray absorption values ​​of the internal auditory canal structures;

[0022] In the cochlear segmentation mask, voxels whose X-ray absorption values ​​are smaller than a set first threshold and larger than the first threshold and smaller than a second threshold are determined as internal auditory canal structures.

[0023] Optionally, the analyzing all cochlear voxel coordinates in the cochlear segmentation mask to determine the cochlear apex region coordinates includes: extracting all cochlear voxel coordinates from the cochlear segmentation mask to construct a first array; performing principal component analysis on the first array to determine the cochlear voxel coordinates with the largest Z value among all cochlear voxel coordinates, and determining the cochlear voxel coordinates with the largest Z value as the cochlear apex region coordinates;

[0024] The adjusting the cochlear posture based on the cochlear apex region coordinates comprises: taking the cochlear apex region coordinates as the center, in the cochlear segmentation mask, cutting out a region of set length, width and height to form a second array; performing principal component analysis on the second array to obtain a plurality of principal component vectors, and constructing an affine transformation matrix based on the plurality of principal component vectors; transforming the first array according to the affine transformation matrix to adjust the cochlear posture so that the modiolar axis direction is perpendicular to the horizontal plane;

[0025] The method of cropping the initial temporal bone region scan data based on the cochlear cropping template to obtain a view of the cochlea and internal auditory canal includes: traversing each voxel in the initial temporal bone region scan data to detect whether the coordinates of each voxel are located in the three-dimensional space region defined by the cochlear cropping template; determining the CT value corresponding to the voxel located in the three-dimensional space region defined by the cochlear cropping template; rearranging the CT values ​​of multiple voxels according to the positions of the corresponding voxels in the three-dimensional space region to generate a data set, and visually displaying the data set to generate a view of the cochlea and internal auditory canal.

[0026] Optionally, before rotating the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain the rotated cochlear nerve foramen view and the internal auditory canal structure, it also includes: solving the corresponding inverse transformation matrix according to the affine transformation matrix, and using the corresponding row vector in the inverse transformation matrix as the modiolar axis direction corresponding to the cochlear posture;

[0027] The method of rotating the cochlear foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain a rotated cochlear foramen view and the internal auditory canal structure includes: querying the cross-sectional plane with the most internal auditory canal voxels along the cross-sectional plane through the initial temporal bone region scan data; determining the rotation angle according to the X component and the Y component in the modiolar axis direction; constructing a rotation matrix around the Z axis based on the rotation angle, and rotating the cochlear voxels in the cochlear foramen view and the internal auditory canal voxels in the internal auditory canal structure in the cross-sectional plane based on the rotation matrix to obtain a rotated cochlear foramen view and the internal auditory canal structure.

[0028] Optionally, determining whether the cochlear nerve foramen is narrow by viewing the rotated cochlear nerve foramen view and the internal auditory canal structure includes:

[0029] In the rotated internal auditory canal structure, the rotated internal auditory canal structure is divided into multiple layers along the Z axis, and the voxel area of ​​each layer of the internal auditory canal is counted layer by layer to determine the slope of the change of the internal auditory canal area;

[0030] Determine the height of the internal auditory canal according to the number of divided layers and the thickness of each layer of the internal auditory canal obtained;

[0031] Determine whether the internal auditory canal height meets the set height, and if not, determine whether the internal auditory canal area change slope meets the preset slope;

[0032] If not, the cochlear nerve foramen is determined to be narrow.

[0033] Optionally, determining whether the cochlear nerve foramen is narrow by viewing the rotated cochlear nerve foramen view and the internal auditory canal structure includes:

[0034] Determine a minimum ordinate in the rotated cochlear foramen view, and offset the minimum ordinate downward by a preset voxel unit to obtain a target ordinate;

[0035] In the rotated internal auditory canal structure, a plurality of data points whose ordinates are the target ordinates are determined, and a maximum abscissa and a minimum abscissa are determined among the plurality of data points to determine a distance between the maximum abscissa and the minimum abscissa;

[0036] Calculate the size of the cochlear foramen according to the obtained voxel side length and the distance between the maximum abscissa and the minimum abscissa;

[0037] If the size of the cochlear nerve foramen does not meet the set size, it is determined that the cochlear nerve foramen is narrow.

[0038] In a second aspect, an embodiment of the present application provides a device for processing a cochlear nerve foramen, comprising:

[0039] An acquisition module is used to acquire initial temporal bone region scan data, and obtain image data of the lateral semicircular canal based on segmentation of the initial temporal bone region scan data; rotate the initial temporal bone region scan data according to the image data of the lateral semicircular canal to obtain target temporal bone region scan data in which the lateral semicircular canal is parallel to the cross section, wherein the target temporal bone region scan data contains three-dimensional inner ear data; input the target temporal bone region scan data into a pre-trained cochlear segmentation model, so as to segment the target temporal bone region scan data through the cochlear segmentation model to obtain a cochlear segmentation mask, wherein the cochlear segmentation mask contains the cochlea and cochlear peripheral structures;

[0040] An adjustment module, used for adjusting the cochlear posture according to the cochlear segmentation mask;

[0041] A determination module, used for determining the size of the cochlea according to the cochlear posture;

[0042] The acquisition module is also used to determine a cochlear cropping template according to the cochlear size and crop the initial temporal bone region scan data to obtain a view of the cochlea and internal auditory canal;

[0043] An extraction module, for defining and extracting the internal auditory canal structure;

[0044] The adjustment module is further used to rotate the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and the internal auditory canal structure, wherein the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views;

[0045] A calculation module is used to determine whether the cochlear nerve foramen is narrow through the rotated cochlear nerve foramen view and the internal auditory canal structure.

[0046] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the method for processing the cochlear nerve foramen as described in the first aspect above.

[0047] In an embodiment of the present application, initial temporal bone region scan data is acquired, and image data of the lateral semicircular canal is obtained based on the segmentation of the initial temporal bone region scan data; the initial temporal bone region scan data is rotated according to the image data of the lateral semicircular canal to obtain target temporal bone region scan data in which the lateral semicircular canal is parallel to the cross section, and the target temporal bone region scan data includes three-dimensional inner ear data; the target temporal bone region scan data is input into a pre-trained cochlear segmentation model, and the target temporal bone region scan data is segmented by the cochlear segmentation model to obtain a cochlear segmentation mask, and the cochlear segmentation mask includes the cochlear and inner ear data. The cochlear peripheral structure; adjusting the cochlear posture according to the cochlear segmentation mask, and determining the cochlear size according to the cochlear posture, determining the cochlear cropping template according to the cochlear size and cropping the initial temporal bone area scan data to obtain the cochlear and internal auditory canal view; defining and extracting the internal auditory canal structure, and rotating the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar direction corresponding to the cochlear posture to obtain the rotated cochlear nerve foramen view and the internal auditory canal structure, the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views; determining whether the cochlear nerve foramen is narrowed by using the rotated cochlear nerve foramen view and the internal auditory canal structure. The technical solution provided by the present application can quickly determine whether the cochlear nerve foramen is narrowed to improve the diagnostic accuracy and efficiency.

[0048] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 A flowchart of a method for processing a cochlear nerve foramen provided in an embodiment of the present application;

[0051] Figure 2 A schematic diagram of the layered structure of the internal auditory canal provided in an embodiment of the present application;

[0052] Figure 3 A schematic diagram of a slope of change of the internal auditory canal area provided in an embodiment of the present application;

[0053] Figure 4 A schematic diagram of the structure of a cochlear nerve foramen processing device provided in an embodiment of the present application;

[0054] Figure 5 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0055] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0056] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0057] The technical solution of the embodiment of the present application can be applied to multiple key scenarios, including but not limited to:

[0058] Clinical diagnosis and evaluation: In clinical departments such as otolaryngology, neurosurgery, and pediatrics, this method can be used as an auxiliary tool for diagnosing unilateral deafness in children and adults, cochlear nerve dysplasia, and other diseases. By accurately measuring the size of the cochlear nerve foramen, it can assist doctors in determining whether there are problems such as stenosis or atresia of the cochlear nerve foramen, and provide patients with timely and accurate diagnosis and treatment recommendations.

[0059] Screening and research of genetic diseases: In the study of hereditary hearing impairment, this method can help scientists identify gene mutations related to the development of the cochlear nerve and further explore the pathogenesis of hereditary hearing impairment through familial or population screening.

[0060] Surgical planning and postoperative evaluation: For patients who need inner ear surgery (such as cochlear implantation), this method can provide detailed information about the cochlear nerve foramen before surgery, helping surgeons to develop more accurate surgical plans and reduce surgical risks. After surgery, it can also be used to evaluate surgical outcomes and monitor morphological changes in the cochlear nerve foramen.

[0061] Imaging teaching and training: In the field of medical education, this method can be used as a teaching tool to help medical students and young doctors intuitively understand the complex anatomical structure of the cochlea and internal auditory canal, learn the standard process of cochlear nerve foramen measurement, and improve their diagnostic skills.

[0062] Medical equipment development and validation: For companies developing new CT scanning equipment or image processing software, this method can be used as one of the standard processes for performance validation to evaluate the accuracy and efficiency of new technologies in measuring the cochlear nerve foramen.

[0063] In summary, this method not only plays an important role in improving clinical diagnostic accuracy and promoting medical research progress, but also has broad application prospects in medical education, surgical safety and medical technology innovation.

[0064] The inventors have found that currently, the clinical evaluation of the cochlear foramen mainly relies on computed tomography (CT) imaging examinations. However, the technology for determining whether the cochlear foramen is narrow is still in its early stages, and the commonly used manual measurement method is dominant. This process includes image acquisition, identification of anatomical structures, and measurement of distance or area directly on CT images. Doctors or technicians need to rely on their rich experience to identify the cochlear foramen in complex anatomical structures, manually demarcate its boundaries, and then measure its width to determine whether it is narrow.

[0065] Although manual measurement has certain application value in clinical practice, this method has significant limitations and shortcomings:

[0066] Subjectivity and error: Manual measurement is highly dependent on the operator’s experience and judgment. There may be large measurement deviations between different operators, affecting the consistency and accuracy of the measurement results.

[0067] Time-consuming and labor-intensive: The manual measurement process is cumbersome and requires careful operation by professionals, which consumes a lot of time, reduces work efficiency, and limits the development of large-scale clinical applications and research.

[0068] Lack of automation and standardization: With the development of medical imaging technology, automated and standardized measurement methods have become a trend, but the existing manual measurement methods have failed to keep up with this pace of development, which is not conducive to achieving fast and accurate diagnosis and evaluation.

[0069] Limited ability to identify subtle structures: For the diagnosis of subtle structural changes such as cochlear foramen stenosis or atresia, the sensitivity and specificity of manual measurement may be insufficient, which can easily lead to missed diagnosis or misdiagnosis.

[0070] In view of the above defects, it is particularly urgent to explore and develop automated measurement technology for the cochlear nerve foramen, aiming to improve diagnostic efficiency and accuracy and provide a powerful tool for early identification and intervention of diseases such as cochlear nerve dysplasia.

[0071] In view of this, an embodiment of the present application provides a method for processing a cochlear nerve foramen, the method comprising: acquiring initial temporal bone region scan data, and segmenting the initial temporal bone region scan data to obtain image data of the lateral semicircular canal based on the image data of the lateral semicircular canal; rotating the initial temporal bone region scan data according to the image data of the lateral semicircular canal to obtain target temporal bone region scan data in which the lateral semicircular canal is parallel to the cross section, the target temporal bone region scan data containing three-dimensional inner ear data; inputting the target temporal bone region scan data into a pre-trained cochlear segmentation model to segment the target temporal bone region scan data through the cochlear segmentation model to obtain a cochlear segmentation mask, the The cochlear segmentation mask includes the cochlea and the surrounding structures of the cochlea; the cochlear posture is adjusted according to the cochlear segmentation mask, and the cochlear size is determined according to the cochlear posture, the cochlear cropping template is determined according to the cochlear size, and the initial temporal bone area scan data is cropped to obtain the cochlear and internal auditory canal views; the internal auditory canal structure is defined and extracted, and the cochlear nerve foramen view and the internal auditory canal structure are rotated based on the modiolar axis direction corresponding to the cochlear posture to obtain the rotated cochlear nerve foramen view and the internal auditory canal structure, and the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views; and the rotated cochlear nerve foramen view and the internal auditory canal structure are used to determine whether the cochlear nerve foramen is narrow.

[0072] The technical solution provided in this application realizes efficient and automated measurement of the size of the cochlear nerve foramen, thereby improving the measurement accuracy.

[0073] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0074] Figure 1 A flowchart of an embodiment of a method for processing a cochlear nerve foramen provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0075] 101. Acquire initial temporal bone region scan data, and segment and obtain image data of the external semicircular canal based on the initial temporal bone region scan data;

[0076] In this step, the "initial temporal bone area scan data" refers to the original image data of the patient's temporal bone area collected by ultra-high resolution computed tomography (UHRCT) technology. These temporal bone area scan data contain a variety of structural information inside and outside the temporal bone. The "external semicircular canal image data" refers to the image part of the structure of the external semicircular canal separated from the initial temporal bone area scan data by image processing technology. The external semicircular canal is part of the inner ear balance system, and its accurate segmentation is crucial for subsequent image adjustment.

[0077] In the embodiment of the present application, the patient's temporal bone is firstly scanned comprehensively using an advanced UHRCT device to obtain an image sequence containing rich details. Subsequently, professional medical image processing software, such as MITK (Medical Imaging and Communication Toolkit) or 3D Slicer, is used to accurately segment the boundary of the lateral semicircular canal from the initial temporal bone region scan data by manually outlining or applying an automatic segmentation algorithm (such as a model based on deep learning), thereby generating an image data set focused on the lateral semicircular canal.

[0078] 102. Rotate the initial temporal bone region scan data according to the image data of the external semicircular canal to obtain target temporal bone region scan data in which the external semicircular canal is parallel to the cross section, wherein the target temporal bone region scan data includes three-dimensional inner ear data;

[0079] In this step, the goal is to ensure that the lateral semicircular canal remains parallel to the cross-section (i.e., scanning plane) in the final image by rotating the initial temporal bone area scan data in three dimensions. This allows for a more intuitive observation and analysis of the inner ear structures, especially the cochlea. The rotation operation is based on the geometric center or a specific axis of the lateral semicircular canal to achieve an ideal viewing angle.

[0080] In the embodiment of the present application, based on the image data of the lateral semicircular canal obtained in the previous step, the main axis of the lateral semicircular canal is calculated, and the rotation angle and axis required to make it parallel to the cross section of the CT scan are determined. Image processing software (such as ITK-SNAP or a custom Python script combined with numpy and vtk libraries) is used to perform this precise three-dimensional rotation operation, thereby generating new temporal bone area scan data, i.e., target temporal bone area scan data, in which the lateral semicircular canal is clearly parallel to the cross section, while retaining the three-dimensional structural information of the inner ear.

[0081] Target temporal bone region scan data Target temporal bone region scan data Target temporal bone region scan data 103. Input the target temporal bone region scan data into a pre-trained cochlear segmentation model, so as to segment the target temporal bone region scan data by the cochlear segmentation model to obtain a cochlear segmentation mask, wherein the cochlear segmentation mask includes the cochlea and the surrounding structures of the cochlea;

[0082] In this step, the "cochlear segmentation model" refers to an algorithm model based on machine learning or deep learning technology (such as convolutional neural network, CNN), which is trained with a large amount of labeled temporal bone area scan data during the development process and learns to recognize and distinguish the cochlear structure from other adjacent tissues. The "cochlear segmentation mask" is a binary image in which the cochlea and its surrounding structures are marked as white or a specific value (representing the area of ​​interest), while other non-cochlear areas are marked as black or another value (background area). This mask is the basis for subsequent analysis, such as morphological measurement, disease diagnosis, etc.

[0083] In the embodiment of the present application, first, a fully trained and verified cochlear segmentation deep learning model is selected, which may be based on the 3D U-Net architecture because of its excellent performance in medical image segmentation tasks. In the training phase, the model uses a high-quality UHRCT scan image data set from different individuals, different age groups, and various cochlear morphologies and pathological states to ensure the generalization ability and accuracy of the model. During implementation, the target temporal bone area scan data obtained in step 102, that is, the CT scan data that has been adjusted to the optimal viewing angle and contains the stereoscopic inner ear structure, is appropriately preprocessed (such as normalization and denoising) and then sent to the cochlear segmentation model as input. During the operation of the model, the key structures of the cochlea and its surroundings are automatically identified and marked, and a fine cochlear segmentation mask is output. This mask not only accurately outlines the cochlear contour, but also covers the relevant structures around the cochlea, such as blood vessels, ossicular connections, etc., providing an accurate basic layer for subsequent cochlear posture adjustment and accurate measurement of the size of the cochlear nerve hole.

[0084] 104. Adjust the cochlear posture according to the cochlear segmentation mask, determine the cochlear size according to the cochlear posture, determine the cochlear cropping template according to the cochlear size, and crop the initial temporal bone region scan data to obtain a cochlear and internal auditory canal view;

[0085] In this step, the PCA algorithm is required to adjust the cochlear posture, wherein the PCA algorithm (Principal Component Analysis) is a statistical method used to reduce the complexity of multidimensional data sets while retaining the main change trends in the data sets.

[0086] In this context, PCA is used to identify the main directions in the cochlear morphology, and the positions of the apex and base of the cochlea are determined by adjusting the posture of the cochlea to ensure that the modiolus (the main axis of the cochlea) is perpendicular to the horizontal plane. The affine transformation matrix is ​​used to implement linear transformations in space, including operations such as rotation, scaling, and translation. Here, it is used to adjust the posture of the cochlea. Voxels are the basic units in a three-dimensional image, similar to pixels in a two-dimensional image. Each voxel corresponds to a volume element in three-dimensional space. The cochlear and internal auditory canal view refers to an image view that is cropped and posture-adjusted to focus only on the cochlea and its key structures for further analysis.

[0087] In the embodiment of the present application, the specific implementation steps of step 104 may include the following steps:

[0088] 1041. Analyze all the voxel coordinates of the cochlea in the cochlear segmentation mask to determine the coordinates of the cochlear apex region, and adjust the posture of the cochlea based on the coordinates of the cochlear apex region so that the modiolus direction is perpendicular to the horizontal plane;

[0089] During the execution of step 1041, first, all cochlear voxel coordinates are extracted from the cochlear segmentation mask to construct a first array; principal component analysis is performed on the first array to determine the cochlear voxel coordinate with the largest Z value among all cochlear voxel coordinates, and the cochlear voxel coordinate with the largest Z value is determined as the cochlear apex region coordinate.

[0090] That is, extract all voxel coordinates marked as cochlea from the cochlear segmentation mask to form a three-dimensional array containing a large number of coordinate points, namely the first array (array1). Apply the PCA algorithm to analyze all the cochlear voxel coordinates in array1, find the cochlear voxel coordinates with the maximum value in the Z-axis direction (usually representing the depth direction), and define the cochlear voxel coordinates with the largest Z value as the cochlear apex region coordinates. This step can be implemented using the skeleton.decomposition.PCA function.

[0091] Secondly, taking the coordinates of the cochlear apex region as the center, in the cochlear segmentation mask, a region of set length, width and height is cropped to form a second array; principal component analysis is performed on the second array to obtain multiple principal component vectors, and an affine transformation matrix is ​​constructed based on the multiple principal component vectors; the first array is transformed according to the affine transformation matrix to adjust the posture of the cochlea so that the modiolar axis direction is perpendicular to the horizontal plane.

[0092] Specifically, with the determined coordinates of the cochlear apex region as the center, a region with a set length, width and height (for example, a region with a length of 15, a width of 15, and a height of 10) is cropped, and the coordinates of all cochlear voxels in the region form a second array (array2). The second array is subjected to PCA analysis again, and the three principal component vectors generated by this analysis are used to construct an affine transformation matrix (transform_matr i x1), wherein the affine transformation matrix (transform_matr i x1) is used to rotate and calibrate the first array to adjust the posture of the cochlea so that the direction of the modiolar axis is perpendicular to the horizontal plane (xoy plane).

[0093] Among them, the process of performing PCA analysis on the second array again and using the three principal component vectors generated by the analysis to construct an affine transformation matrix (transform_matr ix1) can be: ensure that the second array only contains the part of the cochlear structure of interest, and remove the background and other irrelevant structures. Using a similar method as before, perform PCA analysis on the second array. This time PCA not only helps us understand the main direction of the data, but more importantly, the three principal component vectors obtained after PCA analysis will reveal the extension direction of the internal structure of the cochlea, especially the direction most directly related to the modiolus. Based on these three principal component vectors, especially the vector most related to the modiolus direction, we can construct an affine transformation matrix transform_matr ix1. This usually involves the construction of a rotation matrix to ensure that a certain axis (usually the Z axis) of the transformed data set (mainly the first array, that is, the data of the entire cochlea) is perpendicular to the horizontal plane (xoy plane). This matrix will include the necessary rotation and translation parameters to adjust the cochlea to the predetermined observation angle.

[0094] Among them, the process of using the affine transformation matrix (transform_matr ix1) to rotate and calibrate the first array to adjust the cochlear posture so that the modiolar axis is perpendicular to the horizontal plane (xoy plane) can be: using matrix operations, transform_matr ix1 is applied to each point coordinate in the first array. In Python, the numpy.dot() function or similar matrix multiplication operations can be used to multiply the coordinate vector of each point with the transformation matrix to obtain the transformed new coordinates. After the transformation, it is necessary to check the image or data of the cochlea to confirm whether the modiolar axis has been successfully adjusted to the desired posture, that is, close to perpendicular to the xoy plane. This can be verified by visually inspecting the rotated image or calculating the statistical characteristics of the transformed data. After the cochlear posture is adjusted, the cochlear image or data is ready for subsequent analysis, such as measuring the size of the cochlear nerve foramen.

[0095] Through the above process, not only the display posture of the cochlea is accurately adjusted, but also favorable conditions are created for subsequent precise measurement and analysis, ensuring the accuracy and reliability of the research results.

[0096] 1042. In the cochlear posture, measure the Z-axis distance from the coordinates of the cochlear apex region to the coordinates of the cochlear base region to determine the size of the cochlea, where the size of the cochlea at least includes the height of the cochlea;

[0097] During the execution of step 1042, the cochlear height H can be obtained by finding the coordinate point with the maximum Z value (i.e., the coordinate point of the cochlear apex region) and the coordinate point with the minimum Z value (i.e., the coordinate point of the cochlear base region) in the coordinates of the second principal component analysis (i.e., performing PCA analysis on the second array is the second principal component analysis), and then subtracting the Z values ​​of the two.

[0098] 1043. Determine a cochlear cropping template according to the cochlear height, and crop the initial temporal bone region scan data based on the cochlear cropping template to obtain a view of the cochlea and internal auditory canal.

[0099] During the execution of step 1043, considering the position and size of the cochlear nerve foramen, in order to ensure that all regions of interest (ROIs) are included, a cubic cropping template with a length and width of 50 and a height of the cochlear height H+10 can be created, namely, the cochlear cropping template (array3). It should be noted that the cochlear cropping template defines the region of interest (ROI), which can be understood as the cochlear cropping template defines a three-dimensional spatial area, aiming to accurately cover the cochlea and the cochlear nerve foramen below it.

[0100] Furthermore, before cropping the initial temporal bone region scan data based on the cochlear cropping template to obtain the cochlear and internal auditory canal view, it is also necessary to set the cochlear cropping template with the cochlear apex as the center, and use the inverse transformation matrix (i nv_transform_matr ix1) of the affine transformation matrix (transform_matr ix1) to map the coordinates of the cochlear cropping template to the coordinate system of the initial temporal bone region scan data.

[0101] In other words, the inverse matrix inv_transform_matrix1 is used to transform the coordinates of the cube clipping template from the coordinate system after PCA calibration back to the coordinate system of the initial temporal bone region scan data. The purpose of doing so is to ensure that the clipping operation can be accurately applied to the initial temporal bone region scan data, because it is ultimately necessary to perform the clipping operation on the initial temporal bone region scan data. Optionally, the process of coordinate conversion can be achieved by multi-planar reconstruction (MPR) technology.

[0102] Among them, the process of mapping the coordinates of the cochlear cropping template to the coordinate system of the initial temporal bone region scan data using the inverse transformation matrix (i nv_transform_matr ix1) of the affine transformation matrix (transform_matr ix1) can be: after applying the affine transformation matrix transform_matr ix1 to the first array (i.e., all cochlear voxel coordinates in the cochlear segmentation mask) to adjust the cochlear posture, use linear algebra to calculate the inverse transformation matrix of the affine transformation matrix, i.e., i nv_transform_matr i x1. In Python, numpy.line ginv(transform_matr ix1) can be used to implement it. Next, an inverse transformation is applied to each cochlear voxel coordinate in the cochlear cropping template previously constructed according to the adjusted posture. Optionally, each cochlear voxel coordinate in the cochlear cropping template can be multiplied by i nv_transform_matr ix1 to convert them from the cochlear nerve foramen view coordinates back to the initial temporal bone region scan data. This step is critical as it ensures that any subsequent analysis or measurements are based on the exact position of the cochlea in the coordinate system of the initial temporal bone scan data. That is, after applying the inverse transformation, the coordinates of each cochlear voxel in the cochlear crop template are remapped to the frame of the temporal bone scan data.

[0103] After the coordinates of the cochlear cropping template are mapped to the coordinate system of the initial temporal bone region scan data, the position of the cochlear cropping template in the coordinate system of the initial temporal bone region scan data can be obtained. Further, according to the position, the initial temporal bone region scan data can be cropped based on the cochlear cropping template to obtain a view of the cochlea and internal auditory canal.

[0104] Specifically, the process of “cropping the initial temporal bone region scan data based on the cochlear cropping template to obtain a view of the cochlea and the internal auditory canal” may include the following steps:

[0105] Traverse each voxel in the initial temporal bone region scan data to detect whether the coordinates of each voxel are located in the three-dimensional space area defined by the cochlear cropping template: if the coordinates (x', y', z') of a voxel are located in the three-dimensional space area defined by the cochlear cropping template (i.e., x', y', z' are within the length, width, and height of the cochlear cropping template), then this voxel is considered to belong to the region of interest.

[0106] Determine the CT value corresponding to the voxel located in the three-dimensional space area defined by the cochlear cropping template: for each voxel that falls within the range of the cochlear cropping template, record its CT value. This process can be figuratively understood as a "multiplication" operation, but in actual operation it is more accurately called "selective extraction." That is, it is not a multiplication in the mathematical sense, but a selection based on position - if the coordinates of the voxel meet the conditions (that is, it is located in the three-dimensional space area defined by the cochlear cropping template), then "select" (or retain) the CT value of the voxel.

[0107] Rearrange the CT values ​​of multiple voxels according to the positions of the corresponding voxels in the three-dimensional space area to generate a data set, and visualize the data set to generate a view of the cochlea and internal auditory canal: rearrange all "selected" CT values ​​according to the coordinates of their voxels in the three-dimensional space area to form a data set. This data set only contains CT information around the cochlea and the cochlear foramen, and the data in other areas are ignored or regarded as background. Finally, this data set can be displayed in the form of an image to generate a view of the cochlea and internal auditory canal. 105. Define and extract the internal auditory canal structure, and rotate the cochlear foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain a rotated cochlear foramen view and internal auditory canal structure;

[0108] In this step, defining and extracting the internal auditory canal structure involves using known anatomical information and image processing techniques to separate the key structure of the internal auditory canal from the cochlear segmentation mask. The modiolar axis represents the longitudinal axis of the main body of the cochlea. The cochlear and internal auditory canal views and the internal auditory canal structures are rotated to ensure that the analysis perspective of both is consistent with the natural anatomical posture of the cochlea, which facilitates subsequent accurate measurement and analysis. The affine transformation matrix transform_matr ix3 (different from the affine transformation matrix (transform_matr ix1)) plays a key role in this step, through which specific rotation operations are performed to ensure structural alignment.

[0109] In the embodiment of the present application, step 105 may include the following specific steps:

[0110] 1051. Define and extract the internal auditory canal structure;

[0111] Optionally, in an embodiment of the present application, the process of defining and extracting the internal auditory canal structure may include: determining the X-ray absorption value of the internal auditory canal structure; in the cochlear segmentation mask, determining the voxels whose X-ray absorption value is less than a set first threshold and greater than the first threshold and less than a second threshold as the internal auditory canal structure.

[0112] Specifically, by utilizing different Hu values ​​in the scan data (bone data usually has a specific Hu value (i.e., X-ray absorption value) range), in the cochlear segmentation mask, voxels with X-ray absorption values ​​less than a set first threshold and greater than the first threshold and less than a second threshold (for example, voxels less than 0 and greater than 0 and less than 300) are determined as internal auditory canal structures, thereby accurately cropping out the internal auditory canal structure.

[0113] In this step, the X-ray absorption value refers to the degree of X-ray absorption of each voxel in the CT image, usually expressed in HU (Hounsfield Units). Different tissues show different absorption values ​​according to their density and atomic composition. In the analysis of the cochlea and internal auditory canal, it is very important to use these absorption value differences to define the structure. The first threshold and the second threshold are two absorption value boundaries set to distinguish different types of tissue structures. In this way, the internal auditory canal structure can be accurately identified from the complex cochlear segmentation mask because it has a unique density characteristic, which is usually manifested as an absorption value higher than soft tissue but lower than certain specific bone structures.

[0114] In the embodiment of the present application, the specific implementation steps are as follows:

[0115] Determine the X-ray absorption value range of the internal auditory canal structure: Analyze the known CT image database or literature to determine the typical HU value range of the internal auditory canal in CT images. For example, as a part of the bony structure, the internal auditory canal usually has an absorption value greater than that of soft tissue (such as brain parenchyma, about 30-50HU), but lower than that of compact bone (such as skull, which can reach more than 1000HU). Therefore, the first threshold value may be set at 50HU and the second threshold value at 300HU as the basis for preliminary screening of the internal auditory canal structure.

[0116] Apply thresholds to screen the internal auditory canal structure: Use programming languages ​​such as Python and medical image processing libraries (such as Simple ITK or Python) to load the mask image after cochlear segmentation. Traverse each voxel in the mask and check its corresponding CT value (HU value).

[0117] All voxels are divided into three categories: voxels below the first threshold (such as <50HU) are considered soft tissue or other low-density substances; voxels between the first and second thresholds (such as 50HU≤value<300HU) are most likely internal auditory canal structures based on anatomical knowledge; voxels above the second threshold (≥300HU) represent higher-density bones, etc.

[0118] Only those voxels that meet the second category conditions, that is, voxels with X-ray absorption values ​​less than the set first threshold and greater than the first threshold and less than the second threshold, are retained and marked as internal auditory canal structures. The remaining voxels are ignored in the analysis or marked as background.

[0119] Through the above embodiments, the internal auditory canal structure is effectively defined and extracted, providing a clear image basis for the subsequent accurate measurement of fine structures such as the cochlear nerve foramen. This method utilizes the physical properties of CT imaging, combines medical knowledge with image processing technology, and improves the accuracy and efficiency of structure recognition.

[0120] 1052. Determine the view of the cochlear foramen;

[0121] In this step, the cochlear voxels in the cochlear cropping template are determined by using the cochlear cropping template and the cochlear segmentation mask; and the cochlear voxels are separated from the cochlear and internal auditory canal view to generate a cochlear nerve foramen view.

[0122] Specifically, the cochlear segmentation mask (a binary image in which 1 represents the cochlear region and 0 represents the non-cochlear region) obtained in the previous step is used to screen the voxels in the initial temporal bone region scan data to extract the cochlear region. The specific method is to check the segmentation mask at each voxel position. If the value of the position is 1, the corresponding position in the cochlear cropping template is marked as a cochlear voxel, for example, set to a value of 1. According to the above operation, the voxels in the cochlear cropping template that match the corresponding positions of the cochlear segmentation mask are clearly marked as cochlear voxels (voxel values ​​are set to 1). Furthermore, after the cochlear voxel labeling is completed, the data of the internal auditory canal structure is integrated, and this part of the data is assigned different labels (voxel values ​​are set to 2) to distinguish it from the background (0) and cochlear voxels (1). This is to enable the internal auditory canal structure to be processed separately in the subsequent process. Next, in order to separate the cochlear nerve, further operations are required in the identified cochlear region. Specifically, in the Z-axis direction, according to the cochlear voxel positions marked in the cochlear segmentation mask, the cochlear voxels covering the cochlear nerve foramen are identified, and the values ​​of these voxels are set to 0 (or background value), thereby exposing the potential cochlear nerve structure in this area. Finally, based on the gray value (HU value) in the CT data, a threshold value (less than 0 and greater than 0 and less than 300) is set to further crop and separate the structure, especially by using the spatial information left after cochlear segmentation, through binarization processing (usually with 0 as the dividing point), the area around the cochlear nerve can be effectively extracted to form a cochlear nerve foramen view.

[0123] 1053. Rotate the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and the internal auditory canal structure;

[0124] Optionally, before executing step 1043, it also includes: solving the corresponding inverse transformation matrix according to the affine transformation matrix, and using the corresponding row vector in the inverse transformation matrix as the modiolar axis direction corresponding to the cochlear posture; based on this, step 1043 may include: querying the cross-sectional plane with the most internal auditory canal voxels along the cross-sectional plane through the initial temporal bone area scanning data; determining the rotation angle according to the X component and the Y component in the modiolar axis direction; constructing a rotation matrix around the Z axis based on the rotation angle, and rotating the cochlear voxels in the cochlear nerve foramen view, the cochlear and internal auditory canal view, and the internal auditory canal voxels in the internal auditory canal structure in the cross-sectional plane based on the rotation matrix to obtain the rotated cochlear and internal auditory canal view and internal auditory canal structure.

[0125] In this series of operations, the main purpose is to adjust the direction of the internal auditory canal and cochlear foramen view based on the posture of the cochlea, so as to facilitate a more intuitive and accurate analysis of the cochlear foramen. The following is a detailed explanation of the steps:

[0126] Determine the modiolar orientation: First, calculate the inverse transformation matrix using the affine transformation matrix obtained in the previous step (in some cases, this matrix may be obtained when performing PCA to calibrate the cochlear posture). A row vector in the inverse transformation matrix (usually the row representing the main direction) is used to represent the modiolar orientation corresponding to the cochlear posture. This step provides a reference axis for subsequent rotations.

[0127] Identify the most significant cross-section of the internal auditory canal: In the original temporal bone area scan data, by querying layer by layer along the cross-section, find the layer with the largest number of internal auditory canal voxels. This layer is considered to be the layer that best represents the structure of the internal auditory canal and helps with accurate rotation.

[0128] Calculate the rotation angle: Based on the projection of the modiolus direction on the XY plane (i.e., X component and Y component), determine the angle required for rotation. This angle ensures that after rotation, the structures of the cochlea and internal auditory canal are consistent with our desired viewing direction (such as parallel to the cross-section or a specific anatomical reference line).

[0129] Construct a rotation matrix and rotate: Based on the calculated rotation angle, construct a rotation matrix around the Z axis. Then, use this rotation matrix to rotate the cochlear voxels in the cochlear foramen view and the internal auditory canal voxels in the internal auditory canal structure in the most significant cross-section of the internal auditory canal found. The purpose of the rotation is to align the structure to a standard position that is convenient for analysis, such as aligning the cochlear axis with one of the image axes.

[0130] In an embodiment of the present application, a software tool (such as the numpy library in Python) is used to calculate the inverse matrix from the affine transformation matrix obtained from the previous PCA transformation. The row vector representing the direction of the modiolar axis is selected as (X, Y, Z), where the X and Y components will be used to calculate the rotation angle in the next step. Write a script to traverse all cross-sectional CT slices and calculate the number of internal auditory canal voxels in each layer. Find the layer with the largest number as the reference layer and record its Z coordinate. Calculate the rotation angle θ based on X and Y, for example, using the inverse tangent function arctan(b / a). Based on the θ value, construct a 3x3 rotation matrix R (transform_matr i x3), which has non-zero elements only on the Z axis and is used to rotate around the Z axis. Using this rotation matrix, apply the rotation operation to the cochlear voxels and internal auditory canal voxels at the previously identified reference layer. The specific implementation can be completed by matrix multiplication, that is, multiplying the coordinates (x, y, z) of each voxel by the rotation matrix R to obtain the new coordinates after rotation.

[0131] Through the above steps, we finally obtained the cochlear foramen view and internal auditory canal structure view with optimal arrangement along the modiolar axis, which provided a more intuitive and accurate image basis for the subsequent cochlear foramen size measurement.

[0132] 106. Determine whether the cochlear nerve foramen is narrow by using the rotated cochlear nerve foramen view and the internal auditory canal structure.

[0133] As a possible implementation scheme, step 106 may include the following steps:

[0134] 1061. Determine a minimum ordinate in the rotated cochlear foramen view, and offset the minimum ordinate downward by a preset voxel unit to obtain a target ordinate;

[0135] 1062. In the rotated internal auditory canal structure, determine a plurality of data points whose ordinates are the target ordinates, and determine a maximum abscissa and a minimum abscissa among the plurality of data points to determine a distance between the maximum abscissa and the minimum abscissa;

[0136] 1063. Calculate the size of the cochlear foramen according to the obtained voxel side length and the distance between the maximum horizontal coordinate and the minimum horizontal coordinate.

[0137] 1064. If the size of the cochlear nerve foramen does not meet the set size, determine that the cochlear nerve foramen is narrow.

[0138] In this procedure, a method is described for accurately measuring the size of the cochlear foramen in a rotated view of the cochlea and internal auditory canal by identifying specific cross-sections and analyzing the boundaries of the internal auditory canal structures at that level.

[0139] Minimum ordinate and target ordinate: In three-dimensional image analysis, each data point has three-dimensional coordinates (x, y, z), where the ordinate (y-axis) usually represents the depth direction. Determining the minimum ordinate means finding the most anterior (or uppermost) voxel position in the rotated cochlear foramen view. Based on this minimum ordinate, the preset voxel unit is offset downward in order to locate the cross section close to the expected position of the cochlear foramen, which is defined as the target ordinate.

[0140] Data point screening and horizontal coordinate analysis: On the determined target vertical coordinate, all voxel data points with the vertical coordinate are screened from the internal auditory canal structure data. These data points represent the outline of the internal auditory canal in the cross section. Among these data points, the maximum and minimum values ​​of the horizontal coordinate (x-axis) are found. The distance between these two points indirectly reflects the lateral span of the internal auditory canal in the cross section, which is associated with the width of the cochlear nerve foramen.

[0141] Calculate the size of the cochlear foramen: Using the known voxel side length of the CT scan (i.e., spatial resolution, such as 0.1 mm / voxel), combined with the distance between the maximum and minimum abscissas obtained from the above steps, the actual size of the cochlear foramen can be obtained through simple mathematical calculations, and the unit is usually converted to millimeters (mm).

[0142] In the embodiment of the present application, the specific implementation steps are as follows:

[0143] Determine the target ordinate: In the rotated cochlear foramen view, programmatically traverse the ordinates of all voxels to find the minimum ordinate value y_min. Based on the average position and structural characteristics of the internal auditory canal, assume that the preset voxel unit is 5, and shift 5 voxel units downward to calculate the target ordinate y_target = y_min-5.

[0144] Filter data points and calculate the horizontal coordinate distance: In the rotated internal auditory canal structure data, filter out all voxel data points with the vertical coordinate z_target and record the horizontal coordinates of these data points. This screening process can be easily implemented using Python's NumPy library. Calculate the maximum value max_x and the minimum value min_x in the horizontal coordinates of these data points. The difference between the two is the horizontal span of the internal auditory canal in this cross section.

[0145] Calculate the size of the cochlear foramen: Assuming that the voxel side length is known to be 0.1 mm, the size L of the cochlear foramen (unit: mm) can be calculated by the following formula: L = (max_x-min_x)*0.1 mm.

[0146] Through the above steps, not only the size of the cochlear foramen is accurately measured, but also the error caused by human factors is reduced through the automation method, thereby improving the accuracy and efficiency of the measurement. Further, whether the cochlear foramen is narrow is determined based on the preset size and the measured size of the cochlear foramen. The set size can be set according to historical experience.

[0147] As another possible implementation scheme, step 106 may further include the following steps:

[0148] 1061', in the rotated internal auditory canal structure, dividing the rotated internal auditory canal structure into multiple layers along the Z axis, and counting the voxel area of ​​each layer of the internal auditory canal layer by layer to determine the slope of the change of the internal auditory canal area;

[0149] 1062', determining the height of the internal auditory canal according to the number of divided layers and the obtained thickness of each layer of the internal auditory canal;

[0150] 1063', determine whether the internal auditory canal height meets the set height, if not, execute step 1064';

[0151] 1064', determining whether the internal auditory canal area change slope meets a preset slope, if not, executing step 1065';

[0152] 1065', determine that the cochlear nerve foramen is narrow.

[0153] In the above steps, Z-axis division and voxel area statistics: In three-dimensional space, the Z-axis usually represents the depth direction. Dividing the rotated internal auditory canal structure into multiple layers along the Z-axis means slicing the structure into a series of two-dimensional planes, and the thickness of each layer (layer thickness) is determined by the selected resolution. Voxel area refers to the area of ​​the internal auditory canal on each slice. Counting the voxel area of ​​each layer can help us understand how the shape of the internal auditory canal changes with depth.

[0154] IAM area slope: This refers to the rate at which the IAM area changes with depth (in the z-axis direction). If the IAM area slope is abnormal, it may indicate stenosis or other morphological abnormalities.

[0155] Internal auditory canal height: refers to the total length of the internal auditory canal in the Z-axis direction, which can be calculated by multiplying the number of layers by the thickness of each layer.

[0156] For example Figure 2 As shown, after the rotated internal auditory canal structure is calculated in step 105, the rotated internal auditory canal structure is divided into multiple layers along the Z axis (i.e., the vertical direction), and the voxel area of ​​each layer of the internal auditory canal is counted layer by layer, as shown in FIG. Figure 3 As shown in FIG. 1 , the slope of the change in the internal auditory canal area can be determined by counting the relationship between the number of layers and the number of voxels in each layer of the internal auditory canal. Figure 3 The horizontal axis represents the number of layers, and the vertical axis represents the number of voxels corresponding to the auditory canal in each layer.

[0157] Furthermore, the height of the internal auditory canal needs to be determined, wherein the height of the internal auditory canal = number of layers × layer thickness.

[0158] In the embodiment of the present application, it is first necessary to determine whether the height of the internal auditory canal meets the set height. If it does not meet the set height, it is further determined whether the slope of the change in the area of ​​the internal auditory canal meets the preset slope. If the slope of the change in the area of ​​the internal auditory canal also does not meet the preset slope, it is determined that the cochlear nerve foramen is narrow.

[0159] The set height and preset slope can be set according to historical experience. For example, the set height is 2 mm. If it is less than this value, it is determined that there is an abnormality in the cochlear nerve foramen, and further judgment is required based on the slope of the change in the internal auditory canal area. The preset slope is 37. If it is less than this value, it is determined that the cochlear nerve foramen is narrow.

[0160] Figure 4 A schematic diagram of the structure of a cochlear nerve foramen processing device provided in an embodiment of the present application, as shown in FIG. Figure 4 As shown, the device comprises:

[0161] The acquisition module 21 is used to acquire initial temporal bone region scan data, and obtain image data of the lateral semicircular canal based on the initial temporal bone region scan data; rotate the initial temporal bone region scan data according to the image data of the lateral semicircular canal to obtain target temporal bone region scan data in which the lateral semicircular canal is parallel to the cross section, and the target temporal bone region scan data contains three-dimensional inner ear data; input the target temporal bone region scan data into a pre-trained cochlear segmentation model to segment the target temporal bone region scan data through the cochlear segmentation model to obtain a cochlear segmentation mask, and the cochlear segmentation mask contains the cochlea and the surrounding structures of the cochlea;

[0162] An adjustment module 22, configured to adjust the cochlear posture according to the cochlear segmentation mask;

[0163] A determination module 23, configured to determine the size of the cochlea according to the cochlear posture;

[0164] The acquisition module 21 is further used to determine a cochlear cropping template according to the cochlear size and crop the initial temporal bone region scan data to obtain a view of the cochlea and internal auditory canal;

[0165] An extraction module 24, for defining and extracting the internal auditory canal structure;

[0166] The adjustment module 22 is further used to rotate the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and the internal auditory canal structure, wherein the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views;

[0167] The calculation module 25 is used to determine whether the cochlear nerve foramen is narrowed through the rotated cochlear nerve foramen view and the internal auditory canal structure.

[0168] Figure 4 The cochlear nerve foramen processing device can perform Figure 1 The implementation principle and technical effect of the method for processing the cochlear nerve foramen described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the cochlear nerve foramen processing device in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0169] In one possible design, Figure 4 The processing device of the cochlear nerve foramen of the illustrated embodiment may be implemented as a computing device, such as Figure 5 As shown, the computing device may include a storage component 31 and a processing component 32;

[0170] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component.

[0171] The processing component 32 is used to: obtain initial temporal bone region scan data, and obtain image data of the lateral semicircular canal based on the initial temporal bone region scan data; rotate the initial temporal bone region scan data according to the image data of the lateral semicircular canal to obtain target temporal bone region scan data in which the lateral semicircular canal is parallel to the cross section, and the target temporal bone region scan data contains three-dimensional inner ear data; input the target temporal bone region scan data into a pre-trained cochlear segmentation model to segment the target temporal bone region scan data through the cochlear segmentation model to obtain a cochlear segmentation mask, and the cochlear segmentation mask contains the cochlear segmentation data. and the surrounding structures of the cochlea; adjusting the cochlear posture according to the cochlear segmentation mask, determining the size of the cochlea according to the cochlear posture, determining the cochlear cropping template according to the cochlear size and cropping the initial temporal bone region scan data to obtain a view of the cochlea and internal auditory canal; defining and extracting the internal auditory canal structure, and rotating the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and internal auditory canal structure, the cochlear nerve foramen view being obtained by separating the cochlear and internal auditory canal views; determining whether the cochlear nerve foramen is narrow by using the rotated cochlear nerve foramen view and the internal auditory canal structure.

[0172] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0173] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0174] The display component 33 may be an electroluminescent (EL) element, a liquid crystal display or a micro display having a similar structure, or a retinal direct display or a similar laser scanning display.

[0175] Of course, a computing device may also include other components, such as input / output interfaces, communication components, etc.

[0176] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0177] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0178] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0179] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The method for processing the cochlear nerve foramen of the illustrated embodiment.

[0180] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0181] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0182] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for processing a cochlear nerve foramen, characterized in that: include: Acquire initial temporal bone region scan data, and segment and obtain image data of the external semicircular canal based on the initial temporal bone region scan data; Rotating the initial temporal bone region scan data according to the image data of the external semicircular canal to obtain target temporal bone region scan data in which the external semicircular canal is parallel to the cross section, wherein the target temporal bone region scan data includes three-dimensional inner ear data; Inputting the target temporal bone region scan data into a pre-trained cochlear segmentation model, so as to segment the target temporal bone region scan data by the cochlear segmentation model to obtain a cochlear segmentation mask; adjusting the cochlear posture according to the cochlear segmentation mask, determining the size of the cochlear according to the cochlear posture, determining the cochlear cropping template according to the cochlear size and cropping the initial temporal bone region scan data to obtain a cochlear and internal auditory canal view; Defining and extracting the internal auditory canal structure, and rotating the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and the internal auditory canal structure, wherein the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views; Through the rotated view of the cochlear foramen and the internal auditory canal structure, it can be determined whether the cochlear foramen is narrow.

2. The method according to claim 1, characterized in that The step of adjusting the cochlear posture according to the cochlear segmentation mask, determining the size of the cochlear according to the cochlear posture, determining the cochlear cropping template according to the cochlear size, and cropping the initial temporal bone region scan data to obtain a view of the cochlea and the internal auditory canal includes: Analyzing all the voxel coordinates of the cochlea in the cochlear segmentation mask to determine the coordinates of the cochlear apex region, and adjusting the cochlear posture based on the coordinates of the cochlear apex region so that the modiolus direction is perpendicular to the horizontal plane; In the cochlear posture, measuring the Z-axis distance from the coordinates of the cochlear apex region to the coordinates of the cochlear base region to determine the size of the cochlea, wherein the size of the cochlea at least includes the height of the cochlea; A cochlear cropping template is determined according to the cochlear height, and the initial temporal bone region scan data is cropped based on the cochlear cropping template to obtain a view of the cochlea and internal auditory canal.

3. The method according to claim 2, characterized in that Also includes: Determining the cochlear voxels in the cochlear cropping template by using the cochlear cropping template and the cochlear segmentation mask; The cochlear voxels are separated from the cochlear and internal auditory canal view to generate a cochlear neural foramen view.

4. The method according to claim 1, characterized in that: The method of defining and extracting the internal auditory canal structure includes: Determine the X-ray absorption values ​​of the internal auditory canal structures; In the cochlear segmentation mask, voxels whose X-ray absorption values ​​are smaller than a set first threshold and larger than the first threshold and smaller than a second threshold are determined as internal auditory canal structures.

5. The method according to claim 2, characterized in that: The analyzing all the cochlear voxel coordinates in the cochlear segmentation mask to determine the cochlear apex region coordinates includes: extracting all the cochlear voxel coordinates from the cochlear segmentation mask to construct a first array; performing principal component analysis on the first array to determine the cochlear voxel coordinate with the largest Z value among all the cochlear voxel coordinates, and determining the cochlear voxel coordinate with the largest Z value as the cochlear apex region coordinate; The adjusting the cochlear posture based on the cochlear apex region coordinates comprises: taking the cochlear apex region coordinates as the center, in the cochlear segmentation mask, cutting out a region of set length, width and height to form a second array; performing principal component analysis on the second array to obtain a plurality of principal component vectors, and constructing an affine transformation matrix based on the plurality of principal component vectors; transforming the first array according to the affine transformation matrix to adjust the cochlear posture so that the modiolar axis direction is perpendicular to the horizontal plane; The method of cropping the initial temporal bone region scan data based on the cochlear cropping template to obtain a view of the cochlea and internal auditory canal includes: traversing each voxel in the initial temporal bone region scan data to detect whether the coordinates of each voxel are located in the three-dimensional space region defined by the cochlear cropping template; determining the CT value corresponding to the voxel located in the three-dimensional space region defined by the cochlear cropping template; rearranging the CT values ​​of multiple voxels according to the positions of the corresponding voxels in the three-dimensional space region to generate a data set, and visually displaying the data set to generate a view of the cochlea and internal auditory canal.

6. The method according to claim 5, characterized in that Before rotating the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain the rotated cochlear nerve foramen view and the internal auditory canal structure, the method further includes: solving the corresponding inverse transformation matrix according to the affine transformation matrix, and using the corresponding row vector in the inverse transformation matrix as the modiolar axis direction corresponding to the cochlear posture; The method of rotating the cochlear foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain a rotated cochlear foramen view and the internal auditory canal structure includes: querying the cross-sectional plane with the most internal auditory canal voxels along the cross-sectional plane through the initial temporal bone region scan data; determining the rotation angle according to the X component and the Y component in the modiolar axis direction; constructing a rotation matrix around the Z axis based on the rotation angle, and rotating the cochlear foramen view and the internal auditory canal structure in the cross-sectional plane based on the rotation matrix to obtain a rotated cochlear foramen view and the internal auditory canal structure.

7. The method according to claim 6, characterized in that The method of determining whether the cochlear nerve foramen is narrow by viewing the rotated cochlear nerve foramen and the internal auditory canal structure includes: In the rotated internal auditory canal structure, the rotated internal auditory canal structure is divided into multiple layers along the Z axis, and the voxel area of ​​each layer of the internal auditory canal is counted layer by layer to determine the slope of the change of the internal auditory canal area; Determine the height of the internal auditory canal according to the number of divided layers and the thickness of each layer of the internal auditory canal obtained; Determine whether the internal auditory canal height meets the set height, and if not, determine whether the internal auditory canal area change slope meets the preset slope; If not, the cochlear nerve foramen is determined to be narrow.

8. The method according to claim 6, characterized in that The method of determining whether the cochlear nerve foramen is narrow by viewing the rotated cochlear nerve foramen and the internal auditory canal structure includes: Determine a minimum ordinate in the rotated cochlear foramen view, and offset the minimum ordinate downward by a preset voxel unit to obtain a target ordinate; In the rotated internal auditory canal structure, a plurality of data points whose ordinates are the target ordinates are determined, and a maximum abscissa and a minimum abscissa are determined among the plurality of data points to determine a distance between the maximum abscissa and the minimum abscissa; Calculate the size of the cochlear foramen according to the obtained voxel side length and the distance between the maximum abscissa and the minimum abscissa; If the size of the cochlear nerve foramen does not meet the set size, it is determined that the cochlear nerve foramen is narrow.

9. A device for processing a cochlear nerve foramen, characterized in that: include: An acquisition module is used to acquire initial temporal bone region scan data, and obtain image data of the lateral semicircular canal based on segmentation of the initial temporal bone region scan data; rotate the initial temporal bone region scan data according to the image data of the lateral semicircular canal to obtain target temporal bone region scan data in which the lateral semicircular canal is parallel to the cross section, wherein the target temporal bone region scan data contains three-dimensional inner ear data; input the target temporal bone region scan data into a pre-trained cochlear segmentation model, so as to segment the target temporal bone region scan data through the cochlear segmentation model to obtain a cochlear segmentation mask, wherein the cochlear segmentation mask contains the cochlea and cochlear peripheral structures; An adjustment module, used for adjusting the cochlear posture according to the cochlear segmentation mask; A determination module, used for determining the size of the cochlea according to the cochlear posture; The acquisition module is also used to determine a cochlear cropping template according to the cochlear size and crop the initial temporal bone region scan data to obtain a view of the cochlea and internal auditory canal; An extraction module, for defining and extracting the internal auditory canal structure; The adjustment module is further used to rotate the cochlear nerve foramen view and the internal auditory canal structure based on the modiolar axis direction corresponding to the cochlear posture to obtain a rotated cochlear nerve foramen view and the internal auditory canal structure, wherein the cochlear nerve foramen view is obtained by separating the cochlear and internal auditory canal views; A calculation module is used to determine whether the cochlear nerve foramen is narrow through the rotated cochlear nerve foramen view and the internal auditory canal structure.

10. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the method for processing the cochlear nerve foramen as described in any one of claims 1 to 7.

Images