A method and system for registering mouse brain tissue section images

Abstract

The present disclosure relates to a registration method and system for mouse brain tissue slice images. Based on the two-dimensional mouse brain tissue slice images to be registered and the section parameter prediction model, the spatial positioning parameters are determined to effectively replace the cumbersome three-dimensional spatial search and iterative matching in the prior art based on the one-time prediction of the section parameter prediction model, to improve the matching efficiency and give the robust spatial positioning parameters; based on the spatial positioning parameters, the two-dimensional mouse brain tissue slice images and the three-dimensional average mouse brain template atlas, the deformation mapping information is determined, based on the spatial positioning parameters and the three-dimensional standard atlas, the two-dimensional labeled slice is determined, and based on the two-dimensional labeled slice and the deformation mapping information, the registered mouse brain tissue slice images are determined, to perform subsequent fine alignment processing in the two-dimensional level, which not only greatly reduces the computational complexity and algorithm difficulty, but also forms an automatic workflow of intelligent three-dimensional positioning and efficient two-dimensional registration, thereby providing strong support for high-throughput analysis and clinical application.

Description

Technical Field

[0001] The embodiments in this specification belong to the field of biomedical image registration technology, and specifically relate to a registration method and system for rat brain tissue slice images. Background Technology

[0002] Brain atlas registration is a fundamental and platform-based technique in neuroscience research. It is usually a preliminary step in forming a unified interpretive framework for massive and diverse neurobiological data. Its core value lies in overcoming individual differences in the brain to achieve standardized data integration. Quantitative analysis establishes a bridge between microscopic images and macroscopic interpretation of brain function.

[0003] In neuroscience research, registering experimentally obtained brain tissue slice images with standard brain atlases is a crucial step in achieving data standardization, quantitative analysis, and knowledge integration. However, in actual scientific research, due to the high cost and complex procedures of three-dimensional imaging technologies (such as optical microscopy tomography), researchers more often obtain two-dimensional (2D) histological slice images through experiments. Currently, widely used standard brain atlases mainly employ atlases such as the Allen CCF (Allen Common Coordinate Frame) or Waxholm Space atlas (both high-resolution three-dimensional (3D) anatomical structures).

[0004] Currently, the main methods for registering brain tissue slice images with standard brain atlases are categorized into manual registration-dominated, traditional image registration, and registration prediction using deep learning networks. Manual registration-dominated methods suffer from high subjectivity, poor repeatability, and over-reliance on human experience. Traditional image registration methods are unable to achieve robust registration with complete standard atlases, have low matching efficiency, and exhibit local optima for highly symmetric or repetitive structures. Deep learning network registration prediction methods suffer from low matching efficiency and insufficient robustness. In summary, existing brain atlas registration methods for handling the common problem of matching and registering 2D histological slices to 3D atlases with potential physical defects suffer from one or more technical drawbacks, including low automation, low efficiency, high subjectivity, high image quality requirements, and insufficient robustness. Summary of the Invention

[0005] The embodiments of this disclosure present a registration method and system for rat brain tissue slice images.

[0006] In a first aspect of this disclosure, a registration method for mouse brain tissue slice images is provided. The method includes determining spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the image to be registered and a section parameter prediction model. The method also includes determining deformation mapping information based on the spatial localization parameters, the two-dimensional mouse brain tissue slice image, and a three-dimensional average mouse brain template atlas. Furthermore, the method includes determining two-dimensional labeled slices based on the spatial localization parameters and a three-dimensional standard atlas, and determining the registration of the mouse brain tissue slice image based on the two-dimensional labeled slices and the deformation mapping information.

[0007] In a second aspect of this disclosure, a registration system for mouse brain tissue slice images is provided. The system includes a parameter determination module configured to determine spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the image to be registered and a section parameter prediction model. The system also includes a deformation field determination module configured to determine deformation mapping information based on the spatial localization parameters, the two-dimensional mouse brain tissue slice image, and a three-dimensional average mouse brain template atlas. Furthermore, the system includes a registration image determination module configured to determine two-dimensional labeled slices based on the spatial localization parameters and a three-dimensional standard atlas, and to determine the registration of the mouse brain tissue slice image based on the two-dimensional labeled slices and the deformation mapping information.

[0008] In a third aspect of this disclosure, a computer program product is provided, comprising a computer program that is executed by a processor to implement the method according to the first aspect.

[0009] In a fourth aspect of this disclosure, a machine-readable storage medium is provided. The machine-readable storage medium stores machine-executable instructions, which are executed by a processor to implement the method provided according to a first aspect of this disclosure.

[0010] It should be understood that the description in the Summary of the Invention section is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0011] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:

[0012] Figure 1 A schematic diagram of an example environment in which some embodiments of this disclosure may be implemented is shown;

[0013] Figure 2 A flowchart illustrating a registration method for mouse brain tissue slice images according to some embodiments of this disclosure is shown;

[0014] Figure 3 A schematic diagram of spatial positioning parameters based on three-dimensional space is shown, representing some embodiments of this disclosure.

[0015] Figure 4 A schematic diagram of a symmetrical mouse brain tissue slice image is shown, representing some embodiments of the present disclosure.

[0016] Figure 5 A schematic diagram of a registered mouse brain tissue slice image is shown, representing some embodiments of the present disclosure.

[0017] Figure 6 A block diagram of a registration system for rat brain tissue slice images, according to some embodiments of the present disclosure, is shown.

[0018] Figure 7 A block diagram of an electronic device that can implement several embodiments of the present disclosure is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0020] The terms “comprising” and “having”, and any variations thereof, in this specification, claims, and the foregoing drawings are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. Depending on the context, the word “if” as it applies herein may be interpreted as “when”, “when”, “in response to determination”, or “in response to detection”.

[0021] As mentioned above, current methods for registering brain tissue slice images with standard brain atlases mainly fall into three categories: manual registration-led, traditional image registration, and registration prediction using deep learning networks. Among these, the manual registration-led method involves researchers determining the approximate location of the brain slice (such as the coronal, sagittal, or horizontal plane) and then manually selecting the standard brain atlas that best matches the brain slice for registration. However, a two-dimensional brain tissue slice image may correspond to an infinite number of possible locations and angles in a three-dimensional standard brain atlas, which leads to problems such as being time-consuming, highly subjective, having poor repeatability, and relying too heavily on human experience.

[0022] Traditional image registration methods typically rely on low-level image features, such as finding and matching feature points with scale-invariant feature transformations (SMTs), or using optimization algorithms to maximize mutual information or correlation coefficients between images to calculate spatial transformation parameters. However, this approach faces significant challenges when applied to histological slices. In real histological processing, brain slices inevitably introduce artifacts such as tears, folds, or localized defects. When faced with such incomplete brain slices, registration methods based on intact structural features or statistical information often fail. This results in not only inefficient and unrobust registration with complete standard atlases, but also the potential for local optima (such as similar but incorrect structures) for highly symmetrical or repetitive structures (e.g., the distribution characteristics of the mouse brain).

[0023] Using deep learning networks for registration prediction primarily involves directly predicting transformation parameters between images end-to-end. While this approach has potential, it also has limitations. On one hand, to achieve high accuracy, currently designed deep learning networks are typically very complex, requiring enormous computational resources and lengthy training times, resulting in persistently low registration efficiency. On the other hand, current deep learning networks have limited generalization capabilities. When the incomplete patterns or morphological features of the brain tissue slice images to be registered differ significantly from the training data, the deep learning network may fail to correctly regress reasonable transformation parameters, leading to insufficient robustness.

[0024] Furthermore, even if the current method of registering brain tissue slice images with standard brain atlases finds a general matching location, the subsequent fine registration process often requires complex iterative calculations, making it difficult to achieve a fast and automated workflow.

[0025] Therefore, embodiments of this disclosure propose a registration method for mouse brain tissue slice images. The method includes determining spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the image to be registered and a section parameter prediction model. The method also includes determining deformation mapping information based on the spatial localization parameters, the two-dimensional mouse brain tissue slice image, and a three-dimensional average mouse brain template atlas. Furthermore, the method includes determining two-dimensional labeled slices based on the spatial localization parameters and a three-dimensional standard atlas, and determining the registration of the mouse brain tissue slice image based on the two-dimensional labeled slices and the deformation mapping information.

[0026] This approach effectively replaces the cumbersome 3D spatial search and iterative matching in existing technologies with a one-time prediction based on a pre-trained section parameter prediction model. This not only improves matching efficiency and provides robust spatial localization parameters, but also lays a solid foundation for subsequent nonlinear registration. Furthermore, based on spatial localization parameters, 2D mouse brain tissue slice images, 3D average mouse brain template atlas, and 3D standard atlas, deformation mapping information and 2D labeled slices are obtained. Based on this nonlinear deformation field and 2D labeled slices, the mouse brain tissue slice images are registered for subsequent fine alignment processing at the 2D level. This not only greatly reduces computational complexity and algorithmic difficulty, but also forms an automated workflow of intelligent 3D localization and efficient 2D registration, thus providing strong support for high-throughput analysis and clinical applications.

[0027] Figure 1 Schematic diagrams are shown illustrating example environments in which some embodiments of this disclosure can be implemented. For example... Figure 1 As shown, the example environment 100 may include a data acquisition device 101, which is used to acquire two-dimensional mouse brain tissue slice images to be registered, as well as three-dimensional standard atlases and three-dimensional average mouse brain template atlases as registration references. Here, the data acquisition device 101 can acquire two-dimensional mouse brain tissue slice images acquired by the digital slide scanner after the researchers have sliced ​​the mouse brain tissue. It is understood that the two-dimensional mouse brain tissue slice images may have local slice defects caused by the sampling or slide preparation process, and if local slice defects exist, the image integrity and the accuracy of subsequent data processing can be effectively ensured by repairing the two-dimensional mouse brain tissue slice images.

[0028] In addition, data acquisition device 101 can also obtain the 3D atlas data of the third edition of the Universal Coordinate Framework for the Adult Mouse Brain (AllenCCFv3) released by the Allen Institute for Brain Science through the Allen Institute website. This includes the 3D standard atlas and the 3D average mouse brain template atlas. The 3D standard atlas can be understood as a 3D voxel grid used to describe the average morphological structure of the mouse brain. Each voxel in this 3D voxel grid has a corresponding spatial location and is assigned a unique anatomical structure ID. This anatomical structure ID can be, for example, a unique ID corresponding to any one of the 43 cortical regions, 329 subcortical nuclei, 81 fiber tracts, and 8 ventricular structures. The 3D average mouse brain template atlas can also be understood as a 3D voxel grid used to describe the average morphological structure of the mouse brain. Each voxel in this 3D voxel grid has a corresponding spatial location and grayscale value. It is understandable that the side length of each voxel in the three-dimensional standard map and the three-dimensional average mouse brain template map can be 10 micrometers, and the spatial coordinate system corresponding to the three-dimensional standard map is the same as the spatial coordinate system corresponding to the three-dimensional average mouse brain template map.

[0029] Example environment 100 may further include processing terminal 102, which establishes a communication connection with data acquisition device 101 to acquire two-dimensional mouse brain tissue slice images to be registered, and determines the spatial localization parameters of the two-dimensional mouse brain tissue slice images based on the two-dimensional mouse brain tissue slice images and the section parameter prediction model. Here, the section parameter prediction model can be understood as a pre-trained deep learning prediction network that can effectively extract global and local features of two-dimensional images. By downsampling the two-dimensional mouse brain tissue slice images to adjust their resolution, the downsampled two-dimensional mouse brain tissue slice images are input into the section parameter prediction model, thereby predicting the vector or vector group that is closest to the position of a certain two-dimensional plane in the three-dimensional standard atlas (or three-dimensional average mouse brain template atlas) in the corresponding spatial coordinate system, i.e., the spatial localization parameters.

[0030] Understandably, the spatial positioning parameters may include an origin vector parameter, a width vector parameter, and a height vector parameter. The origin vector parameter is used to define the starting position of the two-dimensional mouse brain tissue slice image to be registered in the spatial coordinate system of the three-dimensional standard atlas. The width vector parameter is used to define the plane width direction and corresponding length of the two-dimensional mouse brain tissue slice image to be registered in the spatial coordinate system of the three-dimensional standard atlas. The height vector parameter is used to define the plane height direction and corresponding length of the two-dimensional mouse brain tissue slice image to be registered in the spatial coordinate system of the three-dimensional standard atlas.

[0031] Furthermore, the processing terminal 102 can also acquire a three-dimensional average mouse brain template atlas through the data acquisition device 101, and determine deformation mapping information based on the three-dimensional average mouse brain template atlas, the aforementioned spatial positioning parameters, and the two-dimensional mouse brain tissue slice image. Here, deformation mapping information can be understood as a vector matrix (i.e., a nonlinear deformation field) used to describe the differences in local anatomical structures between the registered mouse brain tissue slice image and the two-dimensional labeled slice corresponding to the three-dimensional standard atlas. This nonlinear deformation field has the same number of pixels as the two-dimensional mouse brain tissue slice image, and each pixel position has a corresponding displacement vector.

[0032] Furthermore, the processing terminal 102 can determine two-dimensional labeled slices based on spatial positioning parameters and three-dimensional standard atlases, and determine the registration of mouse brain tissue slice images based on the two-dimensional labeled slices and deformation mapping information. Here, the two-dimensional labeled slice can be understood as a planar region extracted from the three-dimensional standard atlas that corresponds to the two-dimensional mouse brain tissue slice image. This planar region has multiple pixel positions and anatomical structure IDs corresponding to each pixel position.

[0033] Understandably, the registered mouse brain tissue slice image can be a two-dimensional atlas that is anatomically aligned with the two-dimensional mouse brain tissue slice image and has high-precision annotations (i.e., anatomical structure IDs). It has multiple pixel positions and the anatomical structure IDs corresponding to each pixel position. When researchers need to query which brain region a certain position in the two-dimensional mouse brain tissue slice image belongs to, they only need to query the anatomical structure ID corresponding to the corresponding pixel position in the registered mouse brain tissue slice image to quickly determine the name of the corresponding brain region.

[0034] This approach effectively replaces the cumbersome 3D spatial search and iterative matching in existing technologies with a one-time prediction based on a pre-trained section parameter prediction model. This not only improves matching efficiency and provides robust spatial localization parameters, but also lays a solid foundation for subsequent nonlinear registration. Furthermore, based on spatial localization parameters, 2D mouse brain tissue slice images, 3D average mouse brain template atlas, and 3D standard atlas, deformation mapping information and 2D labeled slices are obtained. Based on this nonlinear deformation field and 2D labeled slices, the mouse brain tissue slice images are registered for subsequent fine alignment processing at the 2D level. This not only greatly reduces computational complexity and algorithmic difficulty, but also forms an automated workflow of intelligent 3D localization and efficient 2D registration, thus providing strong support for high-throughput analysis and clinical applications.

[0035] It should be understood that the architecture and functionality in example environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of this disclosure. Embodiments of this disclosure can also be applied to other environments with different architectures and / or functionalities.

[0036] Figure 2A flowchart illustrating a registration method for mouse brain tissue slice images according to some embodiments of this disclosure is shown. Method 200 may be, for example, derived from... Figure 1 The processing terminal in the example environment shown executes. For example... Figure 2 As shown in box 202, method 200 can determine the spatial localization parameters of a two-dimensional mouse brain tissue slice image based on the image to be registered and a section parameter prediction model. Here, the processing terminal can... Figure 1 The data acquisition device in the example environment shown establishes a communication connection to acquire two-dimensional mouse brain tissue slice images to be registered. The data acquisition device can acquire two-dimensional mouse brain tissue slice images acquired by the digital slide scanner after the researchers have processed the mouse brain tissue into slices. For example, it can be a single coronal histological slice image of a mouse brain.

[0037] It is understandable that the section parameter prediction model can be understood as a pre-trained deep learning prediction network that can effectively extract global and local features of two-dimensional images. By downsampling the two-dimensional mouse brain tissue slice image to adjust the resolution of the two-dimensional mouse brain tissue slice image, and inputting the downsampled two-dimensional mouse brain tissue slice image into the section parameter prediction model, the model can predict the vector or vector group that is closest to the position of a certain two-dimensional plane in the three-dimensional standard atlas (or three-dimensional average mouse brain template atlas) in the corresponding spatial coordinate system, which is also the spatial localization parameter.

[0038] In addition, the spatial positioning parameters may include an origin vector parameter, a width vector parameter, and a height vector parameter. The origin vector parameter is used to define the starting position of the two-dimensional mouse brain tissue slice image to be registered in the spatial coordinate system of the three-dimensional standard atlas. The width vector parameter is used to define the plane width direction and corresponding length of the two-dimensional mouse brain tissue slice image to be registered in the spatial coordinate system of the three-dimensional standard atlas. The height vector parameter is used to define the plane height direction and corresponding length of the two-dimensional mouse brain tissue slice image to be registered in the spatial coordinate system of the three-dimensional standard atlas.

[0039] Please see Figure 3 The diagram shown illustrates spatial positioning parameters based on three-dimensional space. (See attached diagram.) Figure 3 As shown, the spatial coordinate system established based on the x-axis, y-axis, and z-axis can be represented as the spatial coordinate system of a three-dimensional standard atlas. The origin vector parameter can be represented, for example, as (o x o y o z The width vector parameter can be represented as (u) for example. x u y u z The height vector parameter can be represented as (v) for example.x v y v z ).

[0040] In some implementations, the section parameter prediction model can be an architecture-adjusted deep separable convolutional (Xception) neural network, specifically including an input layer, a backbone network, and an output layer containing a regression head consisting of three fully connected layers. The input layer may include, for example, an initial convolutional layer and three residual separable convolutional modules to downsample the two-dimensional mouse brain tissue slice image (e.g., reducing the dimensionality of the input two-dimensional mouse brain tissue slice image to a standard size of (299, 299)) while increasing the number of channels. The backbone network may use the feature extraction module in the Xception neural network pre-trained on ImageNet as the feature extractor, that is, to ensure that the backbone network has a strong general image feature extraction capability through transfer learning. For example, it may include a core module repeated eight times (composed of depthwise separable convolutional layers and linear residual connection layers) to perform feature extraction processing on the downsampled two-dimensional mouse brain tissue slice image to obtain a high-dimensional feature vector. The output layer may include, for example, a regression head composed of three fully connected layers, a residual separable convolutional layer, a global average pooling layer, and a fully connected layer to perform layer-by-layer linear transformation processing on the high-dimensional feature vector, thereby reducing the dimensionality of the feature vector, and performing regression processing on the high-dimensional feature vector after linear transformation processing to obtain spatial localization parameters.

[0041] The regression head consists of three layers. The first fully connected layer receives the high-dimensional feature vector (e.g., 2048 dimensions) output from the backbone network. After linear transformation (e.g., weighted summation + bias), it is processed using the ReLU activation function for non-linear mapping to obtain a dimensionality-reduced feature vector (e.g., 1024 dimensions). The second fully connected layer performs another linear transformation (e.g., weighted summation + bias) on the dimensionality-reduced feature vector and then uses the ReLU activation function for non-linear mapping to obtain a dimensionality-reduced feature vector (e.g., 512 dimensions). The third fully connected layer performs another linear transformation (e.g., weighted summation + bias) on the dimensionality-reduced feature vector and then uses a linear activation function for linear processing to obtain a dimensionality-reduced feature vector (e.g., 9 dimensions).

[0042] Understandably, when training the section parameter prediction model, the training set may include 10,000 sets of effective localization parameters (each with origin vector parameters, width vector parameters, and height vector parameters) randomly generated based on the 3D average mouse brain template atlas, 2D image slice samples extracted using each effective localization parameter, 5,000 real fluorescent slice samples stained with dapi, and localization parameter labels corresponding to each real fluorescent slice sample. The loss function can be set to calculate the mean square error between the predicted localization parameters and the corresponding effective localization parameters (or localization parameter labels). By predicting the corresponding localization parameters based on each input 2D image slice sample and each real fluorescent slice sample, the loss function calculates the total loss based on the predicted localization parameters and the corresponding effective localization parameters (or localization parameter labels). By calculating the gradient of the total loss with respect to the weights of the section parameter prediction model, the weights of the section parameter prediction model are updated using an optimizer (e.g., using the Adam optimizer with an initial learning rate of 1e-4) to minimize the total loss until the maximum number of training epochs is reached.

[0043] Because researchers often encounter partially damaged or completely horizontally aligned brain slices during sample preparation, making it difficult to find corresponding atlases and thus affecting registration results, the embodiments of this disclosure can leverage the symmetry of the brain to provide a suitable registration scheme for partially damaged and incomplete two-dimensional mouse brain tissue slice images obtained from actual sample preparation. Furthermore, spatial parameters can be used to extract brain slices tilted at any angle, greatly expanding the types of brain slices that can be registered and demonstrating strong versatility.

[0044] In some implementations, when the processing terminal determines the spatial localization parameters of a two-dimensional mouse brain tissue slice image based on the image to be registered and a section parameter prediction model, it can perform downsampling processing on the two-dimensional mouse brain tissue slice image. Here, the two-dimensional mouse brain tissue slice image is typically a high-resolution TIFF format grayscale image. Downsampling processing can sample the two-dimensional mouse brain tissue slice image to a size matching the voxel resolution of the three-dimensional average mouse brain template atlas and the three-dimensional standard atlas. For example, the pixel resolution of the downsampled two-dimensional mouse brain tissue slice image can be 10 micrometers or close to 10 micrometers. Of course, embodiments of this disclosure can also perform other preprocessing on the two-dimensional mouse brain tissue slice image, and are not limited to this.

[0045] Subsequently, the processing terminal can identify the central axis of the downsampled two-dimensional mouse brain tissue slice image and determine the complete lateral slice image based on the number of non-background image pixels on both sides of the central axis. Here, the processing terminal can perform image recognition processing on the downsampled two-dimensional mouse brain tissue slice image to identify the central axis of the downsampled two-dimensional mouse brain tissue slice image and count the number of non-background image pixels in the two-dimensional mouse brain tissue slice images on both sides of the central axis. For example, the number of all pixels with a pixel value greater than 20 can be used as the number of non-background image pixels. It is understood that the image recognition processing process mentioned above is a well-known technique in the field. Of course, researchers can also manually mark the central axis based on the vertical center line of the downsampled two-dimensional mouse brain tissue slice image, which will not be elaborated further here.

[0046] In one example, when determining the complete side slice image based on the number of non-background image pixels on both sides of the central axis, specifically, the number of non-background image pixels on one side can be determined by whether it is less than the product of the number of non-background image pixels on the other side and a preset ratio parameter. It can be understood that the preset ratio parameter can be set to, for example, 80%, that is, judging the relationship between the number of non-background image pixels on one side of the central axis and 80% of the number of non-background image pixels on the other side. When it is detected that the number of non-background image pixels on one side is less than 80% of the number of non-background image pixels on the other side, it indicates that the two-dimensional mouse brain tissue slice image has partial damage and incompleteness, thus requiring integrity restoration processing. When it is detected that the number of non-background image pixels on one side is greater than or equal to 80% of the number of non-background image pixels on the other side, it indicates that the two-dimensional mouse brain tissue slice image does not have partial damage and incompleteness, and therefore no integrity restoration processing is required.

[0047] Next, in response to the determination that the number of non-background image pixels on one side is less than the product of the number of non-background image pixels on the other side and a preset ratio parameter, the two-dimensional mouse brain tissue slice image on the other side of the central axis is identified as a complete side slice image. Here, a complete side slice image can be understood as a two-dimensional mouse brain tissue slice image on the other side of the central axis where the brain slice is relatively intact overall, without any obvious damage or missing parts.

[0048] Subsequently, the processing terminal can perform horizontal mirroring of the complete lateral slice image based on the central axis to obtain a symmetrical mouse brain tissue slice image. This symmetrical mouse brain tissue slice image is then input into the section parameter prediction model to obtain the spatial localization parameters of the two-dimensional mouse brain tissue slice image. In one example, taking the two-dimensional mouse brain tissue slice image to the right of the central axis as the complete lateral slice image, it can be horizontally mirrored based on the central axis. The resulting mirrored mouse brain tissue slice image is then used to cover the two-dimensional mouse brain tissue slice image to the left of the central axis to obtain a symmetrical mouse brain tissue slice image. This provides a high-quality input image for the subsequent accurate prediction of the section parameter prediction model.

[0049] Please see Figure 4 The illustration shows a schematic diagram of a symmetrical mouse brain tissue slice image according to some embodiments of the present disclosure. Figure 4 As shown, the left side shows a two-dimensional mouse brain tissue slice image with some damage and incompleteness (that is, the two-dimensional mouse brain tissue slice image on the left side of the central axis), and the right side shows a symmetrical mouse brain tissue slice image obtained by horizontally mirroring the two-dimensional mouse brain tissue slice image on the right side of the central axis. It can be seen that the symmetrical mouse brain tissue slice image is more complete overall and does not have any obvious damage or incompleteness.

[0050] Understandably, when it is determined that a two-dimensional mouse brain tissue slice image is partially damaged or incomplete, the aforementioned symmetrical mouse brain tissue slice image can be used for subsequent data processing to further ensure the accuracy of the registration results.

[0051] In box 204, method 200 can determine deformation mapping information based on spatial positioning parameters, two-dimensional mouse brain tissue slice images, and a three-dimensional average mouse brain template atlas. Here, the processing terminal can acquire a three-dimensional average mouse brain template atlas through a data acquisition device. This three-dimensional average mouse brain template atlas can be understood as a three-dimensional voxel grid used to describe the average morphological structure of the mouse brain. Each voxel in this three-dimensional voxel grid has a corresponding spatial position and gray value, and the side length of each voxel in this three-dimensional standard atlas is 10 micrometers. Of course, embodiments of this disclosure can also allow the processing terminal to directly obtain the three-dimensional atlas data of the Adult Mouse Brain Universal Coordinate Frame Version 3 (Allen CCFv3) published by the Allen Institute for Brain Science through the Allen Institute website, and are not limited to this.

[0052] It can be understood that deformation mapping information can be interpreted as a vector matrix (i.e., a nonlinear deformation field) used to describe the differences in local anatomical structures between a registered mouse brain tissue slice image and a two-dimensional labeled slice corresponding to a three-dimensional standard atlas. This nonlinear deformation field has the same number of pixels as the two-dimensional mouse brain tissue slice image, and each pixel position has a corresponding displacement vector.

[0053] In some implementations, taking the spatial positioning parameters as having an origin vector parameter, a width vector parameter, and a height vector parameter as an example, when the processing terminal determines deformation mapping information based on the spatial positioning parameters, two-dimensional mouse brain tissue slice images, and three-dimensional average mouse brain template atlas, it can determine the slice size based on the width vector parameter and the height vector parameter. In one example, the width vector parameter is represented as (u x u y u z The height vector parameter is represented as (v x v y v z For example, the width vector parameter and the height vector parameter can be substituted into the slice size calculation formula shown below to determine the slice width and slice height respectively:

[0054]

[0055] In the above formula, W is the slice width and H is the slice height.

[0056] Subsequently, the processing terminal can determine the three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image based on the slice size, the origin vector parameter, the pixel positions in the two-dimensional mouse brain tissue slice image, and the standard size corresponding to the three-dimensional average mouse brain template atlas. In one example, the initial three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image can be determined based on the slice size, the origin vector parameter, and the pixel positions in the two-dimensional mouse brain tissue slice image. Here, the origin vector parameter is represented as (o... x o y o z The pixel positions in a two-dimensional mouse brain tissue slice image are represented as ( , For example, the origin vector parameter, the pixel position in the two-dimensional mouse brain tissue slice image, and the slice size can be substituted into the three-dimensional coordinate calculation formula shown below to obtain the corresponding initial three-dimensional coordinates:

[0057]

[0058]

[0059]

[0060] In the above formula, , and Here are the initial 3D coordinates, W is the slice width, and H is the slice height.

[0061] Next, boundary values ​​can be determined based on the initial three-dimensional coordinates and the standard dimensions corresponding to the three-dimensional average mouse brain template atlas. It is understood that the standard dimensions corresponding to the three-dimensional average mouse brain template atlas have standard lengths for the x-axis, y-axis, and z-axis. When the x-axis coordinate in the initial three-dimensional coordinates is greater than or equal to 0 and less than or equal to the standard length of the x-axis, the y-axis coordinate is greater than or equal to 0 and less than or equal to the standard length of the y-axis, and the z-axis coordinate is greater than or equal to 0 and less than or equal to the standard length of the z-axis, the boundary value of the corresponding initial three-dimensional coordinates can be determined as 1; otherwise, the boundary value of the corresponding initial three-dimensional coordinates can be determined as 0.

[0062] Next, based on the initial three-dimensional coordinates and the corresponding boundary values, the three-dimensional coordinates corresponding to the corresponding pixel positions in the two-dimensional mouse brain tissue slice image can be determined. Here, when the boundary value corresponding to a pixel position in the two-dimensional mouse brain tissue slice image is 1, the corresponding initial three-dimensional coordinates can be determined as the three-dimensional coordinates corresponding to that pixel position; when the boundary value corresponding to a pixel position in the two-dimensional mouse brain tissue slice image is 0, the three-dimensional coordinates corresponding to that pixel position can be determined as (0, 0, 0).

[0063] Subsequently, the processing terminal can process all three-dimensional coordinates and the three-dimensional average mouse brain template image using a trilinear interpolation algorithm to obtain two-dimensional image slices. Here, each three-dimensional coordinate has a corresponding pixel position in the aforementioned two-dimensional mouse brain tissue slice image (or a symmetrical mouse brain tissue slice image obtained after integrity restoration). The trilinear interpolation algorithm can identify eight adjacent three-dimensional coordinates corresponding to each three-dimensional coordinate from the three-dimensional average mouse brain template image, and perform a triple linear weighted average processing on the gray values ​​of these eight adjacent three-dimensional coordinates to obtain the gray value corresponding to the corresponding pixel position. Then, the two-dimensional image slice can be obtained based on all pixel positions and the gray values ​​corresponding to each pixel position.

[0064] Understandably, when performing a triple linear weighted average based on the gray values ​​of eight adjacent 3D coordinates to obtain the gray value corresponding to the corresponding pixel position, the gray values ​​of the eight adjacent 3D coordinates can be linearly interpolated along the x-axis to obtain four new points and their corresponding gray values. Then, the gray values ​​of the four new points can be linearly interpolated along the y-axis to obtain two new points and their corresponding gray values. Finally, the gray values ​​of the two new points obtained again can be linearly interpolated along the z-axis to obtain the gray value corresponding to the corresponding pixel position.

[0065] In this way, the extracted two-dimensional atlas slices are visually smooth and continuous, more closely resembling real anatomical sections, and providing a high-quality reference image with continuous gradient changes for subsequent registration processing.

[0066] Subsequently, the processing terminal can determine deformation mapping information based on the two-dimensional atlas slices and two-dimensional mouse brain tissue slices. Understandably, the two-dimensional atlas slices are used as fixed images, and the two-dimensional mouse brain tissue slices (or symmetrical mouse brain tissue slices obtained through integrity restoration) are used as floating images. A nonlinear registration algorithm is then used to iteratively optimize both the fixed and floating images to obtain deformation mapping information, thereby achieving pixel-level precise alignment.

[0067] In one example, a B-spline-based nonlinear registration algorithm can be used to process two-dimensional atlas slices and two-dimensional mouse brain tissue slices to obtain a nonlinear deformation field, which is then identified as deformation mapping information. This nonlinear deformation field has multiple pixel positions corresponding to the two-dimensional mouse brain tissue slice image and displacement vectors corresponding to each pixel position. Here, the B-spline-based nonlinear registration algorithm can first parameterize the deformation field using a B-spline control point grid (e.g., a 50×50 grid) covering the entire two-dimensional mouse brain tissue slice image. Then, an optimization algorithm (e.g., L-BFGS) is used to iteratively optimize the fixed and floating images. The optimization objective can be set to maximize the normalized mutual information between the fixed and floating images, ultimately resulting in a dense two-dimensional nonlinear deformation field.

[0068] Of course, other nonlinear registration algorithms can also be used to determine the nonlinear deformation field in the embodiments of this disclosure, and other nonlinear registration algorithms and the nonlinear registration algorithm of B splines mentioned above are all well known in the art, and will not be described in detail here.

[0069] In box 206, method 200 can determine two-dimensional labeled slices based on spatial positioning parameters and a three-dimensional standard atlas, and determine registered mouse brain tissue slice images based on the two-dimensional labeled slices and deformation mapping information. Here, the processing terminal can acquire a three-dimensional standard atlas through a data acquisition device. This three-dimensional standard atlas can be understood as a three-dimensional voxel grid used to describe the average morphological structure of the mouse brain. Each voxel in this three-dimensional voxel grid has a corresponding spatial location and is assigned a unique anatomical structure ID, and the side length of each voxel in this three-dimensional standard atlas is 10 micrometers. Of course, embodiments of this disclosure can also allow the processing terminal to directly obtain the three-dimensional atlas data of the Adult Mouse Brain Universal Coordinate Frame Version 3 (Allen CCFv3) published by the Allen Institute for Brain Science through the Allen Institute website, and are not limited to this.

[0070] Understandably, a two-dimensional labeled slice can be a planar region extracted from a three-dimensional standard atlas that corresponds to a two-dimensional mouse brain tissue slice image. This planar region has multiple pixel locations and anatomical structure IDs corresponding to each pixel location. A registered mouse brain tissue slice image can be a two-dimensional atlas that is anatomically aligned with the two-dimensional mouse brain tissue slice image and has high-precision annotations (i.e., anatomical structure IDs). It has multiple pixel locations and anatomical structure IDs corresponding to each pixel location. When researchers need to query which brain region a certain location in a two-dimensional mouse brain tissue slice image belongs to, they only need to query the anatomical structure ID corresponding to the corresponding pixel location in the registered mouse brain tissue slice image to quickly determine the name of the corresponding brain region.

[0071] Of course, embodiments of this disclosure can also map the registered mouse brain tissue slice image onto the two-dimensional mouse brain tissue slice image by upsampling, so as to obtain mouse brain information with high resolution and high magnification, thereby providing strong support for subsequent high-throughput analysis and clinical applications.

[0072] In some implementations, when determining two-dimensional labeled slices based on spatial positioning parameters and three-dimensional standard atlases, the processing terminal can determine the slice size based on the width vector parameter and the height vector parameter. Then, based on the slice size, the origin vector parameter, the pixel positions in the two-dimensional mouse brain tissue slice image, and the standard size corresponding to the three-dimensional average mouse brain template atlas, the three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image can be determined. Then, all three-dimensional coordinates and three-dimensional standard atlases are processed based on the nearest neighbor interpolation algorithm to obtain the two-dimensional labeled slices.

[0073] It is understandable that each three-dimensional coordinate has a corresponding pixel position in the aforementioned two-dimensional mouse brain tissue slice image (or a symmetrical mouse brain tissue slice image obtained after integrity restoration). The nearest neighbor interpolation algorithm can identify eight adjacent three-dimensional coordinates corresponding to each three-dimensional coordinate from the three-dimensional standard atlas. The nearest adjacent three-dimensional coordinate is determined based on the geometric distance between the eight adjacent three-dimensional coordinates and the corresponding three-dimensional coordinates. The anatomical structure ID of the nearest adjacent three-dimensional coordinate in the three-dimensional standard atlas is directly used as the anatomical structure ID corresponding to the corresponding pixel position. Thus, two-dimensional labeled slices can be obtained based on all pixel positions and the anatomical structure IDs corresponding to each pixel position.

[0074] In this way, it is possible to ensure that the IDs of each anatomical structure in the extracted two-dimensional labeled slices are not destroyed and accurately reflect the names of brain regions, and to provide high-quality labeled images with accurate ID definitions for subsequent registration processing.

[0075] In some implementations, when the processing terminal determines the registration of a mouse brain tissue slice image based on the two-dimensional labeled slice and deformation mapping information, it can construct a blank image based on the two-dimensional mouse brain tissue slice image and resample the blank image based on the deformation mapping information (e.g., the aforementioned nonlinear deformation field). It is understood that the blank image has the same size and pixel resolution as the two-dimensional mouse brain tissue slice image (or a symmetrical mouse brain tissue slice image obtained after integrity restoration), that is, it has the same number of pixels and pixel positions as the two-dimensional mouse brain tissue slice image, and each pixel position has a corresponding storage vector (e.g., the same as the planar coordinates of the pixel position).

[0076] In one example, the corresponding pixel positions in the blank image can be resampled based on the displacement vectors corresponding to each pixel position in the nonlinear deformation field. For example, the difference between the displacement vectors corresponding to each pixel position and the corresponding storage vectors can be calculated, and the storage vectors of the corresponding pixel positions in the blank image can be updated based on the difference.

[0077] Subsequently, the processing terminal can process the resampled blank image and the two-dimensional standard slice based on the nearest neighbor interpolation algorithm to obtain a registered mouse brain tissue slice image. It can be understood that the nearest neighbor interpolation algorithm can identify eight neighboring two-dimensional coordinates corresponding to the stored vectors of each pixel position in the resampled blank image from the two-dimensional standard slice. The nearest neighboring two-dimensional coordinates are then determined based on the geometric distance between these eight neighboring two-dimensional coordinates and their corresponding stored vectors. The anatomical structure ID of this nearest neighboring two-dimensional coordinate in the two-dimensional standard slice is directly used as the anatomical structure ID of the corresponding pixel position in the resampled blank image. Therefore, the registered mouse brain tissue slice image can be obtained based on all pixel positions in the resampled blank image and the corresponding anatomical structure IDs.

[0078] Please see Figure 5 A schematic diagram of a registered mouse brain tissue slice image according to some embodiments of the present disclosure is shown. Figure 5 As shown, the left side displays the dapi-stained fluorescent section image to be registered (i.e., a two-dimensional mouse brain tissue section image), and the right side displays the registered mouse brain tissue section image obtained after registration processing. It can be seen that the registered mouse brain tissue section image contains multiple brain regions of the mouse brain, each with a unique anatomical structure ID, and each brain region can be composed of all pixels corresponding to the corresponding anatomical structure ID.

[0079] Figure 6A block diagram of a registration system for rat brain tissue slice images according to some embodiments of the present disclosure is shown. The various embodiments in this specification are described in a progressive manner, with reference to each other for similar or identical parts. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple, and relevant parts can be referred to the description of the method embodiments. Figure 6 As shown, the registration system 600 for mouse brain tissue slice images may include at least a parameter determination module 602, configured to determine the spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the two-dimensional mouse brain tissue slice image to be registered and a section parameter prediction model. The registration system 600 also includes a deformation field determination module 604, configured to determine deformation mapping information based on the spatial localization parameters, the two-dimensional mouse brain tissue slice image, and a three-dimensional average mouse brain template atlas. The registration system 600 further includes a registration image determination module 606, configured to determine two-dimensional labeled slices based on the spatial localization parameters and a three-dimensional standard atlas, and to determine the registered mouse brain tissue slice image based on the two-dimensional labeled slices and the deformation mapping information.

[0080] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0081] Figure 7Block diagrams of electronic devices that can implement various embodiments of the present disclosure are shown. For example... Figure 7 As shown, the electronic device 700 includes a processor 701, which can perform various appropriate actions and processes based on computer program instructions loaded into random access memory (RAM) 703 according to computer program instructions stored in read-only memory (ROM) 702. The RAM 703 may also store various programs and data required for the operation of the electronic device 700. The processor 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0082] The various processes and procedures described above, such as method 200, can be executed by processor 701. For example, in some embodiments, method 200 may be implemented as a software program tangibly contained in a machine-readable medium. In some embodiments, part or all of the software program may be loaded into and / or installed onto electronic device 700 via ROM 702. When the software program is loaded into RAM 703 and executed by processor 701, one or more actions of method 200 described above may be performed.

[0083] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload programmable logic devices (CPLDs), and so on.

[0084] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0085] This disclosure can be a method, apparatus, system, and / or program product. The program product may include a machine-readable storage medium on which machine-readable program instructions for performing various aspects of this disclosure are loaded. The machine-readable program instructions described herein can be downloaded from the machine-readable storage medium to various computing / processing devices, or downloaded via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the machine-readable program instructions from the network and forwards them to the machine-readable storage medium in the respective computing / processing device.

[0086] Machine program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. Machine-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as a standalone software package, partially on a user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the machine-readable program instructions to implement various aspects of this disclosure.

[0087] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. Furthermore, although operations are depicted in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0088] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A registration method for images of mouse brain tissue slices, characterized in that, include: Based on the two-dimensional mouse brain tissue slice image to be registered and the section parameter prediction model, the spatial localization parameters of the two-dimensional mouse brain tissue slice image are determined. Based on the spatial positioning parameters, the two-dimensional mouse brain tissue slice image, and the three-dimensional average mouse brain template atlas, deformation mapping information is determined; as well as Based on the spatial positioning parameters and the three-dimensional standard atlas, two-dimensional labeled slices are determined, and based on the two-dimensional labeled slices and the deformation mapping information, registered mouse brain tissue slice images are determined. The method for determining the spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the two-dimensional mouse brain tissue slice image to be registered and the section parameter prediction model includes: Based on the two-dimensional mouse brain tissue slice image to be registered, the two-dimensional mouse brain tissue slice image is downsampled, and the downsampled two-dimensional mouse brain tissue slice image is matched with the voxel resolution of the three-dimensional average mouse brain template atlas and the three-dimensional standard atlas. The midline of the downsampled two-dimensional mouse brain tissue slice image was identified, and the complete lateral slice image was determined based on the number of non-background image pixels on both sides of the midline; and The complete lateral slice image is horizontally mirrored based on the central axis to obtain a symmetrical mouse brain tissue slice image. The symmetrical mouse brain tissue slice image is then input into the section parameter prediction model to obtain the spatial positioning parameters of the two-dimensional mouse brain tissue slice image.

2. The method according to claim 1, characterized in that, Determining the complete side slice image based on the number of non-background image pixels on both sides of the central axis includes: Based on the number of non-background image pixels on both sides of the central axis, determine whether the number of non-background image pixels on one side is less than the product of the number of non-background image pixels on the other side and a preset ratio parameter; and In response to the determination that the number of non-background image pixels on one side is less than the product of the number of non-background image pixels on the other side and a preset ratio parameter, the two-dimensional mouse brain tissue slice image on the other side of the central axis is determined as a complete side slice image.

3. The method according to any one of claims 1-2, characterized in that, The spatial positioning parameters include an origin vector parameter, a width vector parameter, and a height vector parameter; The determination of deformation mapping information based on the spatial positioning parameters, the two-dimensional mouse brain tissue slice image, and the three-dimensional average mouse brain template atlas includes: The slice size is determined based on the width vector parameter and the height vector parameter; Based on the slice size, the origin vector parameter, the pixel position in the two-dimensional mouse brain tissue slice image, and the standard size corresponding to the three-dimensional average mouse brain template map, the three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image are determined. Based on the trilinear interpolation algorithm, all the three-dimensional coordinates and the three-dimensional average mouse brain template map are processed to obtain two-dimensional map slices; and Based on the two-dimensional atlas slices and the two-dimensional mouse brain tissue slice images, deformation mapping information is determined.

4. The method according to claim 3, characterized in that, The step of determining the three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image based on the slice size, the origin vector parameter, the pixel position in the two-dimensional mouse brain tissue slice image, and the standard size corresponding to the three-dimensional average mouse brain template atlas includes: Based on the slice size, the origin vector parameter, and the pixel position in the two-dimensional mouse brain tissue slice image, determine the initial three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image; Based on the initial three-dimensional coordinates and the standard dimensions corresponding to the three-dimensional average mouse brain template map, boundary values ​​are determined; and Based on the initial three-dimensional coordinates and the corresponding boundary values, the three-dimensional coordinates corresponding to the pixel positions in the two-dimensional mouse brain tissue slice image are determined.

5. The method according to claim 3, characterized in that, The determination of deformation mapping information based on the two-dimensional atlas slices and the two-dimensional mouse brain tissue slice images includes: A nonlinear registration algorithm based on B-splines is used to process the two-dimensional atlas slices and the two-dimensional mouse brain tissue slices to obtain a nonlinear deformation field. This nonlinear deformation field has multiple pixel positions corresponding to the two-dimensional mouse brain tissue slices and a displacement vector corresponding to each pixel position. The nonlinear deformation field is determined as deformation mapping information.

6. The method according to claim 3, characterized in that, The process of determining two-dimensional labeled slices based on the spatial positioning parameters and the three-dimensional standard map includes: The slice size is determined based on the width vector parameter and the height vector parameter; Based on the slice size, the origin vector parameter, the pixel positions in the two-dimensional mouse brain tissue slice image, and the standard size corresponding to the three-dimensional average mouse brain template atlas, determine the three-dimensional coordinates corresponding to each pixel position in the two-dimensional mouse brain tissue slice image; and The nearest neighbor interpolation algorithm is used to process all the three-dimensional coordinates and three-dimensional standard maps to obtain two-dimensional labeled slices.

7. The method according to claim 1, characterized in that, The step of determining the registration of mouse brain tissue slice images based on the two-dimensional labeled slices and the deformation mapping information includes: A blank image is constructed based on the two-dimensional mouse brain tissue slice image, and the blank image is resampled based on the deformation mapping information; and The resampled blank image and the two-dimensional labeled slice are processed using the nearest neighbor interpolation algorithm to obtain a registered mouse brain tissue slice image.

8. The method according to claim 1, characterized in that, The section parameter prediction model includes an input layer, a backbone network, and an output layer containing a regression head composed of three fully connected layers. The input layer is used to downsample the two-dimensional mouse brain tissue slice image. The backbone network is used to extract features from the downsampled two-dimensional mouse brain tissue slice image to obtain a high-dimensional feature vector. The output layer is used to perform layer-by-layer linear transformation on the high-dimensional feature vector and perform regression processing on the high-dimensional feature vector after linear transformation.

9. A registration system for images of mouse brain tissue slices, characterized in that, include: The parameter determination module is configured to determine the spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the two-dimensional mouse brain tissue slice image to be registered and the section parameter prediction model. The deformation field determination module is configured to determine deformation mapping information based on the spatial positioning parameters, the two-dimensional mouse brain tissue slice image, and the three-dimensional average mouse brain template atlas. as well as The registration image determination module is configured to determine two-dimensional labeled slices based on the spatial positioning parameters and the three-dimensional standard atlas, and to determine registered mouse brain tissue slice images based on the two-dimensional labeled slices and the deformation mapping information. The method for determining the spatial localization parameters of the two-dimensional mouse brain tissue slice image based on the two-dimensional mouse brain tissue slice image to be registered and the section parameter prediction model includes: Based on the two-dimensional mouse brain tissue slice image to be registered, the two-dimensional mouse brain tissue slice image is downsampled, and the downsampled two-dimensional mouse brain tissue slice image is matched with the voxel resolution of the three-dimensional average mouse brain template atlas and the three-dimensional standard atlas. The midline of the downsampled two-dimensional mouse brain tissue slice image was identified, and the complete lateral slice image was determined based on the number of non-background image pixels on both sides of the midline; and The complete lateral slice image is horizontally mirrored based on the central axis to obtain a symmetrical mouse brain tissue slice image. The symmetrical mouse brain tissue slice image is then input into the section parameter prediction model to obtain the spatial positioning parameters of the two-dimensional mouse brain tissue slice image.

10. A computer-readable storage medium having a computer program stored thereon, the computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the steps of the method as claimed in any one of claims 1-8.

11. An electronic device, characterized in that, include: One or more processors, and A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method according to any one of claims 1-8.

Images