Automatic-manual cooperative registration and brain area analysis method and system for mouse brain image

Abstract

The invention relates to the technical field of biomedical engineering, in particular to an automatic-manual cooperative registration and brain area analysis method and system for a mouse brain image, and the method comprises the following steps: carrying out the automatic registration of the mouse brain image through an automatic registration tool, and obtaining an automatic registration result; performing manual registration on the registered brain atlas in the individual brain space; in the manual registration process, a napari-based manual correction method is used, brain region boundary adjustment is conducted on an automatic registration result, then automatic registration is conducted on a brain region template obtained after manual correction again, and a final deformation field of the complete registration process is obtained; and carrying out coordinate transformation by using a deformation field obtained by registration, transforming cell or axon information in the brain of the individual into a standard brain space, and then carrying out brain region analysis. According to the method, the automatic registration result is manually corrected through smearing type correction, the automatic registration result can be corrected more easily in combination with information of the real mouse brain, and complete processing is carried out in cooperation with a follow-up analysis method.

Description

Technical Field

[0001] The present application relates to the field of biomedical engineering technology, and in particular to a method and system for automatic-manual collaborative registration and brain region analysis of mouse brain images. Background Art

[0002] The fundamental principle of brain atlas registration is to build a bridge between different individual brain images or data and a common standard space (i.e., the space represented by the brain atlas). This bridge enables all individual data to be compared and analyzed in the same coordinate system, thus overcoming research obstacles caused by individual differences.

[0003] Choosing the right transformation method is crucial in achieving registration. Among them, linear transformations such as translation, rotation, and scaling are common methods. Translation adjusts the image's position in space, rotation changes its orientation, and scaling adjusts its size. These simple transformations can be useful for images with relatively regular shapes and small differences. However, due to the complex individual differences in mouse brains, relying solely on linear transformations is insufficient for precise registration. Therefore, non-rigid deformation methods are often used for mouse brain image registration. Non-rigid deformations overcome the limitations of linear transformations and can flexibly adjust the shape of local areas of the image. This property allows them to precisely match subtle differences in brain structure between individuals, ensuring a high degree of alignment between the spatial position and shape of individual brain images and the standard atlas. This greatly improves the accuracy and effectiveness of registration, providing a solid and reliable data foundation for subsequent neuroscience research based on unified standards.

[0004] In recent years, significant progress has been made in registration technology, resulting in the development of numerous tools for mouse brain registration, such as mBrainAligner, brainreg, and visor_reconstruction. However, automatic registration using these tools often results in discrepancies, necessitating manual corrections, often referred to as manual registration. Existing automatic registration tools are not always equipped with a corresponding manual registration tool, making manual corrections impossible during manual registration. This results in an inability to accurately fine-tune brain region boundaries, compromising the accuracy and precision of brain region delineation. Summary of the Invention

[0005] The embodiments of the present application provide a method for automatic-manual collaborative registration and brain region analysis of mouse brain images. Manual registration is achieved using simple functions, and the automatic and manual registration steps are then linked in series. Taking data acquired through VISoR imaging as an example, the automatic registration results are manually corrected using an automatic registration tool. This eliminates issues such as format mismatch and missing information, making it easier to use the registered information and to fully process it with subsequent analysis methods.

[0006] To solve the above technical problems, an embodiment of the present application provides a method for automatic-manual collaborative registration and brain region analysis of mouse brain images, comprising the following steps: first, using an automatic registration tool to automatically register the mouse brain image to obtain an automatic registration result; then, manually registering the registered brain map in the individual brain space; finally, using the deformation field obtained by the registration to perform coordinate transformation, transforming the cell or axon information in the individual brain into the standard brain space, and then performing brain region analysis; wherein the manual registration process includes: first, using a manual correction method based on napari to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process.

[0007] In some exemplary embodiments, an automatic registration tool is used to automatically register the mouse brain image to obtain an automatic registration result, including: using the automatic registration tool, selecting a registration template, and automatically registering the pre-processed mouse brain image to obtain an automatic registration result.

[0008] In some exemplary embodiments, the automatic registration tool is visor_reconstruction.

[0009] In some exemplary embodiments, during the manual registration process, the boundaries of brain regions are adjusted based on the automatic registration results, including: after the automatic registration of the mouse brain image is completed, the brain region template registered to the individual brain is imported into the napari software; based on the structural information of the original mouse brain image, the brightness and transparency of the original image and the brain region template to be corrected after automatic registration are compared, the deviations in the automatic registration results are adjusted, and the brain regions of interest are screened through two-dimensional and three-dimensional visualization to assist in manual correction; the adjustment includes: manually correcting the brain region results after automatic registration by extracting the brain region ID and smearing or erasing.

[0010] In some exemplary embodiments, the manually corrected brain region template is automatically registered again, including: using the manually corrected brain region template as the fixed image required for registration, and the standard brain region template as the moving image, and automatically registering the manually corrected brain region template and the standard brain region template again through the algorithm of the automatic registration link, so that a new deformation field is formed between the manually corrected brain region template and the standard brain region template. After the automatic registration is completed, the final deformation field of the complete registration process is obtained.

[0011] In some exemplary embodiments, the deformation field obtained by registration is used to perform coordinate transformation, and the cell or axon information in the individual brain is transformed into the standard brain space, and then brain region analysis is performed, including: using different brain regions represented by different values ​​in the brain region template as the basis for division, and locating the brain region by finding the corresponding values ​​of the cell points in the standard brain map; calculating the three-dimensional Euclidean distance between the two points and interpolating according to the order of parent and child nodes, dividing the distance between the two points into multiple line segments, calculating the brain regions where the interpolation points are located, and attributing the corresponding line segment lengths to the brain regions where the interpolation points are located, and performing brain region statistics of axons.

[0012] In some exemplary embodiments, during the brain region positioning of cells, based on the brain region template and the structural information file in the registration link, a one-to-one correspondence between the value and the brain region is achieved by querying the value of a certain point in the standard brain space in the brain region template, and the corresponding result between the required coordinate point and the brain region where it is located is obtained, thereby positioning the brain region; in the structural information file, each brain region has a unique ID, and different IDs are presented in the brain region template with different values.

[0013] In some exemplary embodiments, during the statistical analysis of axons in different brain regions, axonal structural points of interest are selected according to the parent-child relationship, and for each parent-child node pair, the three-dimensional Euclidean distance between the two points is calculated and the points are interpolated.

[0014] In some exemplary embodiments, points are interpolated with a step size of 1 μm, and the distance between two points is divided into multiple line segments; the length from the parent node to the last interpolation point, and the length from the last interpolation point to the child node are calculated in sequence; the ownership of the 1 μm axon length is determined by the brain region where the current node is located, so as to complete the statistics of the total axon length and the axon length of each brain region.

[0015] In the second aspect, the embodiments of the present application also provide an automatic-manual collaborative registration and brain region analysis system for mouse brain images, which uses the automatic-manual collaborative registration and brain region analysis method for mouse brain images described in the above embodiments to perform automatic-manual collaborative registration and brain region analysis, including: an automatic registration module, a manual registration module and a brain region analysis module connected in sequence; the automatic registration module is used to use an automatic registration tool to automatically register the mouse brain image to obtain an automatic registration result; the manual registration module is used to manually register the registered brain map in the individual brain space; wherein, the manual registration process includes: first using a napari-based manual correction method to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process; the brain region analysis module is used to use the deformation field obtained by the registration to perform coordinate transformation, transform the cell or axon information in the individual brain into the standard brain space, and then perform brain region analysis.

[0016] The technical solution provided by the embodiments of the present application has at least the following advantages:

[0017] An embodiment of the present application provides a method for automatic-manual collaborative registration and brain region analysis of mouse brain images, which includes the following steps: first, using an automatic registration tool to automatically register the mouse brain image to obtain an automatic registration result; then, manually registering the registered brain map in the individual brain space; finally, using the deformation field obtained by the registration to perform coordinate transformation, transforming the cell or axon information in the individual brain into the standard brain space, and then performing brain region analysis; wherein the manual registration process includes: using a manual correction method based on napari to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process.

[0018] This application aims to propose a method of automatic-manual collaborative registration to achieve fast and convenient registration of mouse brain images and fine correction of regions of interest. After registration, a method of brain region analysis is proposed to achieve brain region positioning of various cells such as neurons in the mouse brain and brain region statistics of axon length. Through this application, the intermediate steps from obtaining mouse brain images to obtaining neuronal information are finally completed, a new registration idea is proposed, and the process of mouse brain registration and data analysis is standardized. Take VISoR (high-throughput three-dimensional fluorescence microscopy) imaging as an example, but other imaging methods are also applicable. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] One or more embodiments are exemplarily described by the pictures in the corresponding drawings. These exemplifications do not constitute limitations on the embodiments. Unless otherwise stated, the pictures in the drawings do not constitute proportional limitations.

[0020] Figure 1 A flowchart of a method for automatic-manual co-registration and brain region analysis of mouse brain images provided in one embodiment of the present application.

[0021] Figure 2 A flowchart of the automatic-manual collaborative correction strategy provided in one embodiment of the present application.

[0022] Figure 3 A schematic diagram of the results of manual correction based on napari provided in one embodiment of the present application.

[0023] Figure 4 A comparison diagram before and after manual correction provided in an embodiment of the present application.

[0024] Figure 5 This is a two-dimensional and three-dimensional visualization diagram of a brain region of interest provided in one embodiment of the present application.

[0025] Figure 6 This is a schematic diagram of the principle of accumulating axon length in different brain regions provided in one embodiment of the present application.

[0026] Figure 7 A schematic structural diagram of a system for automatic-manual collaborative registration and brain region analysis of mouse brain images provided in one embodiment of the present application. DETAILED DESCRIPTION

[0027] As can be seen from the background technology, existing automatic registration tools are not all equipped with matching manual registration tools. There is a technical problem that manual correction cannot be made during manual registration, resulting in the inability to accurately fine-tune the boundaries of brain areas, thereby affecting the accuracy and precision of brain area division.

[0028] There are various methods for mouse brain registration, but automatic registration using registration tools often results in certain deviations, necessitating manual correction, a process known as manual registration. Different manual registration methods have varying advantages and disadvantages. mBrainAligner, a commonly used registration software, automatically aligns individual brain images with the Allen CCFv3 (Allen Mouse Brain Common Coordinate Framework version 3) standard brain atlas and provides visualization of the registration results. This software, to a certain extent, meets the basic requirements of existing research for brain atlas registration. However, in-depth analysis reveals that this software still has room for improvement in terms of accuracy and efficiency. From the perspective of imaging modality, mouse brain images acquired using different imaging techniques exhibit structural differences. VISoR imaging technology is characterized by minimal data deformation, while the fMOST fusion template used by mBrainAligner exhibits relatively large deformation. Although sufficient research has yet to fully clarify the differences in mBrainAligner's adaptability to imaging data with varying degrees of deformation, the deformation characteristics of imaging data theoretically influence the registration performance, and its adaptability to data from different imaging modalities warrants further investigation. From a practical application perspective, mBrainAligner's operational workflow is relatively complex, involving the configuration and adjustment of multiple parameters, requiring a high level of user expertise. Furthermore, the time required to complete a complete registration task is long, significantly reducing research efficiency in large-scale data processing scenarios. Furthermore, while the process of registering individual brain images to a standard brain atlas achieves data standardization, facilitating group-level comparison and analysis, it inevitably compromises direct observation and analysis of original neuronal morphology and trajectories. For example, it is impossible to determine the cortical layer to which a neuron belongs based on its actual location and the structural characteristics of the original mouse brain, making it impossible to determine whether the registered neuron is accurately positioned. Although some post-processing techniques can restore or preserve some of the original information to a certain extent, the risk of information loss compared to the original data remains, hindering in-depth studies of the fine structure of neurons, connectivity patterns, and the integrity of neural pathways.

[0029] mBrainAligner's companion manual registration tool is called Semi3D. The core workflow of Semi3D is to generate feature points from point clouds. Point cloud data, as a discrete representation of three-dimensional spatial data, contains a large amount of 3D coordinate information and can accurately describe the surface morphology of an object. After generating feature points, Semi3D performs drag operations on the point cloud to implement specific functions. This operation has a localized effect on specific surrounding areas in 3D space, with varying degrees of impact depending on factors such as drag force and direction, and the spatial relationship between point clouds. Based on this varying degree of impact, the tool performs image contour matching on the boundary point clouds to better determine the boundaries and morphology of brain regions. However, Semi3D has also exposed some significant limitations in practical applications. First, the tool is highly dependent on the automatic registration results of mBrainAligner and has strict requirements on the data format, making it unsuitable for standalone use. Second, the tool's operation is relatively complex and prone to lag when operating on more detailed brain region localization results. This not only affects research efficiency but also creates a poor user experience for researchers. In addition, Semi3D cannot perform fine movement of adjacent boundary point clouds within the same brain region. If you try to adjust the boundary point cloud on one side, it will often cause the other side to move in conjunction, resulting in the inability to accurately fine-tune the brain region boundary, thereby affecting the accuracy and precision of brain region division and positioning.

[0030] BrainGlobe's registration tool, brainreg, integrates visualization into the registration process, greatly facilitating the observation of registration results. Developed as a plugin for the napari multidimensional image viewer, brainreg leverages the existing scientific Python ecosystem and aims to lower the technical barriers to brain atlas registration, enabling researchers to perform this work without extensive computational expertise. It offers a convenient data loading method, supporting drag-and-drop import of raw mouse brain data. It also allows for one-click registration of raw mouse brains with a variety of brain atlases accessed through the BrainGlobe Atlas API. During the registration process, brainreg first performs resampling and filtering preprocessing on the loaded atlas and raw data. It then achieves precise alignment through affine transformations and nonlinear free-form deformations. The resulting registration allows visualization of the brain atlas overlaid on the raw data and also maps the raw data into the atlas space, facilitating data analysis in a unified coordinate space. This also allows for the study of individual brain structures and morphology within the raw data space. Registration takes less than ten minutes, significantly improving the efficiency of mouse brain registration.

[0031] However, brainreg is not perfect and has limitations when processing data acquired with certain specialized imaging techniques. Specifically, brainreg is sensitive to the quality of the original image. Problems such as background noise, blur, or insufficient contrast in the original image can directly impact the accuracy and reliability of registration. Because the registration algorithm relies on image feature information for matching, low-quality images can lead to inaccurate feature extraction, affecting the final registration results. Furthermore, brainreg lacks a manual correction function.

[0032] Our laboratory uses the already developed visor_reconstruction automated registration tool. Initially, individual brains undergo preprocessing, including resampling and cropping, to adapt the reconstructed 4μm whole-brain images to the 25μm resolution of the mouse brain atlas. The subsequent registration process combines a nonrigid B-spline transform with multiresolution registration to achieve more precise image alignment.

[0033] Freesia, a software currently under development, performs deformation operations on mouse brain images based on a standard atlas. By concatenating the registration results from visor_reconstruction, Freesia achieves deformation operations from point A to point B in the mouse brain image. The registration results from visor_reconstruction provide a basic framework for Freesia, enabling the software to accurately locate each point in the image within a standard coordinate system. Based on this, Freesia guides image deformation by adding deformation reference points, updating the deformation field of the entire brain through non-rigid deformation. However, Freesia is still in its developmental stages, and practical applications still require resolving several issues. Primarily, data format compatibility is difficult to achieve with diverse data formats, and integration with upstream and downstream operations presents significant challenges. This requires researchers to expend additional effort and time on data format conversion when using the software. This not only increases workload and limits overall research efficiency and consistency, but can also lead to data loss or corruption during the conversion process, potentially affecting subsequent image analysis results.

[0034] In order to solve the above technical problems, the embodiment of the present application provides a method and system for automatic-manual co-registration and brain region analysis of mouse brain images, which includes the following steps: first, using an automatic registration tool to automatically register the mouse brain image to obtain an automatic registration result; then, manually registering the registered brain map in the individual brain space; finally, using the deformation field obtained by the registration to perform coordinate transformation, transforming the cell or axon information in the individual brain into the standard brain space, and then performing brain region analysis; wherein, the manual registration process includes: using a manual correction method based on napari to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process. The present application aims to propose a method for automatic-manual co-registration to achieve fast and convenient registration of mouse brain images and fine correction of the region of interest. After registration, a method for brain region analysis is proposed to achieve brain region positioning of various cells such as neurons in the mouse brain and brain region statistics of axon length. This application ultimately completes the intermediate steps from acquiring mouse brain images to obtaining neuronal information, proposes a new alignment approach, and standardizes the process of mouse brain alignment and data analysis. VISoR (high-throughput three-dimensional fluorescence microscopy) imaging is used as an example, but other imaging methods are also applicable.

[0035] The following detailed description of the various embodiments of the present application is provided in conjunction with the accompanying drawings. However, those skilled in the art will appreciate that many technical details are provided in the various embodiments of the present application to facilitate a better understanding of the present application. However, even without these technical details and the various variations and modifications based on the following embodiments, the technical solutions claimed in the present application can still be implemented.

[0036] See Figure 1 The present invention provides a method for automatic-manual co-registration and brain region analysis of mouse brain images, comprising the following steps:

[0037] Step S1: Automatically register the mouse brain image using an automatic registration tool to obtain an automatic registration result.

[0038] Step S2: Manually register the registered brain atlas in the individual brain space.

[0039] Step S3: Using the deformation field obtained by registration to perform coordinate transformation, the cell or axon information in the individual brain is transformed into the standard brain space, and then brain region analysis is performed.

[0040] The manual registration process in step S2 includes: using a manual correction method based on napari to adjust the brain region boundaries of the automatic registration results, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process.

[0041] In some embodiments, the automatic registration tool is visor_reconstruction.

[0042] It should be noted that this application uses the existing automatic registration tool visor_reconstruction for automatic registration.

[0043] The purpose of this application is to use simple functions to achieve manual registration, and then connect the two links of automatic registration. Taking the data obtained by VISoR imaging as an example, the existing automatic registration tool visor_reconstruction is used to manually correct the automatic registration results. This avoids problems such as format mismatch and missing information, making it easier to use the information obtained from the registration and complete processing with subsequent analysis methods.

[0044] In some embodiments, an automatic registration tool is used to automatically register the mouse brain image to obtain an automatic registration result, including: using the automatic registration tool, selecting a suitable registration template, and automatically registering the preprocessed mouse brain image to obtain an automatic registration result.

[0045] After the automatic registration is completed, the registered brain map in the individual brain space is manually registered (also called semi-automatic interactive correction). This application uses the image editing and annotation plug-in SegmentAnnotation Plugin based on the software napari to manually correct the brain area results after automatic registration by extracting the brain area ID and smearing and erasing. During the correction process, the researchers adjusted the brightness and transparency of the original image and the brain area template to be modified based on the structural information of the original mouse brain image and their own professional knowledge, and adjusted the parts with deviations in the registration results. For example, for some areas with blurred or mismatched brain area boundaries, researchers can manually draw accurate boundaries to ensure more accurate division of brain areas. The researchers adjusted the brain area boundaries and other aspects of the automatic registration results, and then performed a second automatic registration on the manually corrected brain area template to obtain the final deformation field of the complete registration process, which is convenient for subsequent analysis.

[0046] Finally, brain region analysis is performed. This is divided into brain region positioning of cells and brain region statistics of axons. Using "different brain regions represented by different values ​​in the brain region template" as the basis for division, brain regions are classified by finding the corresponding value of the cell point in the brain atlas after registration. The brain region statistics of axons are calculated according to the order of parent and child nodes, and the three-dimensional Euclidean distance between the two points is interpolated. The distance between the two points is divided into multiple line segments, and the brain region where the interpolation point is located is calculated separately, and the corresponding line segment length is attributed to the brain region.

[0047] like Figure 2As shown, this application uses a manual correction method based on napari, which allows researchers to adjust the brain region boundaries of the automatic registration results, and then automatically align the manually corrected brain region template again to obtain the final deformation field of the complete registration process, which is convenient for subsequent analysis.

[0048] Specifically, manual registration includes manual correction and the process of obtaining the deformation field through secondary automatic registration.

[0049] In some embodiments, during the manual registration process, the brain region boundaries of the automatic registration results are adjusted, including: after the automatic registration of the mouse brain image is completed, the brain region template registered to the individual brain is imported into the napari software; based on the structural information of the original mouse brain image, the brightness and transparency of the original image and the brain region template to be corrected after automatic registration are compared, the deviations in the automatic registration results are adjusted, and the brain regions of interest are screened through two-dimensional and three-dimensional visualization to assist in manual correction; the adjustment includes: manually correcting the brain region results after automatic registration by extracting the brain region ID and smearing or erasing.

[0050] Specifically, after the automatic registration is completed, the brain region template registered to the individual brain is imported into the napari software. Researchers can use the napari-based image editing and annotation plug-in Segment AnnotationPlugin developed by the laboratory to manually correct the brain region results after automatic registration by extracting the brain region ID and smearing or erasing it. The results are as follows: Figure 3 During the correction process, the researchers adjusted the brightness and transparency of the original image and the template of the brain region to be modified based on the structural information of the original mouse brain image and their own professional knowledge, and adjusted the deviations in the registration results. For example, for some areas with blurred or mismatched brain region boundaries, researchers can manually draw accurate boundaries to ensure more accurate division of brain regions. The comparison before and after manual correction is shown in the figure below. Figure 4 As shown, it should be noted that the red box is the corrected area and the red dotted line marks the cortical boundary. In addition, the brain area of ​​interest can also be screened through two-dimensional and three-dimensional visualization, such as Figure 5 As shown in , it is another indicator to judge the accuracy of the registration results. Figure 5 The left image in the middle shows a 2D coronal section of the VISp5 brain region, and the right image shows a 3D schematic of the VISp5 and VISa5 brain regions. The light blue region represents the VISp5 brain region, and the purple region represents the VISa5 brain region.

[0051] By manually correcting the automatically registered brain region template according to the above operation, a more accurate individual brain region template can be obtained, which provides a feasible operation for improving the accuracy of the data of the local region of interest.

[0052] To obtain the coordinate correspondence between the two spaces, a secondary automatic registration strategy was developed. This involves reusing the automatic registration tool to register the manually corrected brain region template with the standard brain region template to obtain the deformation field relationship and complete the coordinate mapping between the two spaces. In other words, after automatically registering the brain region atlas deformed into the individual brain space, this atlas can be imported into napari, compared with the original image, and smear-style corrections can be performed using a plug-in. The specific correction process can be assisted by screening brain regions of interest. Finally, the modified brain region atlas replaces the data file required by the registration tool, and a secondary automatic registration is performed. The final result is the deformation relationship between the two spaces, which is the deformation field information. This allows points in the individual brain to be located in the standard brain space, while the information in the brain map can be used to locate the corresponding brain region.

[0053] In some embodiments, the manually corrected brain area template is automatically registered again, including: using the manually corrected brain area template as the fixed image required for registration, and the standard brain area template as the moving image, and through the algorithm of the automatic registration link, the manually corrected brain area template and the standard brain area template are automatically registered again, so that a new deformation field is formed between the manually corrected brain area template and the standard brain area template. After the automatic registration is completed, the final deformation field of the complete registration process is obtained.

[0054] Specifically, considering the data format and image characteristics of the manually corrected brain template, the materials required for the original automatic registration need to be replaced before the second automatic registration. That is, the manually corrected brain template is used as the fixed image required for registration, and the standard brain template is used as the moving image. The two are automatically registered through the algorithm of the automatic registration link. By aligning the two brain templates, the inconsistency of features between the anatomical image and the brain image is avoided. It also avoids the unrecognizable features caused by too little change in the brain template before and after manual correction, which can mistakenly be considered to be completely identical before and after manual correction, and ultimately lead to invalid registration failure.

[0055] Secondary automatic registration creates a new deformation field between the manually corrected brain region template and the standard brain template. This is the deformation relationship between the individual brain and the standard brain after the manual registration process, thereby helping researchers obtain new coordinate mapping results.

[0056] In some embodiments, the deformation field obtained by registration is used to perform coordinate transformation, and the cell or axon information in the individual brain is transformed into the standard brain space, and then brain region analysis is performed, including: performing coordinate transformation to transform the individual brain coordinates into the standard brain space; using different brain regions represented by different values ​​in the brain region template as the basis for division, and locating the brain region by finding the corresponding values ​​of the cell points in the standard brain map; calculating the three-dimensional Euclidean distance between the two points and interpolating according to the order of parent and child nodes, dividing the distance between the two points into multiple line segments, calculating the brain regions where the interpolation points are located, and attributing the corresponding line segment lengths to the brain regions where the interpolation points are located, and performing brain region statistics of axons.

[0057] In some embodiments, during the brain region positioning process of cells, based on the brain region template and the structural information file of the registration link, the value of a certain point in the standard brain space in the brain region template is queried to achieve a one-to-one correspondence between the value value and the brain region, and the corresponding result between the required coordinate point and the brain region where it is located is obtained, thereby positioning the brain region; in the structural information file, each brain region has a unique id, and different ids are presented in the brain region template with different value values.

[0058] Specifically, the implementation of brain region localization relies primarily on the information provided by the brain region template and the structural information file during the registration phase. In the structural information file, each brain region has a unique ID, and different IDs are represented by different values ​​in the brain region template. Therefore, by querying the value of a point in the standard brain space in the brain region template, a one-to-one correspondence between that value and the brain region is established, resulting in a correspondence between the desired coordinate point and its corresponding brain region.

[0059] In some embodiments, during the process of counting axons in different brain regions, axonal structural points of interest are selected according to the parent-child relationship, and for each parent-child node pair, the three-dimensional Euclidean distance between the two points is calculated and the points are interpolated.

[0060] For the function of neuron length statistics, this application selects the axon and other structural points of interest according to the parent-child relationship, and for each parent-child node pair, calculates the three-dimensional Euclidean distance between the two points (the unit of this distance follows the coordinate system unit, which is usually μm in this process).

[0061] In some embodiments, the interpolation of points is performed with a step size of 1 μm, and the distance between two points is divided into multiple line segments; the length from the parent node to the last interpolation point, and the length from the last interpolation point to the child node are calculated in sequence; since 1 μm is usually the smallest unit of interest in the process, the axon length of 1 μm can be determined by the brain region where the current node is located, so as to complete the statistics of the total axon length and the axon length of each brain region. For example, Figure 6As shown, the three-dimensional Euclidean distance between the parent node A and the child node B is 3.8μm. Starting from point A, insert three points P1, P2, and P3 in the direction of the connecting line at a distance of 1μm. Determine the brain regions where A, P1, P2, and P3 are located, and add 1μm, 1μm, 1μm, and 0.8μm to their respective brain regions.

[0062] See Figure 7 , the embodiment of the present application also provides an automatic-manual collaborative registration and brain region analysis system for mouse brain images, which uses the automatic-manual collaborative registration and brain region analysis method for mouse brain images described in the above embodiment to perform automatic-manual collaborative registration and brain region analysis, including: an automatic registration module 101, a manual registration module 102 and a brain region analysis module 103 connected in sequence; the automatic registration module 101 is used to use an automatic registration tool to automatically register the mouse brain image to obtain an automatic registration result; the manual registration module 102 is used to manually register the registered brain map in the individual brain space; wherein, the manual registration process includes: using a napari-based manual correction method to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process; the brain region analysis module 103 is used to use the deformation field obtained by the registration to perform coordinate transformation, transform the cell or axon information in the individual brain into the standard brain space, and then perform brain region analysis.

[0063] The method for automatic-manual collaborative registration and brain region analysis of mouse brain images provided in this application fully realizes the process from pre-registration of mouse brain results to analyzed data. It aims to use simple functions to achieve manual registration, and then connect the two links of automatic registration and manual registration in series. Taking the data obtained by VISoR imaging as an example, the existing automatic registration tool visor_reconstruction is used to manually correct the automatic registration result. Specifically, the idea of ​​manual registration is to smear the brain region template after registration, and then automatically register it again to obtain the deformation field. This application uses the Segment Annotation Plugin, an image editing and annotation plug-in based on the software napari, to manually correct the brain region results after automatic registration by extracting the brain region ID and smearing or erasing it.

[0064] Compared with the existing technology, the automatic-manual co-registration and brain region analysis method for mouse brain images provided by this application has the following advantages:

[0065] (1) The input required by the manual registration link has a low degree of dependence on automatic registration. This means that not only can the visor_reconstruction automatic registration tool be used, but as long as a brain region template that can be registered to the individual brain can be obtained, this manual registration method can be used. Therefore, other automatic registration algorithms can be used, which is scalable.

[0066] (2) The operation idea is simple. The brain area template is modified only by smearing. There is no complicated point cloud computing and deformation field calculation, which saves computing resources and reduces the occurrence of lag.

[0067] The automatic-manual collaborative registration and brain region analysis method for mouse brain images provided in this application has been verified to be feasible through experimental simulation. In the specific hypothalamus research topic, it has been observed that the individual brain has some relatively obvious boundaries in the near-ventral nucleus structure, and the distance between the ventral nucleus structure such as VMH and the hypothalamus boundary is very small, which is an area that can be manually reviewed. The registration results near the left and right hemispheres and the third ventricle can also be used as the focus of manual review. In addition, the cell points that are automatically classified as brain area 0 (i.e., not belonging to any brain area) are worth checking to determine the cause of the non-hypothalamic distribution. Therefore, manual registration can be performed by manually correcting the near-ventral nucleus and hypothalamus boundary, adjusting the left and right hemisphere boundaries, and paying attention to the position of non-hypothalamic cells.

[0068] After semi-automatic interactive correction, the number of cells in each brain region was re-counted. Compared with the data results of automatic registration, the proportion of cells in the ARH brain region decreased from 13.1% to 12.8%, while the proportion of cells in the Hypo brain region increased from 45% to 46%. Although the change is relatively slight, this change means that the initial automatic registration defines some hypothalamic cells that do not belong to the ARH brain region as ARH brain region cells, but there should actually be more cells distributed in the undefined hypothalamic region. After manual correction, only 4.2% of the brain region 0 cells were corrected. It was verified that most of them were cells in areas with large deviations from the existing brain region templates, which means that directly deleting all brain region 0 cell points will only have an impact of less than 5% on the results.

[0069] It should be noted that this automatic registration tool is not limited to visor_reconstruction, an automatic registration tool for VISoR images, but can also be used with different imaging methods. Because manual registration has a low reliance on automatic registration, it can be replaced with other automatic registration tools suitable for different modalities and use the same secondary automatic registration concept, but the data needs to be replaced based on the new automatic registration tool. This concept can be used for registration and analysis not only of mouse brains, but also of other animal brains.

[0070] Based on the above technical solution, the embodiment of the present application provides a method and system for automatic-manual collaborative registration and brain region analysis of mouse brain images, which includes the following steps: first, using an automatic registration tool to automatically register the mouse brain image to obtain an automatic registration result; then, manually registering the registered brain map in the individual brain space; finally, using the deformation field obtained by the registration to perform coordinate transformation, transforming the cell or axon information in the individual brain into the standard brain space, and then performing brain region analysis; wherein the manual correction process includes: using a napari-based manual correction method to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process.

[0071] This application aims to propose a method of automatic-manual collaborative registration to achieve fast and convenient registration of mouse brain images and fine correction of regions of interest. After registration, a method of brain region analysis is proposed to achieve brain region positioning of various cells such as neurons in the mouse brain and brain region statistics of axon length. Through this application, the intermediate steps from obtaining mouse brain images to obtaining neuronal information are finally completed, a new registration idea is proposed, and the process of mouse brain registration and data analysis is standardized. Take VISoR (high-throughput three-dimensional fluorescence microscopy) imaging as an example, but other imaging methods are also applicable.

[0072] Those skilled in the art will appreciate that the above-described embodiments are specific examples for implementing the present application, and that in actual applications, various changes in form and detail may be made thereto without departing from the spirit and scope of the present application. Any person skilled in the art may make changes and modifications without departing from the spirit and scope of the present application. Therefore, the scope of protection of the present application shall be subject to the scope defined in the claims.

Claims

1. A method for automatic-manual co-registration and brain region analysis of mouse brain images, characterized in that: The following steps are involved: Automatic registration tools are used to automatically register the mouse brain images and obtain the automatic registration results; Manual registration of the registered posterior brain atlas within individual brain space; The deformation field obtained by registration is used to perform coordinate transformation, transforming the cell or axon information in the individual brain into the standard brain space, and then performing brain region analysis; The manual registration process includes: firstly, using a manual correction method based on napari to adjust the brain region boundary of the automatic registration result, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process.

2. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 1, characterized in that: Automatic registration tools were used to automatically register the mouse brain images and the automatic registration results were obtained, including: Automatic registration tools were used to select registration templates and automatically register the pre-processed mouse brain images to obtain automatic registration results.

3. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 1, characterized in that: The automatic registration tool is visor_reconstruction.

4. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 1, characterized in that: During manual registration, the brain region boundaries are adjusted based on the automatic registration results, including: After completing the automatic registration of the mouse brain images, the brain region template registered to the individual brain was imported into the napari software; Based on the structural information of the original mouse brain image, the brightness and transparency of the original image and the brain region template to be corrected after automatic registration are compared. The deviations in the automatic registration results are adjusted, and the brain regions of interest are screened through 2D and 3D visualization to assist in manual correction. The adjustment includes: manually correcting the brain region result after automatic registration by extracting the brain region ID and smearing or erasing.

5. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 1, characterized in that: Automatic registration is performed again on the manually corrected brain region template, including: The manually corrected brain region template is used as the fixed image required for alignment, and the standard brain region template is used as the moving image. Through the algorithm of the automatic alignment link, the manually corrected brain region template and the standard brain region template are automatically aligned again, so that a new deformation field is created between the manually corrected brain region template and the standard brain region template. After the automatic alignment is completed, the final deformation field of the complete alignment process is obtained.

6. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 1, characterized in that: The deformation field obtained by registration is used to perform coordinate transformation, transforming the cell or axon information in the individual brain into the standard brain space, and then performing brain region analysis, including: The different brain regions represented by different values ​​in the brain region template are used as the basis for division, and the brain regions are located by finding the corresponding values ​​of cell points in the standard brain atlas; According to the order of parent-child nodes, the three-dimensional Euclidean distance between two points is calculated and interpolated. The distance between the two points is divided into multiple line segments. The brain regions where the interpolation points are located are calculated respectively, and the corresponding line segment length is attributed to the brain region where the interpolation point is located, and the axon statistics are performed by brain region.

7. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 6, characterized in that: In the process of brain region localization, based on the brain region template and structure information file in the registration phase, the value of a certain point in the standard brain space in the brain region template is queried to achieve a one-to-one correspondence between the value and the brain region, and the correspondence between the required coordinate point and the brain region in which it is located is obtained, thereby performing brain region localization; In the structure information file, each brain region has a unique ID, and different IDs are presented with different values ​​in the brain region template.

8. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 6, characterized in that: In the process of counting axons in different brain regions, the axonal structural points of interest are selected according to the parent-child relationship. For each parent-child node pair, the three-dimensional Euclidean distance between the two points is calculated and the points are interpolated.

9. The method for automatic-manual co-registration and brain region analysis of mouse brain images according to claim 8, characterized in that: Interpolate the points with a step size of 1 μm and divide the distance between two points into multiple line segments; The length from the parent node to the last interpolation point, as well as the length from the last interpolation point to the child node, are calculated in sequence. The ownership of the 1 μm axon length is determined by the brain region where the current node is located, so as to complete the statistics of the total axon length and the axon length of each brain region.

10. A system for automatic-manual co-registration and brain region analysis of mouse brain images, which uses the method for automatic-manual co-registration and brain region analysis of mouse brain images according to any one of claims 1 to 9 to perform automatic-manual co-registration and brain region analysis, characterized in that: include: An automatic registration module, a manual registration module, and a brain region analysis module connected in sequence; The automatic registration module is used to automatically register the mouse brain image using an automatic registration tool to obtain an automatic registration result; The manual registration module is used to manually register the registered brain atlas within the individual brain space; wherein the manual registration process includes: first, using a manual correction method based on napari to adjust the brain region boundaries of the automatic registration results, and then automatically registering the manually corrected brain region template again to obtain the final deformation field of the complete registration process; The brain region analysis module is used to perform coordinate transformation using the deformation field obtained by registration, transform the cell or axon information in the individual brain into the standard brain space, and then perform brain region analysis.

Images