Image processing methods for brain structural images and three-dimensional neural tract models

By converting DWI diffuse images and T1 structural images to three-dimensional image space, obtaining the nerve bundle model and outputting it to DICOM format, the problem of inability to fuse the three-dimensional model in the existing technology is solved, and the introduction of the three-dimensional nerve bundle model and neurosurgery navigation are realized System application.

CN115101178BActive Publication Date: 2024-11-22SHANGHAI AIJIN ARTIFICIAL INTELLIGENCE TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210860063.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-20
Publication Date
2024-11-22
Estimated Expiration
2042-07-20

AI Technical Summary

Technical Problem

The existing neurosurgical navigation system can only use DICOM data and cannot integrate three-dimensional models reconstructed by other systems into the navigation system, limiting the application of surgical navigation.

Method used

By converting DWI diffuse images and T1 structural images to three-dimensional image space, a nerve bundle model is obtained, and the rotation matrix is ​​converted to DICOM format, and brain structural images with nerve bundle information are output to realize the import of the three-dimensional nerve bundle model.

Benefits of technology

The three-dimensional nerve bundle model is implemented to introduce the neurosurgical navigation system, providing brain structural images with nerve bundle information, and enhancing the application capabilities of the navigation system and the availability of data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115101178B_ABST
    Figure CN115101178B_ABST
Patent Text Reader

Abstract

The present invention discloses an image processing method for a brain structural image and a three-dimensional neural bundle model, which comprises the following steps: converting a DWI diffusion image and a T1 structural image into a three-dimensional image space respectively; obtaining a neural bundle model in the three-dimensional image space; converting the DWI diffusion image, the T1 structural image and the neural bundle model in the three-dimensional image space into a three-dimensional coordinate space respectively through a rotation matrix; superimposing the DWI diffusion image, the T1 structural image and the neural bundle model in the three-dimensional coordinate space; converting the neural bundle model into the T1 structural image space; constructing a T1 structural image containing neural bundle information based on the superimposed voxels of the T1 structural image and the neural bundle model; and outputting the T1 structural image containing neural bundle information into a DICOM format. The present invention achieves the technical effect of importing a three-dimensional neural bundle model into a neurosurgery navigation system for application by integrating the brain structural image with the three-dimensional neural bundle model and outputting the brain structural image with the neural bundle information into a DICOM format.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of image processing technology, and in particular relates to an image processing method for brain structure images and three-dimensional neural bundle models. Background Art

[0002] Image three-dimensional modeling is an image processing technology that is widely used in surgical medical applications. It creates a three-dimensional model of the area of ​​interest through image processing or image cutting of traditional two-dimensional medical images, helping surgical medical staff to view diseases from a three-dimensional spatial perspective. Existing image three-dimensional modeling is generally a built-in system of the neurosurgical navigation system. Based on software maintenance and safety considerations, the existing neurosurgical navigation system can only use DICOM data as the input data source, and can only reconstruct the model through its own image processing system and three-dimensional modeling system for use in surgical navigation. Therefore, it is impossible to integrate the three-dimensional model reconstructed by other systems into the navigation system. Therefore, how to develop a new image processing method to overcome the above-mentioned defects in the prior art is a direction that technicians in this field need to study. Summary of the invention

[0003] The purpose of the present invention is to provide an image processing method for brain structure images and three-dimensional neural bundle models, which can integrate brain structure images with three-dimensional neural bundle models, and output brain structure images with neural bundle information in DICOM format, thereby achieving the technical effect of importing three-dimensional neural bundle models into neurosurgery navigation systems for application.

[0004] The technical solutions adopted are as follows:

[0005] A method for processing brain structure images and three-dimensional neural bundle models, comprising the following steps:

[0006] Step 1: Convert the DWI diffusion image and T1 structural image into three-dimensional image space respectively;

[0007] Step 2: Obtain a neural bundle model in a three-dimensional image space;

[0008] Step 3: The DWI diffusion image, T1 structural image and nerve bundle model in the three-dimensional image space are transformed into the three-dimensional coordinate space through the rotation matrix;

[0009] Step 4: Overlay the DWI diffusion image, T1 structural image and nerve bundle model in the three-dimensional coordinate space;

[0010] Step 5: Use the inverse rotation matrix of the T1 structural image to transform the neural bundle model into the T1 structural image space;

[0011] Step 6: constructing a T1 structural image containing nerve bundle information by overlapping voxels of the T1 structural image and the nerve bundle model in the T1 structural image space;

[0012] Step 7: Export the T1 structural image containing the nerve bundle information into DICOM format.

[0013] Preferably, in the above image processing method, the DWI diffusion image and T1 structural image in step 1 are both in DICOM format, and both are converted into a three-dimensional image space in NIfTI format via dcm2niix.

[0014] Preferably, in the above image processing method, step 2 comprises:

[0015] Step 21: Register the DWI diffusion image with the T1 structural image;

[0016] Step 22: Obtaining a nerve bundle model based on DWI diffusion image processing;

[0017] More preferably, in the above image processing method: in step 21, the DWI diffusion image and the T1 structural image are registered using the ANTs library.

[0018] More preferably, in the above image processing method: the step 22 comprises:

[0019] Step 221: performing DTI model fitting on the DWI diffusion image to obtain the maximum eigenvector V1 and anisotropic index FA of the ellipsoid model;

[0020] Step 222: Assume that the tracking termination parameter FA threshold is ≤ 0.2 and the turning angle is ≥ 60°, and perform whole-brain neural tracking based on the maximum eigenvector V1 and the anisotropic index FA;

[0021] Step 223: intercept the neural bundle according to the region of interest to obtain a neural bundle model.

[0022] Preferably, in the above image processing method: the rotation matrix in step 3 is a 3x4 matrix.

[0023] Preferably, in the above image processing method, step 6 comprises:

[0024] Step 61: Overlaying the T1 structural image with the T1 spatial image of the nerve bundle model;

[0025] Step 62: Replace the overlapping voxels of the T1 structural image and the T1 spatial image of the nerve bundle model with the highlighted values ​​to form a T1 structural image containing nerve bundle information.

[0026] More preferably, in the above image processing method: the calculation process of the highlight value in step 62 is: float(max_image_value)=ceil((float)temp_max_image_value / 100)*100+100;

[0027] Among them, max_image_value is the highlight value, which is saved and calculated as single-precision floating-point data;

[0028] temp_max_image_value is the maximum value of the T1 structure image, which is saved and calculated as single-precision floating-point data.

[0029] Preferably, in the above-mentioned image processing method: step 7 of outputting the T1 structural image with nerve bundle information into DICOM format comprises:

[0030] Step 70: Generate a reading code according to the original DICOM file;

[0031] Step 71: Read the DICOM image according to the read code to obtain the DICOM image matrix and slice position information;

[0032] Step 72: Determine whether the DICOM format is LPS encoding;

[0033] Step 73: Determine whether the scan section type is a sagittal plane, a coronal plane, or an axial plane;

[0034] Step 74: Determine whether the image needs to be mirror-inverted;

[0035] Step 75: rearrange the DICOM image matrix according to the sorting;

[0036] Step 76: read the T1 structural image with neural information and obtain the NIfTI image matrix;

[0037] Step 77: Adjust the NIfTI image matrix order to be consistent with DICOM;

[0038] Step 78: Fill the NIfTI image matrix into the DICOM image matrix according to the section position information;

[0039] Step 79: Output the DICOM image according to the reading code of the DICOM format.

[0040] Compared with the prior art, the image generated by the present invention is in DICOM format, which is suitable for importing into the existing neurosurgery navigation system; the image carries T1 structural image information, which is conducive to matching with other images after importing into the system; the image also carries highlighted nerve bundle model information, which can be used to extract highlighted nerve bundles and provide three-dimensional modeling applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0042] Figure 1 It is a schematic diagram of the process of Example 1;

[0043] Figure 2 This is a schematic diagram of the transformation of step 1 in Example 1;

[0044] Figure 3 This is a schematic diagram of the transformation of step 2 in Example 1;

[0045] Figure 4 This is a schematic diagram of the transformation of step 3 in Example 1;

[0046] Figure 5 This is a schematic diagram of the transformation of step 4 in Example 1;

[0047] Figure 6 This is a schematic diagram of the transformation of steps 5-6 in Example 1;

[0048] Figure 7 This is a schematic diagram of the transformation of step 7 in Example 1. DETAILED DESCRIPTION

[0049] In order to more clearly illustrate the technical solution of the present invention, further description will be given below in conjunction with various embodiments.

[0050] like Figure 1-7 Shown as Example 1:

[0051] A method for processing brain structure images and three-dimensional neural bundle models, comprising the following steps:

[0052] Step 1: Use the dcm2niix library to convert the DWI diffusion image and T1 structural image in DICOM format into the three-dimensional image space in NIfTI format;

[0053] DICOM images are saved in the form of one DICOM image per layer of image. T1 structural images are 192 layers of 256*256 images, and DWI diffusion images are 30 groups of 70 layers of 128*128 images. Therefore, directly reading T1 structural images requires reading 192 DICOMs, while reading DWI diffusion images requires reading 2100 DICOMs. This step converts DICOM data into NIfTI format through the dcm2niix library. During the image processing process, only one set of images needs to be read for subsequent image analysis and processing. This improves the efficiency of image reading and reduces the risk of reading data errors and data transmission.

[0054] Step 2: Obtain a neural bundle model in a three-dimensional image space, including:

[0055] Step 21: Use the ANTs library to register the DWI diffusion image with the T1 structural image;

[0056] Step 22: Obtain a neural bundle model based on DWI diffusion image processing; specifically:

[0057] Step 221: performing DTI (diffusion tensor imaging) model fitting on the DWI diffusion image to obtain the maximum eigenvector (first eigen vector) V1 and the fractional anisotropy (Fractional anisotropy) FA of the ellipsoid model;

[0058] Step 222: Assume that the tracking termination parameter FA threshold is ≤ 0.2 and the turning angle is ≥ 60°, and perform whole-brain neural tracking based on the maximum eigenvector V1 and the anisotropic index FA;

[0059] Step 223: intercept the neural bundle according to the region of interest to obtain a neural bundle model.

[0060] Step 3: The DWI diffusion image, T1 structural image and nerve bundle model in the three-dimensional image space are transformed into the three-dimensional coordinate space through a 3x4 rotation matrix;

[0061] Step 4: Overlay the DWI diffusion image, T1 structural image and nerve bundle model in the three-dimensional coordinate space;

[0062] Step 5: Use the inverse rotation matrix of the T1 structural image to transform the neural bundle model into the T1 structural image space;

[0063] Step 6: In the T1 structural image space, the T1 structural image containing the neural bundle information is constructed by overlapping the voxels of the T1 structural image and the neural bundle model; specifically:

[0064] Step 61: Overlaying the T1 structural image with the T1 spatial image of the nerve bundle model;

[0065] Step 62: Replace the overlapping voxels of the T1 structural image and the T1 spatial image of the nerve bundle model with the highlighted values ​​to form a T1 structural image containing nerve bundle information.

[0066] The calculation process of the highlight value in step 62 is as follows:

[0067] Float(max_image_value)=ceil((float)temp_max_image_value / 100)*100+100;

[0068] Among them, max_image_value is the highlight value, which is saved and calculated as single-precision floating-point data;

[0069] temp_max_image_value is the maximum value of the T1 structure image, which is saved and calculated as single-precision floating-point data.

[0070] In this example: the maximum value of the T1 structural image is 872, and the highlight value is 1000.

[0071] Step 7: Output the T1 structural image containing the nerve bundle information into DICOM format, specifically:

[0072] Step 70: Generate a reading code according to the original DICOM file;

[0073] Step 71: Read the DICOM image according to the read code to obtain the DICOM image matrix and slice position information;

[0074] Step 72: Determine whether the DICOM format is LPS encoding;

[0075] Step 73: Determine whether the scan section type is a sagittal plane, a coronal plane, or an axial plane;

[0076] Step 74: Determine whether the image needs to be mirror-inverted;

[0077] Step 75: rearrange the DICOM image matrix according to the sorting;

[0078] Step 76: read the T1 structural image with neural information and obtain the NIfTI image matrix;

[0079] Step 77: Adjust the NIfTI image matrix order to be consistent with DICOM;

[0080] Step 78: Fill the NIfTI image matrix into the DICOM image matrix according to the section position information;

[0081] Step 79: Output the DICOM image according to the reading code of the DICOM format.

[0082] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the art within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An image processing method for brain structure images and three-dimensional neural bundle models, characterized in that: The steps include: Step 1: Convert the DWI diffusion image and T1 structural image into three-dimensional image space respectively; the DWI diffusion image and T1 structural image are both in DICOM format, and both are converted into three-dimensional image space in NIfTI format by dcm2niix; Step 2: Obtain a neural bundle model in a three-dimensional image space; Step 3: The DWI diffusion image, T1 structural image and nerve bundle model in the three-dimensional image space are transformed into the three-dimensional coordinate space through the rotation matrix; Step 4: Overlay the DWI diffusion image, T1 structural image and nerve bundle model in the three-dimensional coordinate space; Step 5: Use the inverse rotation matrix of the T1 structural image to transform the neural bundle model into the T1 structural image space; Step 6: constructing a T1 structural image containing neural bundle information by overlapping voxels of the T1 structural image and the neural bundle model in the T1 structural image space; the step 6 includes: Step 61: Overlaying the T1 structural image with the T1 spatial image of the nerve bundle model; Step 62: replacing the overlapping voxels of the T1 structural image and the T1 spatial image of the nerve bundle model with the highlighted values ​​to form a T1 structural image containing nerve bundle information; Step 7: Outputting the T1 structural image containing the nerve bundle information into DICOM format, wherein step 7 comprises: Step 70: Generate a reading code according to the original DICOM file; the original DICOM file includes the T1 structural image; Step 71: Read the original DICOM file according to the read code to obtain the image matrix and section position information of the original DICOM file; Step 72: Determine whether the DICOM format of the original DICOM file is LPS encoding; Step 73: Determine whether the scan section type is a sagittal plane, a coronal plane, or an axial plane; Step 74: Determine whether the image needs to be mirror-inverted; Step 75: rearrange the image matrix of the original DICOM file according to the sorting; Step 76: read the T1 structural image containing neural information and obtain the NIfTI image matrix; Step 77: Adjust the NIfTI image matrix order to be consistent with the original DICOM file; Step 78: Fill the NIfTI image matrix into the image matrix of the original DICOM file according to the section position information; Step 79: Output the T1 structural image containing the nerve bundle information into the DICOM format according to the reading code of the DICOM format of the original DICOM file.

2. The image processing method according to claim 1, characterized in that: The step 2 comprises: Step 21: Register the DWI diffusion image with the T1 structural image; Step 22: Obtain a neural bundle model based on DWI diffusion image processing.

3. The image processing method according to claim 2, wherein: In step 21, the ANTs library is used to register the DWI diffusion image with the T1 structural image.

4. The image processing method according to claim 2, wherein: The step 22 comprises: Step 221: performing DTI model fitting on the DWI diffusion image to obtain the maximum eigenvector V1 and anisotropic index FA of the ellipsoid model; Step 222: Assume that the tracking termination parameter FA threshold is ≤ 0.2 and the turning angle is ≥ 60˚, and perform whole-brain neural tracking based on the maximum eigenvector V1 and the anisotropic index FA; Step 223: intercept the neural bundle according to the region of interest to obtain a neural bundle model.

5. The image processing method according to claim 1, wherein: The rotation matrix in step 3 is a 3x4 matrix.

6. The image processing method according to claim 1, characterized in that: The calculation process of the highlight value in step 62 is: float(max_image_value) = ceil((float)temp_max_image_value / 100)*100 + 100; the max_image_value is the highlight value; the temp_max_image_value is the maximum value of the T1 structure image.