A method for batch processing of skull stripping of MRI images of rats and mice

Abstract

The application discloses a kind of MRI image skull stripping batch processing methods suitable for mouse and rat, and the brain magnetic resonance image of test sample is collected;Brain structure magnetic resonance image template is constructed;Draw the parenchyma mask image not containing skull;Structural magnetic resonance image is pretreated;Structural magnetic resonance image is registered to brain structure magnetic resonance image template;The parenchyma mask image is handled, and the parenchyma structure magnetic resonance image of single test sample after skull stripping is generated;The parenchyma functional magnetic resonance image after skull stripping is generated.The application realizes skull stripping batch processing to image file, saves manpower and time cost;Registration quality is high, and effect is good;Multiple use bias field correction, guarantee the accuracy of result.

Description

Technical Field

[0001] This invention relates to the field of image data processing, and more specifically to a batch processing method for skull dissection in MRI images of mice and rats. Background Technology

[0002] Magnetic resonance imaging (MRI) is a non-invasive tissue imaging technique with high spatiotemporal resolution. MRI is widely used in the study of brain structure and function in small animals, playing a crucial role in biological and preclinical medical research. In population studies, to ensure proper alignment between individual sample images, the images need to be transformed to a standard space—that is, standardized—before subsequent group analysis.

[0003] Skull dissection and segmentation are fundamental steps in the spatial standardization of magnetic resonance imaging (MRI) images. To achieve better spatial registration and tissue segmentation of brain imaging data, preprocessing typically involves removing the scalp, skull, and other non-brain tissues from structural images, leaving only the brain tissue. Before skull dissection, bias field correction is generally required. The bias field, caused by inhomogeneities excited by the magnetic field, manifests as changes in brightness between the image center and edges. Removing the bias field effect allows for better differentiation of the boundaries between brain and non-brain tissues.

[0004] For human brain data, several relatively effective skull dissection algorithms have been developed. However, due to the closer proximity of the brain tissue and scalp in mice and rats, and the differences in tissue contrast, these algorithms are not well-suited for mouse and rat data. Therefore, for mouse and rat data, brain parenchyma regions are typically masked manually. This is not only labor-intensive but also introduces human bias issues. Especially when a batch of data is processed by multiple people, differing understandings of brain anatomy among the processors can lead to inconsistent brain tissue extraction results, necessitating rigorous subsequent quality control. To avoid errors introduced by human processing, standardize processes, and significantly improve processing efficiency, especially for large batches of data, a batch processing method for skull dissection of MRI images suitable for mice and rats is needed. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned deficiencies in the existing technology by providing a batch processing method for skull dissection in MRI images of mice and rats.

[0006] The above-mentioned objective of the present invention is achieved through the following technical solution:

[0007] A batch processing method for skull dissection in MRI images of mice and rats includes the following steps:

[0008] Step 1: Collect brain magnetic resonance images of the subjects. Brain magnetic resonance images include structural magnetic resonance images and functional magnetic resonance images.

[0009] Step 2: Construct a brain structural magnetic resonance image template for the subject sample using structural magnetic resonance images;

[0010] Step 3: Draw a mask image of the brain parenchyma that does not include the skull based on the brain structure magnetic resonance image template;

[0011] Step 4: Preprocess the structural magnetic resonance images of a single subject sample;

[0012] Step 5: Register the preprocessed structural magnetic resonance image of a single subject sample onto the brain structural magnetic resonance image template to generate the corresponding transformation matrix, deformation field image, and inverse transformation image of the deformation field;

[0013] Step 6: Based on the transformation matrix, deformation field image and inverse transformation image of the deformation field obtained in Step 5, process the brain parenchyma mask image obtained in Step 3 to generate a magnetic resonance image of the brain parenchyma structure after skull dissection of a single subject sample.

[0014] Step 7: Generate functional magnetic resonance imaging (fMRI) images of the brain parenchyma after skull dissection for a single subject sample.

[0015] Step 2 as described above includes the following steps:

[0016] Step 2.1: Save the structural magnetic resonance images into the created folder;

[0017] Step 2.2: Use the antsMultivariateTemplateConstruction2.sh command of ANTs to process the structural magnetic resonance image to obtain a brain structural magnetic resonance image template;

[0018] Step 2.3: Use the DenoiseImage command of ANTs to perform noise reduction on the brain structure magnetic resonance image template obtained in Step 2.2;

[0019] Step 2.4: Use the N4BiasFieldCorrection command of ANTs to perform bias field correction on the brain structure magnetic resonance image template processed in Step 2.3.

[0020] Step 4 as described above includes the following steps:

[0021] Step 4.1: Use the N4BiasFieldCorrection command of ANTs to perform bias field correction on the structural magnetic resonance image of a single subject sample.

[0022] Step 4.2: Use the fslmaths-inm command of FSL to standardize the numerical values ​​of the structural magnetic resonance image of a single test sample, thereby obtaining the preprocessed structural magnetic resonance image of the test sample.

[0023] Step 6 as described above includes the following steps:

[0024] Step 6.1: Use the antsApplyTransforms command of ANTs to process the brain parenchyma mask image obtained in step 3 using the inverse transformation image and transformation matrix of the deformation field generated in step 5.

[0025] Step 6.2: Use the fslmaths-bin command of FSL to binarize the brain parenchyma mask image obtained in step 6.1 after transformation, and obtain the brain parenchyma mask image after skull dissection of a single subject sample.

[0026] Step 6.3: Using the fslmaths-mas command of FSL, process the brain parenchyma mask image of the single subject sample after skull dissection obtained in step 6.2 with the brain structure magnetic resonance image of the single subject sample to obtain the brain parenchyma structure magnetic resonance image of the single subject sample after skull dissection.

[0027] Step 6.4: Use the DenoiseImage command of ANTs to denoise the brain parenchyma structure magnetic resonance image of the single subject sample obtained in Step 6.3, and obtain the denoised brain parenchyma structure magnetic resonance image of the single subject sample after skull dissection.

[0028] Step 7 as described above includes the following steps:

[0029] Step 7.1: Use the AFNI 3dresample command to resample the brain parenchyma mask image of the single subject sample after skull dissection obtained in Step 6 to the functional magnetic resonance imaging space.

[0030] Step 7.2: Process the functional magnetic resonance image based on the resampled brain parenchyma mask image from Step 7.1 to obtain a preliminary functional magnetic resonance image of skull dissection.

[0031] Step 7.3: Take the functional magnetic resonance image of the preliminary skull dissection obtained in Step 7.2 at a certain time point, and use the antsRegistrationSyN.sh command of ANTs to register it onto the corresponding single subject's brain parenchyma structure magnetic resonance image after skull dissection, and generate the corresponding transformation matrix, deformation field image, and inverse transformation image of the deformation field.

[0032] Step 7.4: Using the antsApplyTransforms command of ANTs, process the brain parenchyma mask image of the single subject sample after skull stripping obtained in Step 6 according to the inverse transformation image and transformation matrix of the deformation field generated in Step 7.3, to obtain the functional magnetic resonance image of the single subject sample after skull stripping of the brain parenchyma mask image.

[0033] Step 7.5: Using the fslmaths-mas command of FSL, process the functional magnetic resonance image of a single subject sample based on the brain parenchyma mask image after skull dissection obtained in Step 7.4 to obtain the functional magnetic resonance image of the brain parenchyma after skull dissection of a single subject sample.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] 1. Enables batch processing of skull dissection for image files, saving manpower and time costs;

[0036] 2. High registration quality and good results;

[0037] 3. Use bias field correction multiple times to ensure the accuracy of the results. Detailed Implementation

[0038] To facilitate understanding and implementation of the present invention by those skilled in the art, the present invention will be further described in detail below with reference to embodiments. It should be understood that the embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0039] The magnetic resonance brain image dataset (rat / mouse magnetic resonance brain image dataset) used in this embodiment was obtained by scanning each subject sample with a Bruker BioSpec 70 / 20USR 7.0T small animal high-field magnetic resonance imaging system within the head coil, producing resting-state functional magnetic resonance imaging (EPI Image) and structural magnetic resonance imaging (T2 Image).

[0040] A batch processing method for skull dissection in MRI images of mice and rats includes the following steps:

[0041] Step 1: Acquire brain magnetic resonance images of the test sample (rat or mouse). The brain magnetic resonance images include structural magnetic resonance images and functional magnetic resonance images, specifically including the following steps:

[0042] Raw BRUKER images of the brains of subjects (rats or mice) were acquired using an MRI scanner. The raw BRUKER images were then converted to NII format structural and functional magnetic resonance imaging (fMRI) images using bru2nii software. The fMRI and structural MRI data were organized, categorized, and uniformly named: structural MRI images were named vbm.nii.gz; fMRI images were named epi.nii.gz. Image quality was checked, and images with poor quality were discarded.

[0043] Step 2: Using structural magnetic resonance imaging (MRI) images, construct a brain structural MRI image template for the test sample (rat or mouse), specifically including the following steps:

[0044] Step 2.1: Create a folder, put all structural MRI images into it, and generate a list of structural MRI images;

[0045] Step 2.2: Open the Linux terminal from the folder containing the structural magnetic resonance images of all subjects created in Step 2.1. Use the antsMultivariateTemplateConstruction2.sh command of ANTs (Advanced Normalization Tools) in the Linux terminal to output the brain structural magnetic resonance image template. Set the parameters and the prefix of the output brain structural magnetic resonance image template file name as needed.

[0046] Step 2.3: Use the DenoiseImage command of ANTs to perform noise reduction on the brain structure magnetic resonance image template obtained in Step 2.2;

[0047] Step 2.4: Use the N4BiasFieldCorrection command of ANTs to perform bias field correction on the brain structure magnetic resonance image template processed in Step 2.3.

[0048] Step 3: Draw a brain parenchyma mask image that does not include the skull based on the brain structure magnetic resonance image template. That is, peel off the skull from the brain structure magnetic resonance image template to generate a brain parenchyma mask image. The brain parenchyma mask image only includes the gray matter, white matter and cerebrospinal fluid.

[0049] Step 4: Preprocess the structural magnetic resonance images of a single subject sample, specifically including the following steps:

[0050] Step 4.1: Use the N4BiasFieldCorrection command of ANTs to perform bias field correction on the structural magnetic resonance image of a single subject sample;

[0051] Step 4.2: Use the fslmaths-inm command of FSL (FMRIB's Software Library) to standardize the numerical values ​​of the structural magnetic resonance images of a single subject sample, thereby obtaining the preprocessed structural magnetic resonance images of the subject sample.

[0052] Step 5: Register the preprocessed structural magnetic resonance image of a single subject sample onto the brain structural magnetic resonance image template, generating the corresponding transformation matrix, deformation field image, and inverse transformation image of the deformation field. This is because structural magnetic resonance images typically have high spatial resolution and inter-tissue contrast, facilitating registration. Specifically, this includes the following steps:

[0053] The preprocessed structural magnetic resonance image of a single subject sample is registered onto a brain structural magnetic resonance image template. Using the antsRegistrationSyN.sh command of ANTs, a transformation matrix reflecting the affine transformation of the image, a warp image, and an inverse warp image are generated.

[0054] Step 6: Based on the transformation matrix, deformation field image, and inverse transformation image of the deformation field obtained in Step 5, process the brain parenchyma mask image obtained in Step 3 to generate a magnetic resonance image of the brain parenchyma structure after skull dissection for a single subject sample. This specifically includes the following steps:

[0055] Step 6.1: Use the antsApplyTransforms command of ANTs to process the brain parenchyma mask image obtained in step 3 using the inverse warp image and transformation matrix of the deformation field generated in step 5.

[0056] Step 6.2: Use the fslmaths-bin command of FSL to binarize the brain parenchyma mask image obtained in step 6.1 after transformation, and set the value of voxels with a non-zero size to 1, thus obtaining the brain parenchyma mask image of a single subject sample after skull dissection.

[0057] Step 6.3: Using the fslmaths-mas command of FSL, process the brain parenchyma mask image of the single subject sample after skull dissection obtained in step 6.2 with the brain structure magnetic resonance image of the single subject sample to obtain the brain parenchyma structure magnetic resonance image of the single subject sample after skull dissection.

[0058] Step 6.4: Use the DenoiseImage command of ANTs to denoise the brain parenchyma structure magnetic resonance image of the single subject sample obtained in Step 6.3, and obtain the denoised brain parenchyma structure magnetic resonance image of the single subject sample after skull dissection.

[0059] Step 7: Register the functional magnetic resonance imaging (fMRI) image of a single subject sample onto the fMRI image of the brain parenchyma structure after skull dissection of the subject sample to obtain the fMRI image after skull dissection. This specifically includes the following steps:

[0060] Step 7.1: Use the AFNI 3dresample command to resample the brain parenchyma mask image of the single subject sample after skull dissection obtained in Step 6 to the functional magnetic resonance imaging space.

[0061] Step 7.2: Process the functional magnetic resonance image based on the resampled brain parenchyma mask image from Step 7.1 to obtain a preliminary functional magnetic resonance image of skull dissection.

[0062] Step 7.3: Take the functional magnetic resonance image of the preliminary skull dissection obtained in Step 7.2 at a certain time point, and use the antsRegistrationSyN.sh command of ANTs to register it onto the corresponding single subject's skull dissection brain parenchyma structure magnetic resonance image to generate a transformation matrix reflecting the affine transformation of the image, a warp image, and an inverse warp image.

[0063] Step 7.4: Using the antsApplyTransforms command of ANTs, process the brain parenchyma mask image of the single subject sample after skull stripping obtained in Step 6 based on the inverse warp image and transformation matrix of the deformation field generated in Step 7.3, to obtain the functional magnetic resonance image of the single subject sample after skull stripping of the brain parenchyma mask image.

[0064] Step 7.5: Using the fslmaths-mas command of FSL, process the functional magnetic resonance image of a single subject sample based on the brain parenchyma mask image after skull stripping obtained in Step 7.4 to obtain the functional magnetic resonance image of brain parenchyma after skull stripping of a single subject sample, and use the slices command to check the quality of the functional magnetic resonance image of brain parenchyma after skull stripping.

[0065] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A method for batch processing of skull stripping of MRI images of rats and mice, characterized in that, The method comprises the following steps: Step 1, collecting brain magnetic resonance images of the subject sample, the brain magnetic resonance images comprising structural magnetic resonance images and functional magnetic resonance images; Step 2, constructing a brain structure magnetic resonance image template of the subject sample by using the structural magnetic resonance images; Step 3, drawing a brain parenchyma mask image not containing a skull according to the brain structure magnetic resonance image template; Step 4, preprocessing the structural magnetic resonance images of the single subject sample; Step 5, registering the single subject sample preprocessed structural magnetic resonance images to the brain structure magnetic resonance image template to generate a corresponding transformation matrix, a deformation field image and an inverse transformation image of the deformation field; Step 6, processing the brain parenchyma mask image obtained in step 3 according to the transformation matrix, the deformation field image and the inverse transformation image of the deformation field obtained in step 5 to generate a skull-stripped brain parenchyma structural magnetic resonance image of the single subject sample; Step 7, generating a skull-stripped brain parenchyma functional magnetic resonance image of the single subject sample, The step 6 comprises the following steps: Step 6.1, using the inverse transformation image of the deformation field and the transformation matrix generated in step 5 to process the brain parenchyma mask image obtained in step 3 by using the antsApplyTransforms command of ANTs; Step 6.2, binarizing the brain parenchyma mask image obtained in step 6.1 after transformation by using the fslmaths -bin command of FSL to obtain a skull-stripped brain parenchyma mask image of the single subject sample; Step 6.3, processing the brain structure magnetic resonance images of the single subject sample by using the skull-stripped brain parenchyma mask image of the single subject sample obtained in step 6.2 by using the fslmaths -mas command of FSL to obtain a skull-stripped brain parenchyma structural magnetic resonance image of the single subject sample; Step 6.4, denoising the skull-stripped brain parenchyma structural magnetic resonance image of the single subject sample obtained in step 6.3 by using the DenoiseImage command of ANTs to obtain a skull-stripped brain parenchyma structural magnetic resonance image of the single subject sample after denoising, The step 7 comprises the following steps: Step 7.1, resampling the skull-stripped brain parenchyma mask image of the single subject sample obtained in step 6 to the functional magnetic resonance image space by using the 3dresample command of AFNI; Step 7.2, processing the functional magnetic resonance images according to the resampled brain parenchyma mask image in step 7.1 to obtain a preliminary skull-stripped functional magnetic resonance image; Step 7.3, taking an image at a time point of the preliminary skull-stripped functional magnetic resonance image obtained in step 7.2, and registering the image to the skull-stripped brain parenchyma structural magnetic resonance image of the corresponding single subject sample by using the antsRegistrationSyN.sh command of ANTs to generate a corresponding transformation matrix, a deformation field image and an inverse transformation image of the deformation field. Step 7.4, using the antsApplyTransforms command of ANTs, processing the skull-stripped brain mask image of the single subject sample obtained in step 6 according to the inverse transformation image and the transformation matrix of the deformation field generated in step 7.3, to obtain the skull-stripped brain mask image of the functional magnetic resonance image of the single subject sample; Step 7.5, using the fslmaths –mas command of FSL, processing the functional magnetic resonance image of the single subject sample according to the skull-stripped brain mask image of the functional magnetic resonance image of the single subject sample obtained in step 7.4, to obtain the skull-stripped brain functional magnetic resonance image of the single subject sample.

2. The method according to claim 1, wherein, The step 2 comprises the following steps: Step 2.1, storing the structural magnetic resonance image into the created folder; Step 2.2, using the antsMultivariateTemplateConstruction2.sh command of ANTs to process the structural magnetic resonance image to obtain a brain structural magnetic resonance image template; Step 2.3, using the DenoiseImage command of ANTs to process the brain structural magnetic resonance image template obtained in step 2.2 to obtain a denoised brain structural magnetic resonance image template; Step 2.4, using the N4BiasFieldCorrection command of ANTs to process the denoised brain structural magnetic resonance image template obtained in step 2.3 to obtain a bias field corrected brain structural magnetic resonance image template.

3. The method according to claim 2, wherein, The step 4 comprises the following steps: Step 4.1, using the N4BiasFieldCorrection command of ANTs to process the structural magnetic resonance image of the single subject sample to obtain a bias field corrected structural magnetic resonance image of the single subject sample; Step 4.2, using the fslmaths –inm command of FSL to standardize the numerical value of the structural magnetic resonance image of the single subject sample, to obtain a pre-processed structural magnetic resonance image of the single subject sample.