MRI image feature extraction method for dynamic evaluation of pediatric brain development

Abstract

The application relates to the technical field of image processing, in particular to an MRI image feature extraction method for dynamic evaluation of child brain development, which comprises the following steps: determining a cerebrospinal fluid coefficient of a child brain; dividing an MRI image into a cerebrospinal fluid region and a residual brain region based on the cerebrospinal fluid coefficient of the child brain of each pixel point; determining a white matter coefficient of each pixel point in the residual brain region according to the gray value of each pixel point in the decayed weighted image of each pixel point in different gradient directions, and the gray value of each pixel point in the first weighted image and the second weighted image in the residual brain region; segmenting the white matter region and the gray matter region of the residual brain region based on the white matter coefficient, and extracting image features in the white matter region and the gray matter region. In this way, the accuracy and reliability of the segmentation result of different tissue regions in the child brain MRI image are improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for extracting features from MRI images for dynamic assessment of pediatric brain development. Background Technology

[0002] Childhood is a critical stage for brain development. Brain structures develop rapidly in the first few years after birth, especially in terms of the infrastructure for functions such as language, motor skills, and cognition. Monitoring brain development allows for the early detection of neurodevelopmental abnormalities or diseases. Magnetic Resonance Imaging (MRI) is a non-invasive imaging technique widely used in neuroimaging. It provides high-resolution images of brain structures, helping researchers observe the developmental process of children's brain structures in detail. Compared to other imaging techniques (such as CT scans), MRI has higher soft tissue resolution, making it particularly suitable for observing subtle changes in brain development.

[0003] In some scenarios, to dynamically assess pediatric brain development, it is necessary to extract features from pediatric brain MRI images. This involves segmenting and extracting different brain tissue regions from the MRI images to dynamically assess brain development. However, because gray and white matter are fully myelinated during pediatric development, the distinction between gray and white matter in MRI images becomes blurred. Furthermore, fluid in some incompletely myelinated white matter can mix with cerebrospinal fluid, making tissue segmentation in MRI images difficult. Consequently, the accuracy and reliability of segmentation results for different brain tissue regions in pediatric brain MRI images are relatively low. Summary of the Invention

[0004] To address the technical problem of low accuracy and reliability in segmentation results of different brain tissue regions in pediatric brain MRI images, the present invention aims to provide a method for extracting MRI image features for dynamic assessment of pediatric brain development.

[0005] To solve the above technical problems, the specific technical solution adopted is as follows: In a first aspect, embodiments of the present invention provide an MRI image feature extraction method for dynamic assessment of pediatric brain development, comprising: determining the cerebrospinal fluid coefficient of the pediatric brain based on the gray values ​​of each pixel in the first weighted image and the second weighted image of the MRI image of the pediatric brain, and the gray values ​​of each pixel in the attenuated weighted image of the MRI image with different gradient directions; dividing the MRI image into a cerebrospinal fluid region and a remaining brain region based on the cerebrospinal fluid coefficient of each pixel; determining the white matter coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the attenuated weighted image with different gradient directions, and the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image respectively; segmenting the white matter region and the gray matter region of the remaining brain region based on the white matter coefficient, and extracting image features from the white matter region and the gray matter region.

[0006] Optionally, determining the cerebrospinal fluid coefficient of the pediatric brain based on the gray values ​​of each pixel in the first and second weighted images of the MRI image of the pediatric brain, and the gray values ​​of each pixel in the attenuation weighted images of different gradient directions in the MRI image, includes: determining the fluid coefficient of each pixel based on the gray values ​​of each pixel in the first and second weighted images of the MRI image of the pediatric brain; determining the attenuation ability of each pixel in each gradient direction based on the gray values ​​of each pixel in the attenuation weighted images of different gradient directions in the MRI image; determining the attenuation rate of each pixel based on the attenuation ability of each pixel in each gradient direction; and determining the cerebrospinal fluid coefficient of the pediatric brain based on the fluid coefficient and the attenuation rate.

[0007] Optionally, determining the liquid coefficient of any pixel based on the gray value of any pixel in the first weighted image and the second weighted image of the pediatric brain MRI image includes: determining a first ratio between the gray value of each pixel in the second weighted image of the pediatric brain MRI image and the gray value in the first weighted image, and normalizing the first ratio to obtain the liquid coefficient.

[0008] Optionally, determining the attenuation capability of each pixel in each gradient direction based on the gray values ​​of each pixel in the attenuation weighted images of different gradient directions in the MRI image includes: using the attenuation weighted image with a preset magnetic field gradient as a reference image, obtaining attenuation weighted images of the reference image under different gradient directions; aligning the reference image with the attenuation weighted images of the reference image under different gradient directions, arranging the aligned attenuation weighted images to obtain an attenuation weighted image sequence; using the sequence of gray value changes of each pixel in the attenuation weighted image sequence as the gray value sequence of each pixel; using the first gray value in the gray value sequence of each pixel as the initial value, and using the second ratio of the initial value to the other gray values ​​in the gray value sequence as the attenuation rate in different gradient directions of the gray value sequence of each pixel.

[0009] Optionally, determining the attenuation rate of each pixel based on its attenuation capability in each gradient direction includes: normalizing the average value of the attenuation capability of each pixel in each gradient direction to obtain the attenuation rate of each pixel.

[0010] Optionally, determining the white matter coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the attenuated weighted images of the remaining brain region in different gradient directions, and the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image respectively, includes: determining the attenuation ability of each pixel in the remaining brain region in each gradient direction based on the gray values ​​of each pixel in the attenuated weighted images of the remaining brain region in different gradient directions in the MRI image; determining the fiber tract coefficient of each pixel in the remaining brain region based on the minimum value of the attenuation ability of each pixel in the remaining brain region in each gradient direction and the attenuation ability of each pixel in the remaining brain region in other gradient directions; determining the lipid coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image respectively; and determining the white matter coefficient of each pixel in the remaining brain region based on the fiber tract coefficient and the lipid coefficient.

[0011] Optionally, determining the fiber bundle coefficient of each pixel in the remaining brain region based on the minimum weakening ability of each pixel in each gradient direction and the weakening ability of each pixel in the remaining brain region in other gradient directions includes: determining a weakening ability difference coefficient based on the weakening ability of each pixel in the remaining brain region in other gradient directions and the minimum weakening ability of each pixel in the remaining brain region in each gradient direction; and determining the fiber bundle coefficient based on the weakening ability difference coefficient and the minimum weakening ability.

[0012] Optionally, determining the lipid coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the first weighted image and the second weighted image includes: determining the gray difference coefficient between the gray values ​​of each pixel in the remaining brain region and the gray values ​​of each pixel in the first weighted image and the second weighted image, and the sum of the gray values ​​of each pixel in the remaining brain region and the gray values ​​of each pixel in the first weighted image and the second weighted image; and determining the lipid coefficient based on the gray difference coefficient and the sum.

[0013] Optionally, determining the white matter coefficient of each pixel in the remaining brain region based on the fiber bundle coefficient and lipid coefficient includes: determining a first weighted coefficient of the fiber bundle coefficient and a second weighted coefficient of the lipid coefficient; and determining the white matter coefficient of each pixel in the remaining brain region based on the fiber bundle coefficient, lipid coefficient, first weighted coefficient, and second weighted coefficient.

[0014] Optionally, segmenting the remaining brain regions into white matter and gray matter regions based on white matter coefficients and extracting image features from the white matter and gray matter regions includes: using the Otsu thresholding method, obtaining the first threshold portion of the white matter coefficient of each pixel as the white matter region and the second threshold portion as the gray matter region, where the first threshold portion is greater than the second threshold portion; extracting the volume of the gray matter region, the volume of the white matter region, the corpus callosum parameters and corticospinal tract parameters on the white matter region; and using the volume of the gray matter region, the volume of the white matter region, the corpus callosum parameters and corticospinal tract parameters on the white matter region as image features.

[0015] This invention offers the following advantages: By combining the pixel grayscale values ​​of the first weighted image and the second weighted image, along with the grayscale information from the multi-gradient direction attenuated weighted image, a cerebrospinal fluid (CSF) coefficient is constructed. This coefficient simultaneously reflects the grayscale differences and diffusion characteristics of gray matter and CSF, effectively avoiding the problem of "confusion between unmyelinated white matter fluid and CSF fluid signals." This makes the boundary between the CSF region and the remaining brain region clearer, significantly reducing the CSF missegmentation rate. After segmenting the remaining brain region, the white matter coefficient is further calculated by combining the grayscale values ​​of the multi-gradient direction attenuated weighted image and the dual-weighted image. Water molecules in white matter diffuse with a certain directionality, while water molecules in gray matter diffuse more randomly. This difference in diffusion characteristics is reflected in the attenuated weighted image. Combined with the grayscale differences between gray and white matter in the dual-weighted image, the white matter coefficient can accurately distinguish between white matter and gray matter, solving the problem of "blurred boundaries caused by incomplete myelination of gray and white matter." The technical solution provided by this invention makes it easier to segment tissues in MRI images, thereby improving the accuracy and reliability of segmentation results for different tissue regions of the brain in pediatric brain MRI images. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an MRI image feature extraction method for dynamic assessment of pediatric brain development, as disclosed in one embodiment of the present invention. Figure 2 A schematic diagram of an MRI image of a child's brain provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of a weakened weighted image of a reference image under different gradient directions, provided as an embodiment of the present invention. Figure 4This is a schematic diagram of the structure of an electronic device disclosed in one embodiment of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an MRI image feature extraction method for dynamic assessment of pediatric brain development proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following describes in detail, with reference to the accompanying drawings, a specific scheme of the MRI image feature extraction method for dynamic assessment of pediatric brain development disclosed in this invention.

[0021] Example 1: Please see Figure 1 The document illustrates a flowchart of an MRI image feature extraction method for dynamic assessment of pediatric brain development according to an embodiment of the present invention, comprising: Step S101: Determine the cerebrospinal fluid coefficient of the child's brain based on the gray values ​​of each pixel in the first weighted image and the second weighted image of the MRI image of the child's brain, and the gray values ​​of each pixel in the attenuated weighted image of different gradient directions in the MRI image.

[0022] Specifically, in the acquisition of MRI images of the pediatric brain in this embodiment of the invention, the specific process is as follows: In a 3.0T superconducting MRI scanner (such as Siemens Prisma) equipped with an 8-channel neonatal-specific head phased array coil, infants aged 0-9 months (grouped according to corrected gestational age: preterm infants ≥29 weeks, full-term infants, and infants aged 1 / 4 / 8 months) who have undergone strict screening and have no contraindications to MRI are sedated in natural sleep or by oral administration of chloral hydrate (50mg / kg body weight). The trunk is wrapped in a thermoplastic vacuum fixation bag, and triple motion inhibition measures are used, including silicone earplugs for noise reduction and a head positioning bracket. Simultaneously, the electrocardiogram (sampling rate 500Hz), blood oxygen saturation (±2% accuracy), and respiratory waveform (chest belt piezoelectric sensor) are monitored.

[0023] The multimodal scanning protocol was then initiated. First, three-plane localization images (TR / TE = 5 / 2ms, slice thickness 5mm) were acquired to determine the scanning reference plane. Then, a high-resolution 3DT1-MPRAGE sequence was executed as the first weighted image (T1-weighted imaging (T1WI)): sagittal scan, TR / TE = 1900 / 2.98ms, flip angle 7°, voxel 1×1×1mm³, matrix 256×256×160, parallel acceleration factor. GRAPPA=2, bandwidth 240Hz / pixel, scan duration 5 minutes 12 seconds; synchronous triggering of isotropic 3DT2-SPACE sequence as second weighted image (T2-weighted imaging (T2WI)): TR / TE=3200 / 409ms, variable flipping angle chain (50°-180°), voxel 1×1×1mm³, Turbo factor 80, fat saturation technique was used to suppress skull base artifacts, scan duration 6 minutes 30 seconds.

[0024] Diminished spectral imaging (DWI) was used as the attenuated weighted image: single-shot EPI, bmax=7000s / mm² (257 direction), 8 frames of b0 images, slice thickness 2.5mm without gaps, FOV=220×220mm, phase encoding direction applied with POCS motion correction, duration 9 minutes 45 seconds; finally, axial T2-FLAIR (TR / TE / TI=9000 / 94 / 2500ms, slice thickness 3mm) and proton density-weighted images (TR / TE=3500 / 15ms) were added for lesion screening; throughout the process, head movement was detected in real time using navigation echo (CLOVER technology) (threshold ≤1mm), if the displacement exceeded the threshold, the current sequence was automatically rescanned, the total scan time was strictly controlled within 38±2 minutes, and the image quality was immediately evaluated after completion (SNR>20, relative uniformity ≥85%), the raw data was archived in DICOM format and backed up to an encrypted hard drive.

[0025] For example, such as Figure 2 As shown, Figure 2This is a schematic diagram of an MRI image of a child's brain provided in one embodiment of the present invention. When dynamically assessing a child's brain development, it is necessary to evaluate the morphology of gray matter, white matter, and cerebrospinal fluid in the MRI images of the child's brain. Specifically, in T1WI, low signal intensity reflects areas with high intracellular fluid content (such as edema or cyst fluid) or low lipid, calcification, or fibrosis content. For example, cerebrospinal fluid shows low signal intensity on T1WI, while soft tissues such as bone show intermediate signal intensity. In T2WI, high signal intensity is usually associated with increased tissue water content or widening of extracellular spaces, such as inflammation, tumors, or effusion. Fluid (such as cerebrospinal fluid or cysts) appears as significantly high signal intensity on T2WI. Furthermore, besides cerebrospinal fluid, the remaining gray and white matter show significant differences in the images; therefore, cerebrospinal fluid can be segmented first, followed by gray and white matter segmentation.

[0026] Furthermore, as an optional embodiment of the present invention, determining the cerebrospinal fluid coefficient of the child's brain based on the gray values ​​of each pixel in the first weighted image and the second weighted image of the MRI image of the child's brain, and the gray values ​​of each pixel in the attenuation weighted image of different gradient directions in the MRI image, includes: determining the fluid coefficient of each pixel based on the gray values ​​of each pixel in the first weighted image and the second weighted image of the MRI image of the child's brain; determining the attenuation ability of each pixel in each gradient direction based on the gray values ​​of each pixel in the attenuation weighted image of different gradient directions in the MRI image; determining the attenuation rate of each pixel based on the attenuation ability of each pixel in each gradient direction; and determining the cerebrospinal fluid coefficient of the child's brain based on the fluid coefficient and the attenuation rate.

[0027] Specifically, this embodiment of the invention first acquires T1WI and T2WI images from MRI images of the child's brain and aligns the two images. Then, a neural network is used to obtain the brain region in the aligned T1WI and T2WI images. Next, the grayscale value of any pixel in the brain region is obtained in the T1WI and T2WI images. This embodiment of the invention records the grayscale value of any pixel in the T1WI image as... The gray value of any pixel in the T2WI image is denoted as .

[0028] Because fluids exhibit low signal intensity in T1WI images and high signal intensity in T2WI images, as an optional embodiment of the present invention, determining the fluid coefficient of any pixel based on the grayscale value of any pixel in the first weighted image and the second weighted image of the pediatric brain MRI image includes: determining a first ratio between the grayscale value of each pixel in the second weighted image and the grayscale value in the first weighted image of the pediatric brain MRI image, and normalizing the first ratio to obtain the fluid coefficient.

[0029] Specifically, the embodiments of the present invention use the following formula to calculate the liquid coefficient: ; In the above formula, Indicates the first Liquid coefficient per pixel. This represents the grayscale value of any pixel in a T1WI image. This represents the grayscale value of any pixel in a T2WI image. A higher value indicates that the pixel has a higher gray value in the T2WI image, representing a high signal, while a lower gray value in the T1WI image represents a low signal. Therefore, the pixel is more consistent with the liquid's appearance in both T1WI and T2WI images, and thus the pixel has a higher liquid coefficient.

[0030] Furthermore, in newborns and early infants, the degree of myelination of gray and white matter is low. In T2WI and T1WI images, parts of gray and white matter may exhibit cerebrospinal fluid image characteristics, making automatic and accurate segmentation of gray matter, white matter, and cerebrospinal fluid extremely challenging. Therefore, it is necessary to extract additional segmentation features from MRI images for enhancement. Thus, this embodiment of the invention acquires diffusion-weighted imaging (DWI) images from MRI images. DWI images capture the microscopic decaying motion of water molecules, thus providing information about brain white matter. Further, Canny edge detection is used to obtain the gradient of pixels in the DWI image. Since DWI images show incomplete myelination of white matter, they can reflect the decay characteristics of white matter in the DWI image; therefore, it is necessary to extract these white matter decay characteristics from the DWI image. DWI images reflect the weakening of water molecules under magnetic field gradients in different directions. When the fiber bundle is parallel to the gradient direction, the signal attenuation is less (higher gray values ​​in the image), while when the fiber bundle is perpendicular to the gradient direction, the signal attenuation is greater (lower gray values ​​in the image). Therefore, as an optional embodiment of the present invention, determining the attenuation capability of each pixel in each gradient direction based on the gray values ​​of each pixel in the attenuation weighted images of different gradient directions in the MRI image includes: using the attenuation weighted image with a preset magnetic field gradient as a reference image, obtaining attenuation weighted images of the reference image under different gradient directions; aligning the reference image with the attenuation weighted images of the reference image under different gradient directions, arranging the aligned attenuation weighted images to obtain an attenuation weighted image sequence; using the sequence of gray value changes of each pixel in the attenuation weighted image sequence as the gray value sequence of each pixel; using the first gray value in the gray value sequence of each pixel as the initial value, and using the second ratio of the initial value to the other gray values ​​in the gray value sequence as the attenuation rate in different gradient directions of the gray value sequence of each pixel.

[0031] Specifically, in this embodiment of the invention, the preset value can be 0. This embodiment selects the weakened weighted image with b=0 as the reference image. Since the magnetic field has no gradient when b=0, it will weaken freely, thus b=0 is used as the reference image. Then, weakened weighted images under different gradient directions are obtained with the reference image as the center. For example, as shown... Figure 3 As shown, Figure 3 This invention provides a schematic diagram of attenuation weighted images of a reference image under different gradient directions, as provided in one embodiment of the present invention. Figure 3 This includes eight DWI images at horizontal, vertical, and diagonal gradients from the baseline image. Furthermore, for DWI images at different gradient directions, they are compared with... Align the reference image at that time. Then, transfer the aligned DWI image from... The process begins by arranging the images to obtain a weakened weighted image sequence. Further, a neural network is used to extract the pediatric brain region from the weakened weighted image sequence. For any pixel in the pediatric brain region of the weakened weighted image sequence, the sequence of grayscale value changes within the weakened weighted image sequence is obtained and denoted as the grayscale sequence for each pixel. Then, the first grayscale value in the grayscale sequence is obtained as the initial value, denoted as [initial value] in this embodiment. Then, for the grayscale sequence of pixels other than the initial values... Other grayscale values Get the initial value Other grayscale values The ratio between Let denoted as the attenuation rate in different directions of the grayscale sequence of a pixel.

[0032] Furthermore, when a pixel exhibits a low decay rate in the current gradient direction, it indicates that the current gradient direction is parallel to the fiber direction, thus the diffusion ability in this direction is strong, and a fiber structure exists in this direction. Since the fiber structures in the gray matter, white matter, and cerebrospinal fluid of a child's brain differ, their decay abilities also differ in different gradient directions. To distinguish cerebrospinal fluid from unmyelinated gray and white matter, it is necessary to analyze their decay rates. As an optional embodiment of the present invention, determining the decay rate of each pixel based on its decay ability in each gradient direction includes: normalizing the average value of the decay ability of each pixel in each gradient direction to obtain the decay rate of each pixel.

[0033] Specifically, embodiments of the present invention obtain the attenuation capability in the grayscale sequence of pixels. The mean of the value is normalized to assess the weakening ability. A higher mean value indicates a stronger decay ability of the pixel in different gradient directions, thus indicating that water molecules are free at the pixel location (possibly cerebrospinal fluid), and vice versa, indicating that water molecules are not free (located in dense cellular areas). The embodiments of this invention will... The attenuation rate of each pixel is denoted as... .

[0034] Furthermore, embodiments of the present invention incorporate the liquid coefficient of any pixel in the pediatric brain region. With decay rate , liquid coefficient With decay rate The product of these two factors is used as the cerebrospinal fluid coefficient of the pediatric brain. .

[0035] Step S102: Based on the cerebrospinal fluid coefficient of the pediatric brain at each pixel, the MRI image is divided into the cerebrospinal fluid region and the remaining brain region.

[0036] Specifically, cerebrospinal fluid (CSF) contains a large amount of fluid and exhibits high attenuation rates in all directions. Therefore, Otsu thresholding can be used to identify the high-threshold portion of the cerebrospinal fluid coefficient in pixels of the pediatric brain from MRI images, denoted as the CSF portion of the pediatric brain. The remaining portion is considered the residual brain region. This residual brain region includes both white and gray matter in the pediatric brain.

[0037] Step S103: Determine the white matter coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the attenuated weighted images in different gradient directions and the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image.

[0038] Specifically, the white and gray matter in a child's brain contain different tissue structures. White matter continuously produces fibrous structures as it myelinates. After myelination, these fibrous structures are formed within the white matter. Because these fibers weaken when parallel to the gradient direction in DWI images, myelinated white matter will show significantly less grayscale weakening in one gradient direction than in others in DWI image sequences. Therefore, to segment the white matter, it is necessary to obtain the weakening direction of pixels in the remaining brain regions.

[0039] Further, as an optional embodiment of the present invention, determining the white matter coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the attenuated weighted images of the remaining brain region in different gradient directions, and the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image respectively, includes: determining the attenuation capability of each pixel in the remaining brain region in each gradient direction based on the gray values ​​of each pixel in the attenuated weighted images of the remaining brain region in different gradient directions in the MRI image; determining the fiber tract coefficient of each pixel in the remaining brain region based on the minimum value of the attenuation capability of each pixel in the remaining brain region in each gradient direction and the attenuation capability of each pixel in the remaining brain region in other gradient directions; determining the lipid coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image respectively; and determining the white matter coefficient of each pixel in the remaining brain region based on the fiber tract coefficient and the lipid coefficient.

[0040] Specifically, embodiments of the present invention acquire the grayscale sequence of each pixel in the remaining brain region and the attenuation ability of each pixel in different gradient directions within the grayscale sequence. The minimum value of the attenuation ability of each pixel in the remaining brain region in different gradient directions within the grayscale sequence is then obtained. K And acquire the decay capability in other gradient directions. and minimum value K Mean of the absolute values ​​of the differences between When a pixel in a DWI image sequence exhibits significantly greater attenuation in one gradient direction than elsewhere, it indicates a strong fiber bundle coefficient for that pixel. Therefore, as an optional embodiment of the present invention, determining the fiber bundle coefficient of each pixel in the remaining brain region based on the minimum attenuation capacity of each pixel in each gradient direction and the attenuation capacity of each pixel in the remaining brain region in other gradient directions includes: determining an attenuation capacity difference coefficient based on the attenuation capacity of each pixel in the remaining brain region in other gradient directions and the minimum attenuation capacity of each pixel in the remaining brain region in each gradient direction; and determining the fiber bundle coefficient based on the attenuation capacity difference coefficient and the minimum attenuation capacity.

[0041] Specifically, in this embodiment of the invention, the fiber bundle coefficient is calculated as follows: the absolute value of a first difference between the attenuation ability of each pixel in the remaining brain region in other gradient directions and the minimum attenuation ability of each pixel in the remaining brain region in each gradient direction is calculated; the product of the minimum attenuation ability and the absolute value of the first difference is normalized to obtain the fiber bundle coefficient. In this embodiment of the invention, the absolute value of the first difference is used as the attenuation ability difference coefficient.

[0042] The fiber bundle coefficient is specifically calculated using the following formula in this embodiment of the invention: ; In the above formula, This represents the fiber bundle coefficient of the j-th pixel in the remaining brain region. This represents the minimum decay capability of the j-th pixel in the remaining brain region across all gradient directions. This represents the absolute value of the difference between the decay ability of the j-th pixel in the remaining brain region and the minimum value in each gradient direction. This represents the normalization function, used to normalize... Normalization was performed. The larger the value, the weaker the attenuation ability of the gradient direction with the greatest attenuation ability in the sequence of pixels. Furthermore, the direction with the weakest attenuation ability differs significantly from other directions. Therefore, it indicates that the gradient direction with the minimum attenuation rate is the main grayscale maintenance direction, thus indicating a higher consistency in grayscale maintenance direction.

[0043] It should be noted that when calculating the fiber bundle coefficient, if the denominator of the calculation formula is 0, a zero-prevention parameter of 0.001 should be added to the denominator before calculation.

[0044] Furthermore, the fiber tract coefficients of the remaining brain regions obtained through the above embodiments of the present invention are due to the presence of fibrous structures within the white matter after myelination; however, because some white matter is not fully myelinated during the development of children, incomplete segmentation may occur. To completely segment white matter and gray matter, it is necessary to determine the degree of myelination in different parts of the brain. During white matter myelination, oligodendrocytes secrete cholesterol and phospholipids, leading to an increase in lipids in the white matter, which results in higher gray levels in T1-weighted images, while the decrease in tissue water content leads to lower gray levels in T2-weighted images. Therefore, the embodiments of the present invention obtain the gray values ​​of pixels in the remaining brain regions in T1-weighted images. Gray values ​​in T2-weighted images Since white matter myelination can result in increased grayscale in T1-weighted images and decreased grayscale in T2-weighted images, the lipid-water ratio conversion can be quantified by comparing the relative intensity changes in T1-weighted and T2-weighted images. Therefore, as an optional embodiment of the present invention, determining the lipid coefficient of each pixel in the remaining brain region based on the grayscale values ​​of each pixel in the first and second weighted images includes: determining the grayscale difference coefficient between the grayscale values ​​of each pixel in the remaining brain region in the first and second weighted images, and the sum of the grayscale values ​​of each pixel in the remaining brain region in the first and second weighted images; and determining the lipid coefficient based on the grayscale difference coefficient and the sum.

[0045] Specifically, the lipid coefficient is calculated in the following manner in this embodiment of the invention: the absolute value of the second difference between the gray values ​​of each pixel in the remaining brain region and the gray values ​​of the first weighted image and the second weighted image are calculated, and the sum of the gray values ​​of each pixel in the remaining brain region and the gray values ​​of the first weighted image and the second weighted image is calculated; the second ratio between the absolute value of the second difference and the sum is determined as the lipid coefficient, wherein the absolute value of the second difference is used as the gray difference coefficient in this embodiment of the invention.

[0046] Specifically, the lipid coefficient is calculated using the following formula in the embodiments of the present invention: ; In the above formula, This represents the lipid coefficient of the j-th pixel in the remaining brain region. This represents the grayscale value of the j-th pixel in the remaining brain region in the first weighted image. This represents the grayscale value of the j-th pixel in the remaining brain region in the second weighted image. A larger value indicates that the difference between the gray levels of the pixels in the first weighted image and the gray levels in the second weighted image is greater than the sum of the gray levels in the first weighted image and the second weighted image. Therefore, it indicates that there is more change in the relative intensity of the gray levels of the pixels in the first weighted image and the second weighted image, indicating that there is more conversion between water and lipid ratios in the pixel area, further indicating that the lipid coefficient of the pixel is larger.

[0047] Furthermore, for the lipid coefficients of each obtained pixel, as an optional embodiment of the present invention, determining the white matter coefficients of each pixel in the remaining brain region based on the fiber bundle coefficients and lipid coefficients includes: determining a first weighting coefficient of the fiber bundle coefficients and a second weighting coefficient of the lipid coefficients; and determining the white matter coefficients of each pixel in the remaining brain region based on the fiber bundle coefficients, lipid coefficients, the first weighting coefficient, and the second weighting coefficient.

[0048] Specifically, in this embodiment of the invention, the weighted sum of the fiber bundle coefficient and the lipid coefficient is used as the white matter coefficient of each pixel in the remaining brain region. This embodiment of the invention utilizes the formula... The white matter coefficient of each pixel was obtained by weighting. ,in, This represents the lipid coefficient of the j-th pixel in the remaining brain region. This represents the fiber tract coefficient of the j-th pixel in the remaining brain region. Indicates the first weighting coefficient. This represents the second weighting coefficient. and The value of can be determined according to the actual situation, as described in the embodiments of the present invention. The value is 0.6. The value is 0.4.

[0049] Step S104: Segment the white matter and gray matter regions of the remaining brain regions based on the white matter coefficient, and extract image features from the white matter and gray matter regions.

[0050] Specifically, in this embodiment of the invention, the Otsu thresholding method is used to obtain the high-threshold portion of the white matter coefficient of each pixel, which is denoted as the white matter region, and the low-threshold portion as the gray matter region. Further, as an optional embodiment of the invention, segmenting the white matter and gray matter regions of the remaining brain region based on the white matter coefficient, and extracting image features from the white matter and gray matter regions includes: based on the Otsu thresholding method, obtaining the first threshold portion of the white matter coefficient of each pixel, denoted as the white matter region, and the second threshold portion as the gray matter region, where the first threshold portion is greater than the second threshold portion; extracting the volume of the gray matter region, the volume of the white matter region, the corpus callosum parameters on the white matter region, and the corticospinal tract parameters; and using the volume of the gray matter region, the volume of the white matter region, the corpus callosum parameters on the white matter region, and the corticospinal tract parameters as image features.

[0051] Specifically, in this embodiment of the invention, after extracting image features, the segmentation results of the white matter region, gray matter region, and cerebrospinal fluid region, along with the extracted image features, are used by doctors for subsequent dynamic assessment of pediatric brain development. Specifically: a third ratio between the volume of the gray matter region and the volume of the white matter region, and a fourth ratio between the volume of the cerebrospinal fluid region and the volumes of the white matter and gray matter regions are calculated; if the third ratio is greater than a first predetermined threshold, it is used by doctors to determine if the pediatric brain has delayed myelination; if the fourth ratio is greater than a second predetermined threshold, it is used by doctors to determine if the pediatric brain has brain atrophy; corpus callosum parameters and corticospinal tract parameters are extracted from the white matter region; if the corpus callosum parameter is less than the third predetermined threshold or the corticospinal tract parameter is greater than the fourth predetermined threshold, it is used by doctors to determine if the pediatric brain has motor pathway damage; if the growth rate of the gray matter region volume is lower than the standard deviation of the growth rate of the gray matter region volume in children of the same age, or if the annual growth rate of the corpus callosum parameter in the white matter region is less than a fifth predetermined threshold, it is used by doctors to determine if the pediatric brain has high-risk developmental deviations.

[0052] Specifically, this embodiment of the invention performs a dynamic assessment of pediatric brain development based on the segmentation results. The first to fifth predetermined thresholds can be determined according to actual circumstances, and this embodiment of the invention does not impose any limitations on them. For example, in this embodiment of the invention, the first predetermined threshold is 1.1, the second predetermined threshold is 15%, the third predetermined threshold is 0.5, the fourth predetermined threshold is 1.2, and the fifth predetermined threshold is 8%.

[0053] More specifically, in this embodiment of the invention, three-dimensional quantization is first performed on the segmented brain region mask: the volume ratio of the whole brain gray matter region to the white matter region (GM / WM Ratio) is calculated. If the ratio is >1.1 in a 2-year-old child (normal <0.9), it indicates delayed myelination. Simultaneously, the cerebrospinal fluid percentage (CSF%) is monitored to indicate brain atrophy if it is >15%, where the CSF percentage refers to the ratio between the CSF region and the white matter and gray matter regions. Next, cortical development is analyzed, and the thickness of the frontal / temporal cortex is measured. In children with developmental delays, the monthly thickness increase is <0.1mm (normal >0.15mm), and the gyriary complexity index (GI) <2.2 indicates delayed maturation. DTI parameters are extracted from the white matter region mask: corpus callosum FA <0.5 (normal >0.6 in 6-month-olds) or corticospinal tract MD >1.2 (× (mm² / s) indicates motor pathway damage. Finally, longitudinal data is integrated for modeling: if the gray matter volume growth rate is less than 2 standard deviations for age-matched individuals or the annual growth rate of white matter corpus callosum FA is <8%, the system automatically marks high-risk developmental deviations and triggers targeted interventions (e.g., initiating rehabilitation treatment when white matter corpus callosum FA decreases by >10% twice consecutively). Through the dynamic correlation of three types of indicators—volume, microstructure, and developmental rate—individualized brain development trajectory assessment is achieved.

[0054] This invention constructs a cerebrospinal fluid (CSF) coefficient by combining the pixel grayscale values ​​of the first weighted image, the second weighted image, and the grayscale information of the multi-gradient direction attenuated weighted image. This coefficient can simultaneously reflect the grayscale differences and diffusion characteristics of gray matter and CSF, effectively avoiding the problem of "confusion between unmyelinated white matter fluid and CSF fluid signals," making the boundary between the CSF region and the remaining brain region clearer and significantly reducing the CSF missegmentation rate. After segmenting the remaining brain region, the white matter coefficient is further calculated by combining the grayscale values ​​of the multi-gradient direction attenuated weighted image and the dual-weighted image. The diffusion of water molecules in white matter has a certain directionality, while the diffusion of water molecules in gray matter is more random. This difference in diffusion characteristics is reflected in the attenuated weighted image. Combined with the grayscale differences of gray and white matter in the dual-weighted image, the white matter coefficient can accurately distinguish between white matter and gray matter, solving the problem of "blurred boundaries caused by incomplete myelination of gray and white matter." The technical solution provided by this invention makes it easier to segment tissues in MRI images, thereby improving the accuracy and reliability of segmentation results for different tissue regions of the brain in pediatric brain MRI images.

[0055] Example 2: Corresponding to the MRI image feature extraction method for dynamic assessment of pediatric brain development provided in the above embodiments, based on the same technical concept, this embodiment of the invention also provides an electronic device for performing the above-described MRI image feature extraction method for dynamic assessment of pediatric brain development. Figure 4 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention, as shown below. Figure 4 As shown. Electronic devices can vary considerably due to differences in configuration or performance, and may include one or more processors 401 and memories 402. The memory 402 stores computer programs that can run on the processor 401, and the processor 401 executes the programs stored in the memory 402 to achieve the above. Figure 1 The various steps in the method embodiment are described. The memory 402 can be temporary or persistent storage. The application stored in the memory 402 may include one or more modules (not shown), each module may include a series of computer-executable instructions for the electronic device.

[0056] Furthermore, the processor 401 may be configured to communicate with the memory 402 and execute a series of computer-executable instructions stored in the memory 402 on the electronic device. The electronic device may also include one or more power supplies 403, one or more wired or wireless network interfaces 404, one or more input / output interfaces 405, and one or more keyboards 406.

[0057] Specifically, in this embodiment, the electronic device includes a processor, a communication interface, a memory, and a communication bus; wherein, the processor, the communication interface, and the memory communicate with each other via the bus; the memory is used to store computer programs; and the processor is used to execute the programs stored in the memory to achieve the above. Figure 1 The various steps in the method embodiments are the same as those in the above method embodiments, and have the same beneficial effects. To avoid repetition, the embodiments of the present invention will not be described again here.

[0058] It should be noted that the electronic device provided in this embodiment of the invention and the MRI image feature extraction method for dynamic assessment of pediatric brain development provided in this embodiment of the invention are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the aforementioned implementation of the MRI image feature extraction method for dynamic assessment of pediatric brain development, and has the same or similar beneficial effects. Repeated parts will not be described again.

[0059] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0061] This invention also provides a computer-readable storage medium that stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform... Figure 1 The methods disclosed in the embodiments shown achieve the functions and beneficial effects of the methods in the preceding method embodiments, and will not be repeated here.

[0062] The computer-readable storage media include read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for extracting MRI image features for dynamic assessment of pediatric brain development, characterized in that, include: The cerebrospinal fluid coefficient of the child's brain is determined based on the gray values ​​of each pixel in the first and second weighted images of the MRI image of the child's brain, and the gray values ​​of each pixel in the attenuated weighted images of different gradient directions in the MRI image. The MRI image is divided into a cerebrospinal fluid region and a remaining brain region based on the cerebrospinal fluid coefficient of each pixel in the pediatric brain. The white matter coefficient of each pixel in the remaining brain region is determined based on the gray values ​​of each pixel in the attenuated weighted images in different gradient directions, and the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image, respectively. Based on the white matter coefficient, the white matter and gray matter regions of the remaining brain region are segmented, and image features of the white matter and gray matter regions are extracted. The determination of the cerebrospinal fluid coefficient of the pediatric brain includes: The liquid coefficient of each pixel is determined based on the gray values ​​of each pixel in the first weighted image and the second weighted image of the MRI image of the child's brain. The attenuation capability of each pixel in each gradient direction is determined based on the gray value of each pixel in the attenuation weighted image of different gradient directions in the MRI image. The attenuation rate of each pixel is determined based on the attenuation capability of each pixel in each gradient direction. The cerebrospinal fluid coefficient of the pediatric brain is determined based on the fluid coefficient and the attenuation rate.

2. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 1, characterized in that, The determination of the liquid coefficient for each pixel based on the grayscale values ​​of each pixel in the first weighted image and the second weighted image of the pediatric brain MRI image includes: A first ratio is determined between the gray value of each pixel in the second weighted image of the MRI image of the child's brain and the gray value in the first weighted image. The first ratio is then normalized to obtain the liquid coefficient.

3. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 1, characterized in that, The step of determining the attenuation capability of each pixel in each gradient direction based on the gray value of each pixel in the attenuation weighted image of different gradient directions in the MRI image includes: Using the weakened weighted image with a preset magnetic field gradient as a reference image, weakened weighted images of the reference image under different gradient directions are obtained. Align the reference image with the attenuated weighted images of the reference image under different gradient directions, and arrange the aligned attenuated weighted images to obtain an attenuated weighted image sequence. The sequence of grayscale value changes of each pixel in the weakened weighted image sequence is taken as the grayscale sequence of each pixel. The first gray value in the grayscale sequence of each pixel is taken as the initial value, and the second ratio of the initial value to the other gray values ​​in the grayscale sequence is taken as the decay rate in different gradient directions in the grayscale sequence of each pixel.

4. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 1, characterized in that, Determining the attenuation rate of each pixel based on its attenuation capability in each gradient direction includes: The average value of the attenuation ability of each pixel in each gradient direction is normalized to obtain the attenuation rate of each pixel.

5. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 1, characterized in that, The step of determining the white matter coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the attenuated weighted images in different gradient directions, and the gray values ​​of each pixel in the remaining brain region in the first weighted image and the second weighted image respectively, includes: Based on the gray values ​​of each pixel in the attenuation weighted image of the remaining brain region in the MRI image under different gradient directions, the attenuation capability of each pixel in the remaining brain region under each gradient direction is determined. The fiber bundle coefficient of each pixel in the remaining brain region is determined based on the minimum value of the attenuation ability of each pixel in each gradient direction and the attenuation ability of each pixel in the remaining brain region in other gradient directions. The lipid coefficient of each pixel in the remaining brain region is determined based on the gray values ​​of each pixel in the first weighted image and the second weighted image. Based on the fiber bundle coefficient and the lipid coefficient, the white matter coefficient of each pixel in the remaining brain region is determined.

6. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 5, characterized in that, The step of determining the fiber bundle coefficient of each pixel in the remaining brain region based on the minimum attenuation capability of each pixel in each gradient direction and the attenuation capability of each pixel in the remaining brain region in other gradient directions includes: The attenuation ability difference coefficient is determined based on the minimum value of the attenuation ability of each pixel in the remaining brain region in other gradient directions and the attenuation ability of each pixel in the remaining brain region in each gradient direction. The fiber bundle coefficient is determined based on the minimum value of the weakening capacity difference coefficient and the weakening capacity.

7. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 5, characterized in that, The step of determining the lipid coefficient of each pixel in the remaining brain region based on the gray values ​​of each pixel in the first weighted image and the second weighted image includes: Determine the grayscale difference coefficient of each pixel in the remaining brain region between the grayscale values ​​of the first weighted image and the second weighted image, and the sum of the grayscale values ​​of each pixel in the remaining brain region between the grayscale values ​​of the first weighted image and the second weighted image; The lipid coefficient is determined based on the grayscale difference coefficient and the sum value.

8. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 5, characterized in that, The determination of the white matter coefficient of each pixel in the remaining brain region based on the fiber tract coefficient and the lipid coefficient includes: Determine the first weighting factor of the fiber bundle coefficient and the second weighting factor of the lipid coefficient; The white matter coefficient of each pixel in the remaining brain region is determined based on the fiber bundle coefficient, the lipid coefficient, the first weighting coefficient, and the second weighting coefficient.

9. The MRI image feature extraction method for dynamic assessment of pediatric brain development according to claim 1, characterized in that, The segmentation of the white matter and gray matter regions of the remaining brain region based on the white matter coefficient, and the extraction of image features from the white matter and gray matter regions, includes: Based on the Otsu threshold segmentation method, the first threshold part of the white matter coefficient of each pixel is recorded as the white matter region, and the second threshold part is recorded as the gray matter region. The first threshold part is greater than the second threshold part. Extract the volume of the gray matter region, the volume of the white matter region, the corpus callosum parameters and corticospinal tract parameters on the white matter region; The volume of the gray matter region, the volume of the white matter region, the corpus callosum parameters and corticospinal tract parameters on the white matter region are used as the image features.

Images