CN1947655A - Quantitative analysis method for cerebral cortex complexity during treating three-D magnetoencepha-resonance data - Google Patents

Abstract

The invention relates to the technical field of three-dimensional structural nuclear magnetic resonance imaging, and relates to a method for analyzing the distribution of discrete data in three-dimensional space by using the fractal information dimension of discrete voxels to quantitatively analyze the complexity of the shape of the cerebral cortex. After collecting the three-dimensional structural MRI data and undergoing necessary preprocessing, the voxels at the junction of all or part of the cerebral cortex gray matter and cerebrospinal fluid are extracted to form a discrete outer surface of cerebral gray matter. Calculate the fractal information dimension of discrete data in three-dimensional space, and thus obtain a quantitative description of the shape complexity of the whole brain or local gray matter cortex. This method is also perfectly applicable for measuring the shape complexity of gray matter endocortex as well as white matter cortex. The invention is mainly used to analyze the abnormal shape of the cerebral cortex caused by mental and nervous diseases. This method is accurate and fast, and can be completed on a common microcomputer. This method can be widely used in clinical and basic research of brain structure MRI.

Description

A Quantitative Analysis Method for Cortical Complexity in 3D Brain Magnetic Resonance Data Processing

technical field

The invention relates to the technical field of magnetic resonance imaging, in particular to a method for quantitatively analyzing cortex complexity in three-dimensional brain magnetic resonance data processing using geometric fractal dimensions to process brain structural magnetic resonance data.

Background technique

Magnetic Resonance Imaging (MRI) mainly uses nuclear magnetic resonance imaging of hydrogen nuclei for imaging. The size of the MRI signal in the image core is mainly determined by the proton density of the tissue, the longitudinal and transverse relaxation time, the inhomogeneity of the magnetic field and other factors. In the brain tissue, the longitudinal relaxation times of white matter, gray matter and cerebrospinal fluid are relatively different, so a clear structural image is formed, which makes quantitative analysis of brain structure possible.

Long-term autopsy studies have proved that the shape complexity of the cerebral cortex is closely related to the development, pathological changes, and cognitive functions of the human brain. In the vast majority of brain structural imaging studies, counting the volume of cortical parenchyma by counting the number of voxels of white and gray matter in the cortex is the only method to measure the complexity of the cortex.

Thompsom et al., (1998) proposed a metric to fine-tune the analysis of cortical complexity. They first manually calibrated the brain sulcus of the cerebral cortex, then fitted each brain sulcus with a parametric surface, and then calculated the degenerated fractal dimension of each parametric surface, and used the size of the degenerated fractal dimension to measure the cortical sulcus Complexity. This approach has some major drawbacks. First of all, there is no strict and uniform standard for manual calibration of brain sulci, which is greatly affected by human factors, and the manual workload is heavy, which is not conducive to general promotion. Second, the parametric surface is a second-order smooth surface, which cannot express the small shape changes described by the data points, so the accuracy is low. To this end, we propose a method to calculate the fractal information dimension of cortical voxel point sets directly based on image data, so as to quantitatively describe the complexity of the shape of the cortex.

references:

Thompson, P., Moussai, J., Zohoori, S., Goldkorn, A., Khan, A., Mega, M., Small, G., Cummings, J., Toga, A..Cortical variability and asymmetry abnormal aging and Alzheimer's disease. Cerebral Cortex, 1998: Vol.8, pp492-509.

Contents of the invention

A method for processing brain structure nuclear magnetic resonance data, using computer equipment, using three-dimensional fractal information dimension to analyze the distribution of scattered voxel collections on the surface of the cerebral cortex, in order to achieve the purpose of detecting the complexity of the shape of the cerebral cortex, three-dimensional scattered The fractal information dimension of the point cloud, using the theory of renormalization group, quantitatively analyzes the geometric shape described by the voxel point cloud of the cerebral cortex, so as to study the complexity of the cerebral cortex.

The method for processing the structural MRI data of the brain measures the complexity of geometric shapes of the whole brain or partial brain gray matter, white matter cortex, and substructure surfaces.

A method for analyzing the distribution of discrete data in three-dimensional space using the fractal information dimension of discrete voxels to quantify the complexity of the shape of the brain cortex. After collecting the three-dimensional structural MRI data and undergoing necessary preprocessing, the voxels at the junction of all or part of the cerebral cortex gray matter and cerebrospinal fluid are extracted to form a discrete outer surface of cerebral gray matter. Calculate the fractal information dimension of discrete data in three-dimensional space, and thus obtain a quantitative description of the shape complexity of the whole brain or local gray matter cortex. This method is also perfectly applicable for measuring the shape complexity of gray matter endocortex as well as white matter cortex. The invention is mainly used to analyze the abnormal shape of the cerebral cortex caused by mental and nervous diseases. This method is accurate and fast, and can be completed on a common microcomputer. This method can be widely used in clinical and basic research of brain structure MRI.

The core part of the present invention is that, after the three-dimensional brain structure nuclear magnetic resonance image is segmented, the voxels located at the boundary between the gray matter and the cerebrospinal fluid are extracted from the whole brain or the brain area of interest, and these voxels are regarded as three-dimensional space. A set of discrete points. The information dimension of the point set is calculated to quantitatively analyze the cortical shape complexity of the brain area. This method is also suitable for analyzing the shape complexity of white matter cortex. The calculation process of this method is fast and stable, and can be completed on an ordinary microcomputer. Its implementation process can be divided into four steps, as shown in Figure 1.

Description of drawings

Fig. 1 is a flow chart of the realization process of the present invention.

Figure 2 is a definition diagram of the prefrontal cortex.

Figure 3 is an extracted image of the left prefrontal gray matter cortex.

Detailed ways

As shown in Figure 1,

Step 1 (S1), acquisition of brain structure MRI data: acquisition of brain functional structure MRI data is completed on an MRI scanner capable of forming T1 or T2 weighted structural images. There are no special requirements for the specific parameters of the imaging, and the spatial resolution is generally several millimeters, such as 1×1mm ² . , There are no special requirements for the cooperation of the subjects, only the subjects need to be told to close their eyes quietly and keep their heads as still as possible.

Step 2 (S2), preprocessing: After the data collection is completed, conventional preprocessing is generally required, including head movement correction, spatial standardization, etc., but these processes are determined according to the purpose of use and are not necessary. After these basic processes are completed, the cranial part, as well as the cerebellum, brainstem, etc., are removed, and the brain is divided into tissues. The brain regions of interest are then segmented.

Step 3 (S3), extracting the voxels of the cerebral cortex: after the preprocessing is completed, extract the voxels on the surface of the interested cortex, calculate the position of each voxel centroid in the three-dimensional space, and regard these positions as a set of discrete points in space .

Step 4 (S4), calculating the fractal information dimension: performing spatial normalization on the spatially discrete point set obtained in Step 3, and then calculating its fractal information dimension. For the cerebral cortex, this information dimension is a value between 1 and 3. The closer this value is to 3, the greater the complexity of the cortical shape, and conversely, if it is closer to 1, the less complex the cortical shape is.

Figure 2 is the definition of the prefrontal lobe (the left side of the vertical line is the prefrontal lobe), an example of the extraction of the gray matter cortex of the left prefrontal lobe, where (a) is the extracted cortical voxel, and (b) is the centroid of these bodies A discrete set of points formed in three-dimensional space.

Figure 3 is an extraction of the left prefrontal gray matter cortex. a: cortical voxel, b: discrete point set formed by the voxel centroid in 3D space.

The implementation process of the present invention is illustrated below by way of example:

Step 1 (S1), data acquisition: the case group consisted of 12 boys, aged 11.67-14.83, who met the diagnostic criteria of ADHD in the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) (DSM-IV). The control group consisted of 11 boys from neighboring secondary schools, aged 12.67-13.92 years. All subjects were right-handed, with an IQ of >80 points. Pervasive developmental disorders, childhood schizophrenia, emotional disorders, and other common childhood mental and nervous system developmental disorders were excluded. Attention caused by organic diseases were excluded. For those with disabilities, preterm infants (<34 weeks) and those with a history of severe traumatic brain injury (causing coma) were excluded. The purpose of this study was to compare the difference in the shape complexity of the left and right prefrontal gray matter in the brains of normal children and children with ADHD, in order to understand the impact of the lesion on the development of the prefrontal gray matter. All imaging data were acquired on a Siemens Trio 3.0T machine, using standard head coils as transmitting and receiving coils. Routine head scans are done after scout scans to rule out brain lesions. The three-dimensional structure image was scanned by SPGR sequence in sagittal position, TR=1770ms, TE=3.92ms, IT (inversion time)=1100ms, inversion angle=12°, scan field (FOV)=256mm×256mm, matrix 512×512, slice thickness = 1mm, slice interval = 0.5mm, a total of 192 slices, scanning time 15 minutes; the subject was asked to keep his head as still as possible during scanning.

Step 2 (S2), pretreatment: first, use MRIcro ( http://www.psychology.nottingham.ac.uk/staff/crl/mricro.html ) software to automatically remove non-brain tissue. Then, manually remove the brainstem, cerebellum, and residual non-brain tissue. Choose a normal person as a template, and use SPM5b ( http://www.fil.ion.ucl.ac.uk/spm/ ) for rigid body registration. This step is to rotate and translate the position of the orthodox head so that it can be used A unified standard for prefrontal lobe segmentation. The definition of the prefrontal lobe is shown in Figure 1, the brain tissue to the left of the vertical line is the prefrontal lobe. Finally, perform tissue segmentation on the prefrontal cortex.

Step 3 (S3), extracting cerebral cortex voxels: after the preprocessing is completed, extract all voxels on the interface surface of prefrontal gray matter and cerebrospinal fluid, calculate the position of each voxel centroid in three-dimensional space, and regard these positions as a spatially discrete point set. Figure 2 is an example of the extraction of the left prefrontal gray matter cortex, where (a) is the extracted cortical voxels, and (b) is the discrete point set formed by these voxel centroids in three-dimensional space.

Step 4 (S4), calculating the fractal information dimension: performing spatial normalization on the spatially discrete point set obtained in Step 3, and then calculating its fractal information dimension. After the t-statistic test was performed on the left and right sides of the prefrontal cortex of the patient group and the control group, it was found that the shape complexity of the left side of the prefrontal cortex was greater than that of the right side in both patients and normal people. However, the shape complexity of the patient's left prefrontal cortex was significantly lower than that of normal people, and its right side was also lower than normal people, but it did not reach the level of significance. These changes reflect that ADHD disease is indeed closely related to the dysplasia of prefrontal cortex shape complexity.

In conclusion, the precise measurement of the shape complexity of the cerebral cortex using the fractal dimension of discrete point sets has a very broad application prospect. Compared with the traditional method, this method has the following main advantages: (1) It avoids the heavy work of manual sulcus calibration and the subjective error caused by the lack of uniform standards; Parametric fitting, thus ensuring that some small changes in the shape of the cortex are investigated, effectively improving the measurement accuracy; (3) The entire processing process is fast and consumes less memory, which is conducive to general promotion and has very important physiological and clinical significance. significance.

Claims (5)

1. A method for processing the structural nuclear magnetic resonance data of the brain, characterized in that computer equipment is used to analyze the distribution of scattered voxel sets on the surface of the cerebral cortex by using three-dimensional fractal information dimensions, so as to achieve the detection of complex shapes of the cerebral cortex In order to study the complexity of the cerebral cortex, the fractal information dimension of the three-dimensional scattered point cloud is quantitatively analyzed by using the theory of renormalization group to analyze the geometric shape described by the voxel point cloud of the cerebral cortex. 2. The method for processing brain structural MRI data according to claim 1, characterized in that the geometric shape complexity of the whole brain or partial brain gray matter, white matter cortex, and substructure surface is measured. 3. The method for processing brain structure nuclear magnetic resonance data according to claim 1, characterized in that, after tissue segmentation of the three-dimensional brain structure nuclear magnetic resonance image, the whole brain or the brain region of interest is extracted from the gray matter. Consider these voxels as a discrete point set in three-dimensional space, and calculate the information dimension of the point set, so as to quantitatively analyze the complexity of the cortical shape of the brain region. 4. The method for processing the structural MRI data of the brain according to claim 1, characterized in that, after collecting the three-dimensional structural MRI data and undergoing necessary preprocessing, all or part of the cerebral cortex gray matter and cerebrospinal fluid are extracted The voxels at the junction form the discrete gray matter outer surface, and calculate the fractal information dimension of the discrete data in three-dimensional space, thus obtaining a quantitative description of the shape complexity of the whole brain or local gray matter outer cortex. 5. The method for processing brain structural MRI data according to claim 1, characterized in that, step S1, acquisition of brain structural MRI data: the acquisition of brain functional structural MRI data can be performed in T1 or Completed on the magnetic resonance scanner of T2 weighted structural image; Step S2, preprocessing: After the data collection is completed, conventional preprocessing is generally required, including head movement correction and spatial standardization; Step S3, extracting the voxels of the cerebral cortex: after the preprocessing is completed, extract the voxels on the surface of the interested cortex, calculate the position of each voxel centroid in three-dimensional space, and regard these positions as a set of spatially discrete points; Step S4, calculating the fractal information dimension: performing spatial normalization on the spatially discrete point set obtained in step S3, and then calculating its fractal information dimension.

Images