A prediction system for sulcal abnormalities

Abstract

The present invention belongs to the field of medical image processing and analysis, and particularly relates to a prediction system for sulcal abnormalities, comprising: acquiring MRI brain image data of autistic children from multiple acquisition sites; reconstructing the white matter surface of the children's brain MRI image data; extracting sulcal depressions based on the reconstructed white matter surface of the brain, and projecting them onto a standard sphere space to construct a sulcal map; extracting sulcal features based on the constructed sulcal map to reflect the geometric characteristics and topological properties of the sulci; an abnormal sulcal pattern analysis unit constructs a sulcal abnormality value based on the sulcal abnormality based on the obtained geometric characteristics and topological properties, and sets an alarm threshold, and makes a prediction based on the relationship between the sulcal abnormality value and the alarm threshold. The present invention derives the sulcal pattern abnormality of autistic children based on the geometric characteristics and topological properties of the abnormal sulci of the autistic children, and the present invention still has a good recognition effect when the sulcal pattern changes and fluctuations in the early stages of autism are small.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing and analysis, and in particular relates to a prediction system for brain sulcus abnormalities. Background Art

[0002] The brain is a complex dynamic system, and its developmental changes are closely related to human higher-order cognitive abilities. Abnormal cerebral cortical morphology can be observed in many neurodevelopmental disorders and cortical malformations. The sulcal pattern not only reflects the various properties of the cerebral cortex, but also reflects the spatial, geometric, and topological relationships between multiple sulci. Studying this pattern helps us understand the optimized organization and arrangement of cortical functional areas and underlying white matter fiber connections. Changes in the global sulcal pattern reflect changes in early brain development and manifest as individual differences in cognitive function, personality traits, or mental disorders. Therefore, studying sulcal patterns has important academic significance and clinical application value for in-depth understanding of the normal developmental mechanisms of the brain or the pathogenesis of neurodevelopmental disorders.

[0003] Early studies of sulcal patterns primarily relied on qualitative visual observation. Abnormal sulcal arrangement, connectivity, or interruptions have been found in a variety of disorders, including schizophrenia, temporal lobe epilepsy, and Turner syndrome. However, this type of qualitative analysis fails to quantify the relationships between sulci and is time-consuming and labor-intensive, severely limiting the amount of data available for analysis. Therefore, there is an urgent need to develop automated methods for quantitatively analyzing sulcal patterns. Sun et al. first proposed using three-dimensional invariant moments to represent sulci and then employed an agglomerative clustering algorithm to quantitatively analyze the dominant sulcal patterns. They then used geodesic distances between sulci after affine registration as a similarity measure, which offers better shape description capabilities. However, this method is highly dependent on registration accuracy and is very sensitive to cortical morphology. In recent years, researchers have proposed using sulcal indentations to study changes in brain folding. Sulcal indentations, defined as the deepest local locations of sulci, are associated with brain function under strict genetic regulation and can be extracted using the watershed method. By modeling local sulcal depressions as sulcal maps to analyze sulcal patterns, the geometric and topological characteristics of the sulci can be effectively reflected. This has successfully established a relationship with genetic factors and described the sulcal characteristics of patients. Although current sulcal pattern analysis methods based on sulcal maps have achieved some success, they still have some shortcomings. They require strong prior assumptions, require a large number of parameters to be debugged, or have high computational costs. How to quickly and accurately analyze sulcal patterns across the entire brain is a key issue that needs to be addressed in this invention.

[0004] In addition, autism is a neurodevelopmental disorder characterized by social interaction disorders, language communication disorders, and behavioral stereotyping. According to the Wucailu Children's Behavior Correction Center, the prevalence of autism in China is approximately 1%, making it the most common disease among children. Studies have found that autistic patients exhibit abnormalities in indicators such as gyrification index, cortical thickness, and surface area. In addition, based on sulcal shape analysis, it was found that autistic patients have abnormalities in the depth and length of the central sulcus, intraparietal sulcus, and medial frontal sulcus, and this abnormality is closely related to the severity of the disease. However, whether and how the sulcal patterns of the autistic population change has not been reported.

[0005] In summary, the existing technical problems are: the current sulcus pattern analysis methods are insufficient in speed and accuracy, there is no research on the sulcus patterns of autism, and it is impossible to clearly identify the sulcus patterns in the early stages of autism when the changes and fluctuations are small. Summary of the Invention

[0006] To solve the above technical problems, the present invention proposes a system for predicting sulcus abnormalities, characterized in that the system includes an abnormal sulcus pattern detection unit, and the abnormal sulcus pattern detection process includes the following steps:

[0007] S1: Acquire brain MRI imaging data of children with autism from multiple acquisition sites;

[0008] S2: Exclude subjects from the acquired MRI imaging data;

[0009] S3: Reconstruct the brain white matter surface based on the MRI image data of the children's brain after the subjects are excluded, and obtain the reconstructed brain white matter surface;

[0010] S4: Extract sulci and concavities based on the reconstructed brain white matter surface, and project the extracted sulci and concavities into the standard spherical space to construct a sulcus map;

[0011] S5: Extract the features of the sulci based on the constructed sulci map to obtain the geometric characteristics and topological properties of the sulci;

[0012] S6: The abnormal sulcus pattern detection unit combines the obtained geometric characteristics and topological properties to construct a sulcus abnormality value based on the sulcus abnormality, sets an alarm threshold, and makes a prediction based on the relationship between the sulcus abnormality value and the alarm threshold.

[0013] Preferably, the acquired MRI image data is subjected to subject exclusion, specifically including:

[0014] 1) MRI imaging data collected at sites with fewer than 20 subjects were excluded;

[0015] 2) MRI imaging data of patients with Asperger syndrome and unspecified pervasive developmental disorder in autism spectrum disorder were excluded;

[0016] 3) MRI imaging data with severe image artifacts and head motion were excluded.

[0017] Preferably, the brain white matter surface reconstruction is performed on the MRI image data of the children's brain after the subjects are excluded, specifically including:

[0018] A filtering-based method is used to correct the unevenness of brain MRI images, remove non-brain tissue parts in the brain images, obtain a uniform brain MRI image, and divide the gray matter and white matter of the brain tissue according to the color intensity of the gray matter and white matter in the brain MRI image, and reconstruct the white matter surface of the brain white matter.

[0019] Preferably, the S4 specifically includes:

[0020] A depth potential energy function is used to estimate the depth of each vertex on the reconstructed brain white matter surface to obtain a depth map. Based on the depth map, a watershed algorithm is used to divide the white matter surface into multiple sulcal basins, and the deepest position of each basin is located to obtain sulcal depressions. A surface-based registration method is used to map the white matter surfaces of all participating subjects into a common spherical space. The white matter surface in the spherical space is uniformly sampled using the Fibonacci point set to obtain reference points on the white matter surface. Overlapping circles are used to uniformly sample the reference points, and each overlapping neighborhood of the Fibonacci points is determined. A sulcal map based on sulcal depressions is constructed based on each overlapping neighborhood of the Fibonacci points. The nodes of the map are the sulcal depressions in the neighborhood, and the edges of the map are the adjacent relationships between sulcal depressions.

[0021] Preferably, the S5 specifically includes:

[0022] For each node in the sulcus map, the position, sulcus depth, boundary length and area of the sulcus basin of the node are extracted. The geometric characteristics of the sulcus map are obtained based on the coordinates, sulcus depth, boundary length and area of the sulcus basin at the node, and the number of connecting edges of the node is extracted to obtain the topological properties of the sulcus map.

[0023] Preferably, the degree of abnormality of the sulcus pattern of autistic children is obtained by combining the obtained geometric characteristics and topological properties, specifically including:

[0024] According to the geometric characteristics and topological properties of the sulcus map, the least squares support vector machine is used to perform discriminant analysis on the sulcus maps in different neighborhoods of each Fibonacci point:

[0025] Constrained to Xw+b=y+∈

[0026] Get the weights of all neighbors of the Fibonacci point: w = Ay

[0027] Calculate the incentive a of all neighbors of the Fibonacci point by weight w:

[0028] a=sCov(X)w

[0029] Combining the incentives and weights of all neighborhoods of the Fibonacci point, we get the outlier value k behind the brain i :

[0030]

[0031] Where X represents the matrix composed of the geometric characteristics and topological properties of the extracted sulcus map, y represents the vector composed of clinical variables, α represents the hyperparameter that controls the fitting effect, ‖‖ represents the regularization operation, ∈ represents the relaxation variable, b represents the bias, s represents the scaling factor, and s>0, that is, a is positively correlated with Cov(X)w.

[0032] Preferably, an alarm threshold is set, and prediction is made based on the relationship between the sulcus abnormality value and the alarm threshold, specifically including: setting the alarm threshold to 0.05, when the sulcus abnormality value is greater than the alarm threshold, it indicates autism sulcus abnormality, and a prediction signal is given.

[0033] Beneficial effects of the present invention: Based on the sulcus pattern analysis method, the present invention observes the abnormalities in the geometric and topological properties of the brain sulci in disease states. The present invention has advantages in speed and accuracy, and provides a new way to reveal the abnormal cerebral cortical morphology of autism. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of the present invention;

[0035] Figure 2 This is a diagram showing an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] A prediction system for sulcal abnormalities, such as Figure 2 As shown, the system includes an abnormal sulcus pattern detection unit, and the abnormal sulcus pattern detection process is as follows: Figure 1 The following steps are shown:

[0038] S1: Acquire brain MRI imaging data of children with autism from multiple acquisition sites;

[0039] S2: Exclude subjects from the acquired MRI imaging data;

[0040] S3: Reconstruct the brain white matter surface based on the MRI image data of the children's brain after the subjects are excluded, and obtain the reconstructed brain white matter surface;

[0041] S4: Extract sulci and concavities based on the reconstructed brain white matter surface, and project the extracted sulci and concavities into the standard spherical space to construct a sulcus map;

[0042] S5: Extract the features of the sulci based on the constructed sulci map to obtain the geometric characteristics and topological properties of the sulci;

[0043] S6: The abnormal sulcus pattern analysis unit combines the obtained geometric characteristics and topological properties to construct a sulcus abnormality value based on the sulcus abnormality, sets an alarm threshold, and makes a prediction based on the relationship between the sulcus abnormality value and the alarm threshold.

[0044] The acquired MRI imaging data were excluded from the following subjects:

[0045] 1) MRI imaging data collected at sites with fewer than 20 subjects were excluded;

[0046] 2) MRI imaging data of patients with Asperger syndrome and unspecified pervasive developmental disorder in autism spectrum disorder were excluded;

[0047] 3) MRI imaging data with severe image artifacts and head motion were excluded.

[0048] The brain white matter surface reconstruction was performed on the MRI image data of the children's brain after the subjects were excluded, including:

[0049] A filtering-based method is used to correct the unevenness of brain MRI images, remove non-brain tissue parts in the brain images, obtain a uniform brain MRI image, and divide the gray matter and white matter of the brain tissue according to the color intensity of the gray matter and white matter in the brain MRI image, and reconstruct the white matter surface of the brain white matter.

[0050] The filtering-based method is used to correct the inhomogeneity of the brain MRI image, which is expressed as:

[0051] V(x)=log(v(x))-LPF(logv(x))+C N

[0052] Among them, V(x) represents the corrected image, v(x) represents the input uneven image, LPF(logv(x)) represents the estimated bias field, C NRepresents the normalization coefficient, which makes up for the low-frequency loss of the input image.

[0053] After correcting for intensity variations due to brain MRI image inhomogeneities, normalized intensity images were created from the MRI dataset. Extracerebral voxels were removed using a "skull stripping" procedure, and the grayscale-normalized, skull-stripped images were segmented based on grayscale interface geometry. Cutting planes were then calculated to separate the cerebral hemispheres and disconnect subcortical structures from cortical components. Any internal voids within the white matter components were filled, resulting in a filled volume for each cortical hemisphere. Finally, the resulting volume was overlaid with a triangular tessellation and deformation to produce an accurate and smooth representation of the gray-white interface and neural cortical surface.

[0054] The S4 specifically includes:

[0055] A depth potential energy function was used to estimate the depth of each vertex on the reconstructed brain white matter surface to obtain a depth map. Based on the depth map, a watershed algorithm was used to divide the white matter surface into multiple sulcal basins, and the deepest position of each basin, i.e., the sulcal concavity, was located. The white matter surfaces with clear sulcal concavities were located. A surface-based registration method was used to map the white matter surfaces of all participating subjects into a common spherical space. The white matter surfaces projected into the spherical space were uniformly sampled using the Fibonacci point set to obtain reference points for statistical mapping. The reference points were uniformly sampled using overlapping circles to determine each overlapping neighborhood of the Fibonacci points. A sulcal map based on sulcal concavities was constructed based on each overlapping neighborhood of the Fibonacci points. The nodes of the map were the sulcal concavities in the neighborhood, and the edges of the map were the proximity relationships between sulcal concavities. The corresponding relationships between the nodes in the sulcal maps of different subjects were determined by the distance between their positions in the spherical space.

[0056] The depth potential energy function is used to estimate the depth of each vertex on the reconstructed brain white matter surface, which is expressed as:

[0057] J=J p +λ A J A +λ d J d

[0058] Among them, J represents the depth of the vertex, J p represents the metric depth for alignment based on cortical depth and curvature information, λ A and λ d As the regularization coefficient, J A represents a topological protection term, J d Indicates the amount of control metric distortion.

[0059] The surface-based registration method specifically includes: first, using information obtained from the white matter surface where clear brain sulci have been located to record the cortical folding pattern; then projecting the recorded cortical folding pattern back to the volume domain of spherical space and using linear elastic constraints to diffuse the folded cortex; this process generates a nonlinear displacement field, which is used to map the target and each subject's cortical surface to spherical space.

[0060] The watershed algorithm processing process includes:

[0061] First, the depth values are sorted and a depth map sorted by depth is created. The vertex with the largest depth value is defined as the initial vertex of the sulcus basin. If the next vertex in the depth map is a neighbor of the initial vertex, it is added to the basin; if not, a new initial vertex is created; thus, multiple sulcus basins are obtained.

[0062] Preferably, the S5 specifically includes:

[0063] For each node in the sulcus map, the position, sulcus depth, boundary length and area of the sulcus basin of the node are extracted. The geometric characteristics of the sulcus map are obtained based on the coordinates, sulcus depth, boundary length and area of the sulcus basin at the node, and the number of connecting edges of the node is extracted to obtain the topological properties of the sulcus map.

[0064] Preferably, the degree of abnormality of the sulcus pattern of autistic children is obtained by combining the obtained geometric characteristics and topological properties, specifically including:

[0065] According to the geometric characteristics and topological properties of the sulcus map, the least squares support vector machine (LSSVM) was used to perform discriminant analysis on the sulcus maps in different neighborhoods of each Fibonacci point:

[0066] Constrained to Xw+b=y+∈

[0067] Get the weights of all neighbors of the Fibonacci point: w = Ay

[0068] Calculate the incentive a of all neighbors of the Fibonacci point by weight w:

[0069] a=sCov(X)w

[0070] Combining the incentives and weights of all neighborhoods of the Fibonacci point, we can get the outlier value k of the sulcus. i :

[0071]

[0072] Where X represents the matrix composed of the geometric characteristics and topological properties of the extracted sulcus map, y represents the vector composed of clinical variables, α represents the hyperparameter that controls the fitting effect, ∈ represents the slack variable, b represents the bias, s represents the scale factor, and s>0, that is, a is positively correlated with Cov(X)w.

[0073] Preferably, an alarm threshold is set, and a prediction is made based on the relationship between the sulcus abnormality value and the alarm threshold, specifically including: setting the sulcus abnormality change degree threshold to 0.05, that is, the alarm threshold is 0.05, when the sulcus abnormality value is greater than the sulcus abnormality change threshold, it indicates autism sulcus abnormality, and a prediction signal is given.

[0074] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A prediction system for sulcus abnormalities, characterized in that: The system includes an abnormal sulcus pattern detection unit. The abnormal sulcus pattern detection process includes the following steps: S1: Acquire brain MRI imaging data of children with autism from multiple acquisition sites; S2: Exclude subjects from the acquired MRI imaging data; S3: Reconstruct the brain white matter surface based on the MRI image data of the children's brain after the subjects are excluded, and obtain the reconstructed brain white matter surface; S4: Extract sulci and concavities based on the reconstructed brain white matter surface, and project the extracted sulci and concavities into the standard spherical space to construct a sulcus map; S5: Extract the features of the sulci based on the constructed sulci map to obtain the geometric characteristics and topological properties of the sulci; For each node in the sulcus map, the position, sulcus depth, boundary length and area of the sulcus basin of the node are extracted. The geometric characteristics of the sulcus map are obtained based on the coordinates, sulcus depth, boundary length and area of the sulcus basin at the node, and the number of connecting edges of the node is extracted to obtain the topological properties of the sulcus map. S6: The abnormal sulcus pattern detection unit combines the obtained geometric characteristics and topological properties to construct a sulcus abnormality value based on the sulcus abnormality, sets an alarm threshold, and makes a prediction based on the relationship between the sulcus abnormality value and the alarm threshold; Combining the obtained geometric characteristics and topological properties, the degree of abnormality of the brain sulcus pattern of children with autism is obtained, including: According to the geometric characteristics and topological properties of the sulcus map, the least squares support vector machine is used to perform discriminant analysis on the sulcus map in different neighborhoods of each Fibonacci point: Constrained to Xw+b=y+∈ Get the weights of all neighbors of the Fibonacci point: w = Ay; Calculate the incentive a of all neighbors of the Fibonacci point by weight w: a=sCov(X)w Combining the incentives and weights of all neighborhoods of the Fibonacci point, we can get the outlier value k of the sulcus. i : Where C represents the matrix composed of the geometric characteristics and topological properties of the extracted sulcus map, y represents the vector composed of clinical variables, α represents the hyperparameter that controls the fitting effect, ‖‖ represents the regularization operation, ∈ represents the relaxation variable, b represents the bias, s represents the scaling factor, and s>0, that is, a is positively correlated with Cov(X)w.

2. A prediction system for sulcal abnormality according to claim 1, characterized in that: The acquired MRI imaging data were excluded from the following subjects: 1) MRI imaging data collected at sites with fewer than 20 subjects were excluded; 2) MRI imaging data of Asperger syndrome and unspecified pervasive developmental disorder in patients with autism spectrum disorder were excluded; 3) MRI imaging data with severe image artifacts and head motion were excluded.

3. The system for predicting sulcal abnormalities according to claim 1, wherein: The brain white matter surface reconstruction was performed on the MRI image data of the children's brain after the subjects were excluded, including: A filtering-based method is used to correct the unevenness of brain MRI images, remove non-brain tissue parts in the brain images, obtain a uniform brain MRI image, and divide the gray matter and white matter of the brain tissue according to the color intensity of the gray matter and white matter in the brain MRI image, and reconstruct the white matter surface of the brain white matter.

4. The system for predicting sulcal abnormalities according to claim 1, wherein: The S4 specifically includes: A depth potential energy function is used to estimate the depth of each vertex on the reconstructed brain white matter surface to obtain a depth map. Based on the depth map, a watershed algorithm is used to divide the white matter surface into multiple sulcal basins, and the deepest position of each basin is located to obtain sulcal depressions. A surface-based registration method is used to map the white matter surfaces of all participating subjects into a common spherical space. The white matter surface in the spherical space is uniformly sampled using the Fibonacci point set to obtain reference points on the white matter surface. Overlapping circles are used to uniformly sample the reference points, and each overlapping neighborhood of the Fibonacci points is determined. A sulcal map based on sulcal depressions is constructed based on each overlapping neighborhood of the Fibonacci points. The nodes of the map are the sulcal depressions in the neighborhood, and the edges of the map are the adjacent relationships between sulcal depressions.

5. The system for predicting sulcal abnormalities according to claim 1, wherein: An alarm threshold is set, and prediction is made based on the relationship between the abnormal sulcus value and the alarm threshold, specifically including: setting the alarm threshold to 0.

05. When the abnormal sulcus value is greater than the alarm threshold, it indicates abnormal sulcus of autism, and a prediction signal is given.

Images