A method for segmenting superficial cerebral veins in images

Abstract

The present invention discloses a method for segmenting superficial cerebral veins in an image, which belongs to the technical field of medical image processing and comprises the following steps: S1: selecting several healthy subjects; S2: performing conventional sequence scanning and susceptibility-weighted imaging; S3: SWI raw data processing; S4: constructing a data set; S5: data set preprocessing; S6: model training; S7: model application; S8: model extension application; the present invention combines the development advantages of susceptibility-weighted imaging, constructs a deep learning model based on MinIP images, and realizes preliminary quantitative analysis of the segmentation of morphological features of superficial cerebral veins, which can provide an objective reference method for the evaluation of superficial cerebral veins in clinical neurological diseases in the future. Based on this model, the hypothesis proposed by tissue anatomy that pathological changes of superficial cerebral veins can cause microinfarctions and thus lead to high signals in the white matter of the brain is discovered from an imaging perspective, and the gender difference in the distribution of the number of blood vessels in the superficial cerebral veins is discovered.

Description

Technical Field

[0001] The invention belongs to the technical field of medical image processing, and in particular relates to a method for segmenting superficial cerebral veins in an image. Background Art

[0002] Foreign researchers have pointed out that there may be a potential association between superficial cerebral veins and cerebral ischemic and hypoxic lesions such as microinfarction. Superficial cerebral veins are more likely to become vulnerable points for cerebrovascular diseases than arteries. However, at present, most research on cerebral veins focuses on deep cerebral veins, while research on superficial cerebral veins is relatively small. Therefore, it is necessary to design a method for segmenting superficial cerebral veins in images to solve the above problems. Summary of the Invention

[0003] To solve the problems raised in the above background technology, the present invention provides a method for segmenting superficial cerebral veins in images, which has the characteristic of providing an objective method for reference in the evaluation of superficial cerebral veins in future clinical neurological diseases.

[0004] To achieve the above object, the present invention provides the following technical solution: a method for segmenting superficial cerebral veins in an image, comprising the following steps:

[0005] S1: Select several healthy subjects who are right-handed, have no history of intracranial mass or cranial surgery, and have no central nervous system diseases that may cause cognitive dysfunction and related symptoms;

[0006] S2: Several subjects' heads were fixed with sponge pads to prevent rotation. Conventional sequential scanning and susceptibility-weighted imaging were then performed on the heads of several subjects using a German Siemens Skyra3.0T MRI scanner and a 32-channel head phased array coil.

[0007] S3: The minimum intensity projection image is reconstructed on the obtained SWI image through volume imaging, which is used as the image data of the model construction dataset;

[0008] S4: Collect MinIP image data from all subjects and import them into the MicroDicom DICOM viewer. Unify and fix the window width and window level to observe the visualization of the bilateral superficial cerebral veins at the level of the lateral ventricles in the MinIP image data. Select any three brain slices between the upper and lower edges of the lateral ventricles of all subjects with the clearest visualization of the superficial cerebral veins as samples for delineating the region of interest to obtain the model construction dataset.

[0009] S5: Convert the DICOM format images into the same number of JPG format images as the number of slices selected after the SWI scan. Each image displays the complete image information of a slice. Quickly locate the slice number and the position of the superficial cerebral veins using the position information and the number of scanned slices. Import the selected slice images into the open source software Labelme, and manually outline all visible superficial cerebral veins one by one to obtain the preprocessed model construction dataset.

[0010] S6: Using four-fold cross-validation, the dataset was randomly divided into training, validation, and test sets in a ratio of 6:2:2. The DeepLabV3+ network was trained on the training and validation sets and validated on the test set. The optimal hyperparameter model with the smallest mean error was selected to obtain the automatic quantification model for superficial cerebral veins.

[0011] S7: Process SWI images from several subjects using the same reconstruction parameters. Select the MinIP images from the highest cerebral level at the level of the lateral ventricles and train them in the automatic quantification model for superficial cerebral veins. This will yield the diameter, curvature, and number of superficial cerebral veins in both cerebral hemispheres. The quantitative data for the superficial cerebral veins in both cerebral hemispheres of these subjects will be collated and recorded for statistical analysis.

[0012] S8: Based on the quantitative data of superficial cerebral veins in both cerebral hemispheres of several subjects, the distribution of deep subcortical white matter lesions in all subjects was determined. The subjects were divided into two groups based on whether they had deep subcortical white matter lesions: a group with deep subcortical white matter lesions and a group without deep subcortical white matter lesions.

[0013] Furthermore, in the present invention, in step S3, the reconstruction parameters of the volume imaging are a layer thickness of 20 mm and a layer spacing of 1 mm.

[0014] Furthermore, in the present invention, in step S4, the range of the window width and the window level is 20-40 Hu.

[0015] Furthermore, in the present invention, in step S6, in order to avoid falling into a local optimum, a method of gradually decreasing the learning rate is adopted, that is, the learning rate is adjusted every 100 iterations.

[0016] Furthermore, in step S8 of the present invention, the specific steps of judging the distribution of deep subcortical white matter lesions of the subjects include using SPSS 25.0 statistical software to perform statistical analysis on the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects, using the Mann-Whitney U test to perform inter-group comparison on the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects, and using a binary logistic regression model to analyze the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] This paper combines the development advantages of magnetic susceptibility weighted imaging and constructs a deep learning model based on MinIP images to achieve preliminary quantitative analysis of the segmentation of the morphological characteristics of superficial cerebral veins. It can provide an objective reference method for the evaluation of superficial cerebral veins in future clinical neurological diseases. Based on this model, it was discovered and verified from an imaging perspective that the hypothesis proposed by tissue anatomy that pathological changes in superficial cerebral veins can cause microinfarctions and thus lead to high signals in the white matter of the brain, and at the same time, gender differences in the distribution of the number of blood vessels in the superficial cerebral veins were discovered. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Flowchart of the method for segmenting superficial cerebral veins in an image according to the present invention;

[0020] Figure 2 This is the framework diagram of the neural network model for automatic quantification of superficial cerebral veins of the present invention;

[0021] Figure 3 This is the index diagram of the optimal model of the present invention in each group of experiments;

[0022] Figure 4 This is the M(QR) diagram of the distribution characteristics of the number of superficial cerebral veins in the bilateral cerebral hemispheres in terms of gender;

[0023] Figure 5 This is a diagram showing the distribution characteristics of the number of superficial cerebral veins in both cerebral hemispheres in terms of gender;

[0024] Figure 6 This is the M(QR) diagram comparing the quantitative indices of bilateral superficial cerebral veins of the two groups of the present invention;

[0025] Figure 7 This is a regression test diagram of the present invention between deep subcortical white matter lesions and curvature of the right superficial cerebral vein. DETAILED DESCRIPTION

[0026] 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.

[0027] See also Figure 1-7 The present invention provides the following technical solution: a method for segmenting superficial cerebral veins in an image, comprising the following steps:

[0028] S1: Select several healthy subjects who are right-handed, have no history of intracranial mass or cranial surgery, and have no central nervous system diseases that may cause cognitive dysfunction and related symptoms;

[0029] S2: Several subjects' heads were fixed with sponge pads to prevent rotation. Conventional sequential scanning and susceptibility-weighted imaging were then performed on the heads of several subjects using a German Siemens Skyra3.0T MRI scanner and a 32-channel head phased array coil.

[0030] S3: The minimum intensity projection image is reconstructed on the obtained SWI image through volume imaging, which is used as the image data of the model construction dataset;

[0031] S4: Collect MinIP image data from all subjects and import them into the MicroDicom DICOM viewer. Unify and fix the window width and window level to observe the visualization of the bilateral superficial cerebral veins at the level of the lateral ventricles in the MinIP image data. Select any three brain slices between the upper and lower edges of the lateral ventricles of all subjects with the clearest visualization of the superficial cerebral veins as samples for delineating the region of interest to obtain the model construction dataset.

[0032] S5: Convert the DICOM format images into the same number of JPG format images as the number of slices selected after the SWI scan. Each image displays the complete image information of a slice. Quickly locate the slice number and the position of the superficial cerebral veins using the position information and the number of scanned slices. Import the selected slice images into the open source software Labelme, and manually outline all visible superficial cerebral veins one by one to obtain the preprocessed model construction dataset.

[0033] S6: Using four-fold cross validation, the data set is randomly divided into training set, validation set and test set in a ratio of 6:2:2. The DeepLabV3+ network is used to train on the training set and validation set, and validated on the test set. The optimal hyperparameter model with the smallest average error is selected. The optimal model is shown in the attached manual for each group of experimental indicators. Figure 3 As shown, the automatic quantification model of superficial cerebral veins was obtained;

[0034] S7: Process SWI images from several subjects using the same reconstruction parameters. Select the MinIP images from the highest cerebral level at the level of the lateral ventricles and train them in the automatic quantification model for superficial cerebral veins. This will yield the diameter, curvature, and number of superficial cerebral veins in both cerebral hemispheres. The quantitative data for the superficial cerebral veins in both cerebral hemispheres of these subjects will be collated and recorded for statistical analysis.

[0035] S8: Based on the quantitative data of superficial cerebral veins in both cerebral hemispheres of several subjects, the distribution of deep subcortical white matter lesions in all subjects was determined. The subjects were divided into two groups based on whether they had deep subcortical white matter lesions: a group with deep subcortical white matter lesions and a group without deep subcortical white matter lesions.

[0036] Specifically, in step S3 , the reconstruction parameters of the volume imaging are a slice thickness of 20 mm and a slice spacing of 1 mm.

[0037] Specifically, in step S4, the range of the window width and the window level is 20-40 Hu.

[0038] Specifically, in step S6, in order to avoid falling into a local optimum, the learning rate is gradually reduced, that is, the learning rate is adjusted every 100 iterations.

[0039] Specifically, in step S8, the specific steps of judging the distribution of deep subcortical white matter lesions of the subjects include using SPSS 25.0 statistical software to perform statistical analysis on the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects, using the Mann-Whitney U test to perform inter-group comparison on the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects, and using a binary logistic regression model to analyze the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects.

[0040] As the instruction manual Figure 4 (Note: ** indicates a statistically significant difference at the P < 0.01 level) and Figures 5 (A: Distribution of superficial cerebral veins in men, B: Distribution of superficial cerebral veins in women, C: The number of superficial cerebral veins in both cerebral hemispheres in men was significantly greater than that in women). There were no differences in the distribution of various quantitative indicators of superficial cerebral veins in the two cerebral hemispheres in terms of age and years of education (P>0.05). However, there was a significant difference in the number of superficial cerebral veins in the two cerebral hemispheres in terms of gender (right P = 0.004, left P = 0.000, P < 0.01), showing that the number of superficial cerebral veins in men was significantly greater than that in women, and this was more pronounced in the left cerebral hemisphere.

[0041] As the instruction manual Figure 6 (N-DWML: no subcortical deep white matter lesion group, DWML: deep subcortical white matter lesion group, ** indicates extremely significant statistical difference at the P < 0.01 level) As shown, the curvature of the right superficial cerebral varicose vein in the group with subcortical deep white matter lesions was significantly larger than that in the group without subcortical deep white matter lesions (P = 0.000, P < 0.01). No statistically significant differences were found in the diameter and number of varicose veins in both cerebral hemispheres between the two groups.

[0042] As the instruction manual Figure 7As shown in the data, the results of binary logistic regression analysis showed that deep subcortical white matter lesions were significantly positively correlated with the curvature of the right superficial cerebral vein (regression coefficient was 2.035, P=0.015, P<0.05), that is, the greater the curvature of the right superficial cerebral vein, the higher the probability of developing deep subcortical white matter lesions. The difference was statistically significant. The data results showed that the probability of developing deep subcortical white matter lesions in patients with higher curvature of the right superficial cerebral vein was 7.649 times that of those with lower curvature.

[0043] 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 method for segmenting superficial cerebral veins in an image, characterized in that: The following steps are involved: S1: Select several healthy subjects who are right-handed, have no history of intracranial mass or cranial surgery, and have no central nervous system diseases that may cause cognitive dysfunction and related symptoms; S2: Several subjects' heads were fixed with sponge pads to prevent rotation. Conventional sequence scanning and susceptibility-weighted imaging (SWI) were then performed on the heads of several subjects using a German Siemens Skyra 3.0T MRI scanner and a 32-channel head phased array coil. S3: The minimum intensity projection map MinIP is reconstructed on the obtained SWI image through volume imaging, which serves as the image data of the model construction dataset; S4: Collect MinIP image data from all subjects and import them into the MicroDicom DICOM viewer. Unify and fix the window width and window level to observe the visualization of the bilateral superficial cerebral veins at the level of the lateral ventricles in the MinIP image data. Select any three brain slices between the upper and lower edges of the lateral ventricles of all subjects with the clearest visualization of the superficial cerebral veins as samples for delineating the region of interest to obtain the model construction dataset. S5: Convert the DICOM format images into the same number of JPG format images as the number of slices selected after the SWI scan. Each image displays the complete image information of a slice. Quickly locate the slice number and the position of the superficial cerebral veins using the position information and the number of scanned slices. Import the selected slice images into the open source software Labelme, and manually outline all visible superficial cerebral veins one by one to obtain the preprocessed model construction dataset. S6: Using four-fold cross-validation, the dataset was randomly divided into training, validation, and test sets in a ratio of 6:2:

2. The DeepLabV3+ network was trained on the training and validation sets and validated on the test set. The optimal hyperparameter model with the smallest mean error was selected to obtain the automatic quantification model for superficial cerebral veins. S7: Process SWI images from several subjects using the same reconstruction parameters. Select the MinIP images from the highest cerebral level at the level of the lateral ventricles and train them in the automatic quantification model for superficial cerebral veins. This will yield the diameter, curvature, and number of superficial cerebral veins in both cerebral hemispheres. The quantitative data for the superficial cerebral veins in both cerebral hemispheres of these subjects will be collated and recorded for statistical analysis. S8: Based on the quantitative data of superficial cerebral veins in both cerebral hemispheres of several subjects, the distribution of deep subcortical white matter lesions in all subjects was determined. The subjects were divided into two groups based on whether they had deep subcortical white matter lesions: a group with deep subcortical white matter lesions and a group without deep subcortical white matter lesions.

2. The method for segmenting superficial cerebral veins in an image according to claim 1, wherein: In step S3, the reconstruction parameters of the volume imaging are a slice thickness of 20 mm and a slice spacing of 1 mm.

3. The method for segmenting superficial cerebral veins in an image according to claim 1, wherein: In step S4, the range of the window width and the window level is 20-40 Hu.

4. The method for segmenting superficial cerebral veins in an image according to claim 1, wherein: In step S6, in order to avoid falling into a local optimum, the learning rate is gradually reduced, that is, the learning rate is adjusted every 100 iterations.

5. The method for segmenting superficial cerebral veins in an image according to claim 1, wherein: In step S8, the specific steps of judging the distribution of deep subcortical white matter lesions of the subjects include using SPSS25.0 statistical software to perform statistical analysis on the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects, using the Mann-Whitney U test to perform inter-group comparison on the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects, and using a binary logistic regression model to analyze the quantitative data of the superficial cerebral veins in the bilateral cerebral hemispheres of several subjects.

Images