CN111415324A - Classification and identification method of brain lesion image space distribution characteristics based on magnetic resonance imaging - Google Patents
Classification and identification method of brain lesion image space distribution characteristics based on magnetic resonance imaging Download PDFInfo
- Publication number
- CN111415324A CN111415324A CN201910737818.4A CN201910737818A CN111415324A CN 111415324 A CN111415324 A CN 111415324A CN 201910737818 A CN201910737818 A CN 201910737818A CN 111415324 A CN111415324 A CN 111415324A
- Authority
- CN
- China
- Prior art keywords
- image
- brain
- focus
- space
- partition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 57
- 206010051290 Central nervous system lesion Diseases 0.000 title claims abstract description 20
- 238000002595 magnetic resonance imaging Methods 0.000 title claims abstract description 18
- 210000004556 brain Anatomy 0.000 claims abstract description 60
- 230000003902 lesion Effects 0.000 claims abstract description 42
- 238000012216 screening Methods 0.000 claims abstract description 16
- 230000011218 segmentation Effects 0.000 claims abstract description 11
- 238000000605 extraction Methods 0.000 claims abstract description 7
- 238000010801 machine learning Methods 0.000 claims abstract description 4
- 238000005192 partition Methods 0.000 claims description 55
- 210000004885 white matter Anatomy 0.000 claims description 11
- 210000000877 corpus callosum Anatomy 0.000 claims description 6
- 210000001259 mesencephalon Anatomy 0.000 claims description 6
- 210000004884 grey matter Anatomy 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 210000001175 cerebrospinal fluid Anatomy 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 238000007477 logistic regression Methods 0.000 claims description 4
- 238000010187 selection method Methods 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 241000257303 Hymenoptera Species 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 230000002146 bilateral effect Effects 0.000 claims description 2
- 210000005013 brain tissue Anatomy 0.000 claims description 2
- 238000000546 chi-square test Methods 0.000 claims description 2
- 238000004891 communication Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 claims description 2
- 230000008030 elimination Effects 0.000 claims description 2
- 238000003379 elimination reaction Methods 0.000 claims description 2
- 238000003709 image segmentation Methods 0.000 claims description 2
- 238000007637 random forest analysis Methods 0.000 claims description 2
- 238000012706 support-vector machine Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 claims description 2
- 238000011160 research Methods 0.000 abstract description 6
- 108090000623 proteins and genes Proteins 0.000 abstract description 5
- 208000014644 Brain disease Diseases 0.000 abstract description 2
- 238000012545 processing Methods 0.000 abstract description 2
- 238000004458 analytical method Methods 0.000 abstract 2
- 201000010099 disease Diseases 0.000 description 8
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 8
- 108010036280 Aquaporin 4 Chemical group 0.000 description 4
- 102000012002 Aquaporin 4 Human genes 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000012850 discrimination method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 208000035977 Rare disease Diseases 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 208000029067 Neuromyelitis optica spectrum disease Diseases 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Radiology & Medical Imaging (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention belongs to the technical field of image processing and application, and particularly relates to a classification and identification method for brain lesion image space distribution characteristics based on magnetic resonance imaging. The method mainly comprises the steps of lesion segmentation, individual image registration, space standardization, standard space template individuation, lesion space distribution characteristic extraction, characteristic screening, modeling and the like, and is characterized in that a set of brain lesion image space distribution characteristic set analysis method is constructed through various characteristic analyses of lesions in an individual space and a standard space, and on the basis, the characteristic screening and modeling are carried out by using machine learning. The method can be used for carrying out brain lesion image classification and identification of different brain diseases or brain states caused by different antibodies, different genes and the like by using the brain magnetic resonance image, and provides effective guidance for clinic and scientific research.
Description
Technical Field
The invention belongs to the field of image processing and application, and particularly relates to a classification and identification method for brain lesion image space distribution characteristics based on magnetic resonance imaging.
Background
In the prior art, in the classification and identification methods of different brain diseases or brain state images, the brain magnetic resonance imaging technology plays an important role in displaying excellent brain focus due to non-invasiveness, timeliness and the like. In clinical practice, doctors usually summarize and summarize different characteristics of lesions of different diseases shown on images through long-term clinical experience, and visually classify, identify and report the lesions. However, the classification and identification based on experience has the defects of low efficiency, difficulty in finding new characteristics of lesions, difficulty in automatically combining a plurality of characteristics and the like, and the accuracy and efficiency of the classification and identification of brain lesion images are greatly reduced in the case of different diseases such as genes, antibodies and the like, different symptoms or states, novel disease subtypes lacking experience or clinical rare diseases with similar clinical manifestations.
The development of the imaging omics and the deep learning method in recent years provides a thought for classifying and identifying brain lesion images from the data analysis perspective. In general, the deep learning method [1] generally requires a large amount of data, and the result often has no interpretability, so that the method is not suitable for small sample amount and exploration research stage of classification problems of similar diseases, rare diseases, new subtypes and the like; the image omics method [2] performs quantitative feature extraction based on image focus gray information, including statistical features, texture features, filtering features and the like, and then adopts a machine learning method to perform model construction; as the features are artificially defined and extracted, the result is interpretable; however, traditional imaging omics tend to focus on features of the lesion itself, rather than its spatial distribution in the brain. In fact, according to clinical and research experiences, the spatial position and distribution characteristics of the focus of diseases caused by different genotypes or antibodies and the like in the brain often have important differences, so that the method has important significance in classification and identification.
At present, no systematic classification and identification method of brain lesion space distribution characteristics based on magnetic resonance imaging exists. Some studies have reported attention to some spatial distribution features, but are generally identified visually by a physician or researcher (e.g., the presence or absence of a lesion near the ventricles), followed by calibration and induction counting [3 ]. On one hand, the extracted features are extremely limited and are not objective enough, and different judgment results of different doctors or researchers are different; on the other hand, automation is not realized, and the efficiency is low.
Based on the defects of the existing classification identification technology in the aspect of focus spatial distribution characteristics, the inventor of the application intends to provide a classification identification method of the spatial distribution characteristics of brain focus images based on magnetic resonance imaging, and intends to obtain a model and a discrimination method suitable for classification identification by calculating and analyzing the multi-type spatial distribution characteristics of the brain focus images based on magnetic resonance imaging, and especially provides effective guidance for clinic and scientific research in the aspect of different disease classifications caused by different genes or antibodies and the like.
The prior art related to the present invention is:
[1]Najafabadi M M，Villanustre F，Khoshgoftaar T M，et al.Deep learningapplications and challenges in big data analytics[J].Journal of Big Data，2015，2(1)：1.
[2]Ma X，Zhang L，Huang D，et al.Quantitative radiomic biomarkers fordiscrimination between neuromyelitis optica spectrum disorder and multiplesclerosis[J].Journal of Magnetic Resonance Imaging，2018.
[3]Jurynczyk M，Geraldes R，Probert F，et al.Distinct brain imagingcharacteristics of autoantibody-mediated CNS conditions and multiplesclerosis[J].Brain，2017，140(3)：617-627.
Disclosure of Invention
The invention aims to provide a classification and identification method of brain lesion image space distribution characteristics based on magnetic resonance imaging based on the defects of the existing classification and identification technology in the aspect of lesion space distribution characteristics, and the method mainly comprises the steps of lesion segmentation, individual image registration, space standardization, standard space template individuation, lesion space distribution characteristic extraction, characteristic screening, modeling and the like (as shown in figure 1). The method of the invention obtains a model and a discrimination method suitable for classification and identification by calculating and analyzing the multi-type spatial distribution characteristics of the magnetic resonance imaging brain lesion image, and provides effective guidance for clinic and scientific research particularly in the aspect of classification of different diseases caused by different genes or antibodies and the like.
Specifically, the classification and identification method of brain lesion image space distribution characteristics based on magnetic resonance imaging comprises the following steps:
1) And focal image segmentation:
1) preparing and selecting brain image data, namely at least two groups of tested images G1 and G2, preparing two images for each tested individual, wherein one image is a 'lesion display image', one mode is selected from clinical brain magnetic resonance images (including but not limited to T1 weighted images, T2 weighted images, F L AIR, DWI, ADC, SWI and the like) of a patient to serve as an image of subsequent lesion segmentation, the other image is a 'brain structural image', and a T1 weighted image is generally selected, wherein the two images can be the same;
1) -2, lesion segmentation: for the 'focus display image', using image segmentation software, a tool kit and a segmentation algorithm, including but not limited to a workstation and segmentation software provided by an MRI manufacturer, MRIcro, MRIcron, MIPAV, ITK-SNAP, MITK, a region growing algorithm and the like, segmenting the whole brain focus, and storing the segmented whole brain focus as a binary image or data to generate a 'focus image';
2) the image registration of the individual, namely rigidly registering the brain structure image of the individual with the lesion display image, wherein available software and methods comprise but are not limited to SPM, FS L and the like;
3) the method comprises the following steps of performing space standardization, namely registering an individual brain structure image to a standard space (such as MNI standard space, Talairach standard space and the like), wherein the used software and methods comprise but are not limited to SPM, FS L, AFNI, ANTs, a deformation field-based registration method and the like, applying transformation parameters generated in the process to a focus image, obtaining a binary standard space focus image by selecting a threshold value (default 0.5 and optional 0-1), and registering the individual brain structure image with a standard symmetrical brain template (a template completely symmetrical between left and right brain spaces) by using the same software and method to obtain a binary symmetrical space focus image;
4) the method comprises the following steps of (1) individualizing a standard space template, wherein the step is suitable for the condition that partition characteristics need to be calculated in an individual space in the subsequent steps of (5) to (4) to (6), if the step is carried out in the standard space, the step can be omitted, extracting each partition in the template into an independent binary template aiming at partition templates (including but not limited to gray matter templates, white matter templates, cerebrospinal fluid templates, brain lobe partition templates, under-screen structure partition templates, AA L partition templates, Brodmann partition templates, Harvard Oxford partition templates, corpus callosum and other white matter partitions, ventricle templates, mesencephalon aqueduct and other self-made partitions and the like), respectively registering all the partition templates to the individual space by utilizing an inverse transformation process from the individual to the standard space in the step (3), and obtaining a binary individual space partition image corresponding to each partition by selecting a threshold (default 0.5, optionally 0 to 1);
5) And extracting the space distribution characteristics of the focus: wherein, the following lesion space distribution characteristic sets of several categories and the extraction method thereof are established:
5) -1, distribution size characteristics: extracting a single 2D or 3D focus by using an individual space focus image according to an inter-voxel communication principle, and calculating size characteristics such as the volume, the longest diameter and the like of the single focus; on the basis, the maximum value, the average value, the sum, the median, the number of the focuses and the large focuses (such as more than 200 mm) of the above characteristics of all the single focuses tested are obtained 3) The number and other statistical characteristics of the lesion to be tested are taken as a distribution size characteristic set of the tested lesion;
5) and (2) symmetry characteristics, namely using a 'symmetrical space focus image' and a corresponding partition template to respectively calculate symmetry characteristics of the focus in the left and right half brains in the whole brain and each group of symmetrical brain partitions, wherein the symmetry characteristics comprise unilateral/bilateral characteristics (the left/double/right sides are respectively marked as-1/0/1), a difference value or a ratio of the number of voxels of the focus in the left and right half brains, a Dice coefficient for measuring distribution similarity and the like, and the Dice coefficient of a certain focus i is defined as twice of the ratio of the number of the voxels in the intersection of the left focus (V L) and the right focus (VR) of the symmetrical focus to the number of the voxels in the union set after the left and right focus images are turned over, namely the Dice coefficient is double of the number of the voxels in the intersection
D(i)＝2×|VL∩VR|/|VL∪VR|
5) -3, probability profile characteristics: for each group of tested objects, calculating the number of nonzero values of each voxel in the group by using a binarized standard space lesion image of all individuals in the group, dividing the number by the total number of the individuals in the group to obtain a lesion space probability distribution map of the group (the voxel value range is 0-1, the closer to 1, the higher the existence probability of the lesion of the voxel is), and obtaining a high probability distribution map of the group by selecting a threshold (default 0.2, optional 0-1); defining the probability distribution diagram characteristic value of the g group of the focus of a certain tested i in M groups as the difference value of the sum of the number of voxels in the intersection of the focus area of i and the high probability distribution focus area of the g group and the intersection number of the focus and the high probability distribution focus area of other groups:
The following steps 5) -4-6, the focus image and the individual space partition image can be used for analyzing in the individual space, the standard space focus image and the partition template of the standard space can be used for analyzing in the standard space, or both;
5) 4, individual brain partition distribution characteristics: for each individual brain partition, the distribution characteristics of the tested brain lesion within that partition are calculated: the presence or absence of the focus in a partition (the presence is 1, the absence is 0), the volume of the focus in the partition, the proportion of the focus in the partition volume to the whole partition volume and the like;
5) -5, paracerebri distribution characteristics: aiming at interested marked brain subareas (such as ventricles, midbrain aqueducts, corpus callosum, white matter areas and the like), performing small area expansion operation (generally not more than 5mm), generating paracerebral areas, and calculating paracerebral distribution characteristics (such as paraventricular, around the midbrain aqueducts, near corpus callosum, near white matter areas and the like) of tested focuses aiming at corresponding brain areas, wherein the distribution characteristics comprise whether paracerebral focuses exist (the existence is 1, the nonexistence is 0), paracerebral focus volumes and the like;
5) -6, distribution of features across brain regions: according to the results of the steps 5) -4, calculating the cross-brain area distribution characteristics of the tested brain lesions, including the number and pairwise proportion of covered brain tissue types (gray matter/white matter/cerebrospinal fluid), the number of covered brain areas and the like;
6) Characteristic screening and modeling:
6) the method comprises the following steps of (1) feature screening, wherein on the basis of a lesion space distribution feature set extracted in the step 5), various feature screening methods in machine learning are adopted for feature screening, and the available feature screening methods comprise variance selection, chi-square test, U test, mutual information, recursive elimination, a feature selection method (such as an L ASSO method) based on a logical regression model with L1 punishment item/L2 punishment item/L1 combined with L2 punishment item, an exhaustion method and the like, wherein some methods (such as the L ASSO method) can simultaneously carry out feature screening and generate a classification identification model, so that the 6- (2) steps can be combined;
6) -2, establishing a classification identification model: taking the screened tested focus characteristics as input, taking the actual tested grouped categories as labels, training a classifier, and generating a linear or nonlinear classification identification model; classifiers that can be used include logistic regression, random forests, support vector machines, artificial neural networks, and the like; after the model is established, the classification and identification effects of the model are evaluated through the equivalent values of AUC value, accuracy, sensitivity and specificity of the ROC curve.
The invention provides a classification identification method of space distribution characteristics of brain focus images based on magnetic resonance imaging, which is used for obtaining a model and a discrimination method suitable for classification identification by calculating and analyzing the space distribution characteristics of multiple types of the brain focus images based on the magnetic resonance imaging, and can provide effective guidance for clinic and scientific research particularly in the aspect of classification of different diseases caused by different genes or antibodies and the like.
Drawings
FIG. 1 is a flow chart of a classification and identification method based on the spatial distribution characteristics of magnetic resonance imaging brain lesion images.
Figure 2 shows a graph of the MOG and AQP4 group lesion segmentation and extraction results in example 1.
Fig. 3 shows the feature selection result of example 1.
FIG. 4 shows the ROC curve for the identification effect of the model classification of example 1.
Detailed Description
Example 1 Classification of features of spatial distribution of brain lesion images in MOG antibody-positive and AQP4 antibody-positive NMOSD patients
Classification and identification:
1) clinical MRI images of two groups of patients, MOG antibody positive and AQP4 antibody positive nmods, were selected as 28 and 57 images, respectively, each containing fl AIR images as "lesion visualization images" and T1 weighted images as "brain structure images";
2) for the F L AIR image, using MRIacron to segment the whole brain focus, and storing the image as a binary image as a focus image;
3) rigidly registering the T1 weighted image of the individual with the F L AIR image using SPM;
4) applying transformation parameters to the F L AIR image, selecting a threshold value of 0.5 to obtain a binary 'standard space focus image';
5) Extracting a single 3D focus (as shown in figure 2) by using an individual space focus image, and calculating the volume of the single focus; obtaining the maximum value, average value, total sum, lesion number and large lesion (more than 200 mm) of the volume of all the single 3D lesions tested in each test 3) The number of (2);
6) Calculating the existence of the tested brain lesion in the partition (the existence is 1, the nonexistence is 0) and the volume of the lesion in the partition aiming at each independent brain partition;
7) Calculating the ratio of gray matter to white matter covered by the tested brain focus and the number of covered brain partitions;
8) and (3) feature screening and modeling, namely taking the features extracted in the step (5) as input, taking the actual tested grouping category as a label (the MOG group is 1, the AQP4 group is 0), using an L ASSO method, wherein 5-fold cross validation is used for feature screening and modeling, and finally, the model comprises 9 parameters, wherein 1 is a constant, and the rest 8 are space distribution features (shown in figure 3).
The results show that the established model has ROC curve AUC of 0.959, accuracy of 0.959, sensitivity of 1 and specificity of 0.86, and has good classification and identification effects (as shown in fig. 4).
Claims (7)
1. The classification and identification method of brain lesion image space distribution characteristics based on magnetic resonance imaging is characterized by comprising the following steps of:
1) And focal image segmentation:
2) carrying out rigid registration on the individual brain structure image and the lesion display image by adopting SPM and FS L software and methods;
3) the method comprises the following steps of performing spatial standardization, namely registering an individual brain structure image to a standard space by adopting software and methods SPM, FS L, AFNI, ANTs and a deformation field-based registration method, applying transformation parameters generated in the process to a focus image, selecting a threshold value to obtain a binary standard space focus image, and registering the individual brain structure image with a standard symmetrical brain template by using the same software and method to obtain a binary symmetrical space focus image;
4) Individuation of the standard space template: aiming at a partition template of a standard space, extracting each partition in the template as an independent binary template, respectively registering all partition templates to an individual space by utilizing an inverse transformation process from the individual to the standard space in the step 3, and obtaining a binary 'individual space partition image' corresponding to each partition by selecting a threshold (default 0.5, optional 0-1);
5) And extracting the space distribution characteristics of the focus: establishing a lesion space distribution characteristic set of several categories and an extraction method thereof:
6) Feature screening and modeling.
2. The method of claim 1, wherein in step 3),
Registering the individual 'brain structure image' to MNI standard space or Talairach standard space; the threshold is 0.5 as a default and is 0-1; the standard symmetrical brain template is a template with completely symmetrical left and right brain spaces.
3. The method of claim 1, wherein said step 4) is adapted to the case where the partition characteristics are calculated in the individual space in the subsequent steps 5) -4-6;
the partition template of the standard space includes but is not limited to a gray matter template, a white matter template, a cerebrospinal fluid template, a brain lobe partition template, a subterminal structure partition template, an AA L partition template, a Brodmann partition template, a Harvard Oxford partition template, a corpus callosum white matter partition, a ventricle template, and a midbrain aqueduct self-made partition.
4. The method of claim 1, wherein said step 1) comprises the substeps of:
1) preparing and selecting brain image data, namely at least two groups of tested images G1 and G2, preparing two images for each tested individual, wherein one image is a 'lesion display image', one modality is selected from obtained clinical brain magnetic resonance images of a patient including but not limited to a T1 weighted image, a T2 weighted image, F L AIR, DWI, ADC and SWI to be used as an image for subsequent lesion segmentation, and the other image is a 'brain structural image' and is selected from a T1 weighted image;
1) -2, lesion segmentation: for "lesion display images," image segmentation software, kits and segmentation algorithms, including but not limited to, workstation and segmentation software provided by MRI vendors, MRIcro, MRIcron, MIPAV, ITK-SNAP, MITK or region growing algorithms, are used to segment whole brain lesion images and saved as binary images or data to generate "lesion images.
5. The method of claim 1, wherein said step 5) comprises the following feature set extraction method:
5) -1, distribution size characteristics: extracting a single 2D or 3D focus by using an individual space focus image according to an inter-voxel communication principle, and calculating the size and longest diameter size characteristics of the single focus image; obtaining the maximum value, the average value, the sum, the median, the number of the focuses and the over 200mm of the above characteristics of all the single focus images tested 3The statistical characteristic of the number of the large focuses is used as a distribution size characteristic set of the tested focus images;
5) and (2) symmetry characteristics, namely using a 'symmetric space focus image' and a corresponding partition template to respectively calculate symmetry characteristics of the focus image in the left and right half brains in the whole brain and each group of symmetric brain partitions, wherein the symmetry characteristics comprise unilateral/bilateral characteristics, the left/double/right sides are respectively marked as-1/0/1, the difference value or ratio of the number of voxels of the focus image in the left and right half brains and a Dice coefficient for measuring distribution similarity, and the Dice coefficient of a certain focus i is defined as twice of the ratio of the number of pixels in the intersection of the left focus (V L) or the right focus (VR) image of the symmetric focus to the number of the pixels in the union set after the left and right sides of the focus image are turned over, namely the two times of the number of the pixels in the intersection of the two
D(i)＝2×|VL∩VR|/|VL∪VR|
5) -3, probability profile characteristics: for each group of tested objects, calculating the number of nonzero values of each voxel in the group by using a binarized standard space lesion image of all individuals in the group, and dividing the number by the total number of the individuals in the group to obtain a lesion space probability distribution map of the group, wherein the closer the voxel value ranges from 0 to 1, the higher the existence probability of the lesion of the voxel is, and the higher the probability of the lesion of the voxel is, by selecting a threshold value, defaulting to 0.2 and selecting 0 to 1, the high probability distribution map of the group is obtained; defining the probability distribution diagram characteristic value of the g group of the focus of a certain tested i in M groups as the difference value of the sum of the number of voxels in the intersection of the focus area of i and the high probability distribution focus area of the g group and the intersection number of the focus and the high probability distribution focus area of other groups:
The following steps 5) -4-6, the focus image and the individual space partition image can be used for analyzing in an individual space, or the standard space focus image and the partition template of the standard space are used for analyzing in the standard space, or both;
5) 4, individual brain partition distribution characteristics: for each individual brain partition, calculating the distribution characteristics of the brain lesion image to be tested in the partition: the lesion exists or not in the partition, the existence is 1, the nonexistence is 0, the volume of the lesion in the partition, and the proportion of the volume of the lesion in the partition to the volume of the whole partition;
5) -5, paracerebri distribution characteristics: aiming at interested marked brain partitions such as ventricles of brains, midbrain aqueducts, corpus callosum and white matter areas, performing small area expansion operation, not more than 5mm, generating paracerebral areas, and calculating paracerebral distribution characteristics of tested focuses aiming at the corresponding brain areas, including paraventricular, around midbrain aqueducts, near corpus callosum and near white matter areas, including whether paracerebral focuses exist, the existence is 1, the nonexistence is 0 and paracerebral focus volume;
5) -6, distribution of features across brain regions: according to the results of the steps 5) -4, calculating the cross-brain area distribution characteristics of the tested brain focus image, including covered brain tissue types: the number and pairwise proportion of gray matter/white matter/cerebrospinal fluid and the number of covered brain partitions.
6. The method of claim 1, wherein said step 6) comprises the substeps of:
6) the method comprises the following steps of-1, feature screening, namely screening features by adopting various feature screening methods in machine learning on the basis of a focus image space distribution feature set extracted in the step 5), wherein the used feature screening methods comprise variance selection, chi-square test, U test, mutual information, recursive elimination method, a feature selection method or an exhaustion method based on a logistic regression model combining L1 punishment item/L2 punishment item/L1 with L2 punishment item, wherein the L ASSO method can simultaneously screen the features and generate a classification identification model, so that the step 6- (2) can be combined;
6) -2, establishing a classification identification model: taking the screened image characteristics of the tested focus as input, taking the actual tested grouped categories as labels, training a classifier, and generating a linear or nonlinear classification identification model; the used classifier comprises a logistic regression, a random forest, a support vector machine or an artificial neural network;
After the model is established, the classification identification effect of the model is evaluated through the AUC value, accuracy, sensitivity and specificity value of the ROC curve.
7. the method of claim 6, wherein the characteristic selection method of the logistic regression model in the step 6) -1 is an L ASSO method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910737818.4A CN111415324B (en) | 2019-08-09 | 2019-08-09 | Classification and identification method for brain disease focus image space distribution characteristics based on magnetic resonance imaging |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910737818.4A CN111415324B (en) | 2019-08-09 | 2019-08-09 | Classification and identification method for brain disease focus image space distribution characteristics based on magnetic resonance imaging |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111415324A true CN111415324A (en) | 2020-07-14 |
| CN111415324B CN111415324B (en) | 2024-03-08 |
Family
ID=71494186
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910737818.4A Active CN111415324B (en) | 2019-08-09 | 2019-08-09 | Classification and identification method for brain disease focus image space distribution characteristics based on magnetic resonance imaging |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111415324B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112075927A (en) * | 2020-10-15 | 2020-12-15 | 首都医科大学附属北京天坛医院 | Method and device for classifying causes of cerebral apoplexy |
| CN112651976A (en) * | 2020-10-13 | 2021-04-13 | 首都医科大学附属北京天坛医院 | Focal region brain network determination method and system based on low-resolution nuclear magnetic data |
| CN113066113A (en) * | 2021-04-27 | 2021-07-02 | 河北医科大学第二医院 | Method and equipment for constructing spontaneous hypertensive rat brain template and map set |
| CN113192014A (en) * | 2021-04-16 | 2021-07-30 | 深圳市第二人民医院(深圳市转化医学研究院) | Training method, device, electronic equipment and medium for improving ventricle segmentation model |
| CN113283465A (en) * | 2021-04-02 | 2021-08-20 | 电子科技大学 | Diffusion tensor imaging data analysis method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106156484A (en) * | 2016-06-08 | 2016-11-23 | 中国科学院自动化研究所 | Disease of brain individuation Forecasting Methodology based on nuclear magnetic resonance image and system |
| WO2017210873A1 (en) * | 2016-06-08 | 2017-12-14 | 中国科学院自动化研究所 | Individualized brain disease prediction method and system based on magnetic resonance imaging |
-
2019
- 2019-08-09 CN CN201910737818.4A patent/CN111415324B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106156484A (en) * | 2016-06-08 | 2016-11-23 | 中国科学院自动化研究所 | Disease of brain individuation Forecasting Methodology based on nuclear magnetic resonance image and system |
| WO2017210873A1 (en) * | 2016-06-08 | 2017-12-14 | 中国科学院自动化研究所 | Individualized brain disease prediction method and system based on magnetic resonance imaging |
Non-Patent Citations (2)
| Title |
|---|
| 庞清华等: "基于VBM-DARTEL的局灶性皮质发育不良脑灰质体积异常检测方法", 《中国医学物理学杂志》 * |
| 张衡等: "纹理分析在甲状腺结节影像学中的研究进展", 《中国医学计算机成像杂志》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112651976A (en) * | 2020-10-13 | 2021-04-13 | 首都医科大学附属北京天坛医院 | Focal region brain network determination method and system based on low-resolution nuclear magnetic data |
| CN112651976B (en) * | 2020-10-13 | 2022-07-19 | 首都医科大学附属北京天坛医院 | Focal region brain network determination method and system based on low-resolution nuclear magnetic data |
| CN112075927A (en) * | 2020-10-15 | 2020-12-15 | 首都医科大学附属北京天坛医院 | Method and device for classifying causes of cerebral apoplexy |
| CN112075927B (en) * | 2020-10-15 | 2024-05-14 | 首都医科大学附属北京天坛医院 | Etiology classification method and device for cerebral apoplexy |
| CN113283465A (en) * | 2021-04-02 | 2021-08-20 | 电子科技大学 | Diffusion tensor imaging data analysis method and device |
| CN113283465B (en) * | 2021-04-02 | 2022-04-29 | 电子科技大学 | Diffusion tensor imaging data analysis method and device |
| CN113192014A (en) * | 2021-04-16 | 2021-07-30 | 深圳市第二人民医院(深圳市转化医学研究院) | Training method, device, electronic equipment and medium for improving ventricle segmentation model |
| CN113192014B (en) * | 2021-04-16 | 2024-01-30 | 深圳市第二人民医院(深圳市转化医学研究院) | Training method and device for improving ventricle segmentation model, electronic equipment and medium |
| CN113066113A (en) * | 2021-04-27 | 2021-07-02 | 河北医科大学第二医院 | Method and equipment for constructing spontaneous hypertensive rat brain template and map set |
| CN113066113B (en) * | 2021-04-27 | 2022-01-11 | 河北医科大学第二医院 | Method and equipment for constructing spontaneous hypertensive rat brain template and map set |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111415324B (en) | 2024-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111415324B (en) | Classification and identification method for brain disease focus image space distribution characteristics based on magnetic resonance imaging | |
| CN107909117B (en) | Classification device for early and late mild cognitive impairment based on brain function network characteristics | |
| US8831327B2 (en) | Systems and methods for tissue classification using attributes of a biomarker enhanced tissue network (BETN) | |
| CN112101451B (en) | Breast cancer tissue pathological type classification method based on generation of antagonism network screening image block | |
| Bjornsson et al. | Associative image analysis: a method for automated quantification of 3D multi-parameter images of brain tissue | |
| CN113454733A (en) | Multi-instance learner for prognostic tissue pattern recognition | |
| CN109840554B (en) | Alzheimer's disease MRI image classification method based on SVM-RFE-MRMR algorithm | |
| CN114730463A (en) | Multi-instance learner for tissue image classification | |
| CN110827242A (en) | Intracranial aneurysm detection method and system based on convolutional neural network | |
| Song et al. | Using HOG-LBP features and MMP learning to recognize imaging signs of lung lesions | |
| US10423820B2 (en) | Systems and methods for automatic generation of training sets for machine interpretation of images | |
| CN110796661B (en) | Fungal microscopic image segmentation detection method and system based on convolutional neural network | |
| CN112215799A (en) | Automatic classification method and system for grinded glass lung nodules | |
| US20140235487A1 (en) | Oral cancer risk scoring | |
| CN114600155A (en) | Weakly supervised multitask learning for cell detection and segmentation | |
| CN114782372B (en) | DNA fluorescence in situ hybridization BCR/ABL fusion state detection method and detection system | |
| CN110288577A (en) | A kind of 3D realizes the control method and control device of multiple organ lesion integration | |
| CN108805181B (en) | Image classification device and method based on multi-classification model | |
| Vijendran et al. | Optimal segmentation and fusion of multi-modal brain images using clustering based deep learning algorithm | |
| US20210313064A1 (en) | Tau protein accumulation prediction apparatus using machine learning and tau protein accumulation prediction method using the same | |
| Ravi et al. | Machine Learning-based Classification and Analysis of Breast Cancer Pathological Images | |
| Bhuvaneswary et al. | Multi modal image fusion technique for detecting brain tumor | |
| Majumder et al. | Digital image analysis for early diagnosis of cancer: Identification of pre-cancerous state | |
| Tasoulis et al. | Fractal analysis and fuzzy c-means clustering for quantification of fibrotic microscopy images | |
| Nandy et al. | Supervised learning framework for screening nuclei in tissue sections |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |