CN113256654A - White matter high signal extraction method and system based on lesion diffusion algorithm - Google Patents

White matter high signal extraction method and system based on lesion diffusion algorithm Download PDF

Info

Publication number
CN113256654A
CN113256654A CN202110579377.7A CN202110579377A CN113256654A CN 113256654 A CN113256654 A CN 113256654A CN 202110579377 A CN202110579377 A CN 202110579377A CN 113256654 A CN113256654 A CN 113256654A
Authority
CN
China
Prior art keywords
lesion
white matter
map
voxel
high signal
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
Application number
CN202110579377.7A
Other languages
Chinese (zh)
Other versions
CN113256654B (en
Inventor
白丽君
潘怡臻
熊峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN202110579377.7A priority Critical patent/CN113256654B/en
Publication of CN113256654A publication Critical patent/CN113256654A/en
Application granted granted Critical
Publication of CN113256654B publication Critical patent/CN113256654B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • 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
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • 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
    • G06T2207/20076—Probabilistic image processing
    • 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
    • G06T2207/30096—Tumor; Lesion

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Processing (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

The invention discloses a white matter high signal extraction method and a white matter high signal extraction system based on a lesion diffusion algorithm, wherein an original image comprising a FLAIR image, a T1 weighted image and a white matter tissue prior probability map is input; preprocessing an input original image; performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map, and performing initialization and white matter high signal segmentation on the obtained tissue signal intensity distribution confidence map to obtain a white matter high signal probability map; and performing image binarization on the white matter high signal probability map generated after the segmentation processing through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map. The method is easy to operate, has strong generalization, and is a white matter high signal extraction method which has stable effect and strong practicability and is very effective.

Description

White matter high signal extraction method and system based on lesion diffusion algorithm
Technical Field
The invention belongs to the technical field of medical imaging, and particularly relates to a white matter high signal extraction method and system based on a lesion diffusion algorithm.
Background
The processing of white matter high signals is typically visualized by imaging, most commonly using a magnetic resonance imaging fluid attenuated inversion recovery sequence (FLAIR) for visualization. In the FLAIR sequence, cerebrospinal fluid is inhibited to be low signal, contrast of high-signal lesion is enhanced, and the imaging effect on white matter high signal of paraventricular, ventricular and deep parts is better than that of other imaging modes.
Currently, the evaluation of white high signals is divided into manual evaluation and automatic evaluation. The manual evaluation means that special personnel observe the magnetic resonance image with naked eyes, and then evaluate the high signal grade of the white matter of the patient by using a cerebral white matter disease scale such as Fazekas and the like. The automatic evaluation refers to a method for quantitatively extracting white matter high signals by utilizing an automatic white matter high signal detection algorithm. In recent years, due to the rise of artificial intelligence, many machine learning and deep learning-based white matter high signal segmentation methods have been generated, such as machine-based methods based on a support vector machine, K-neighborhood, random forest, and the like, and deep learning-based methods based on a VGG network (super resolution test sequence), a 2D full convolution network, a 3D convolution network, and the like. These methods are typically long in implementation cycle, have high requirements on the modality of the original image, and are generally optimal for a particular sample and have insufficient generalization.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a white matter high signal extraction method and system based on a lesion diffusion algorithm, which is reliable in extraction result, strong in generalization and easy to use, in view of the above-mentioned deficiencies in the prior art.
The invention adopts the following technical scheme:
a white matter high signal extraction method based on a lesion diffusion algorithm comprises the following steps:
s1, inputting an original image comprising a FLAIR image, a T1 weighted image and a tissue prior probability map of white matter;
s2, carrying out preprocessing operation on the original image input in the step S1;
s3, performing abnormal signal product operation on the image preprocessed in the step S2 to obtain a tissue signal intensity distribution confidence map;
s4, initializing the tissue signal intensity distribution confidence map obtained in the step S3 and carrying out white matter high signal segmentation to obtain a white matter high signal probability map;
and S5, performing image binarization on the white matter high signal probability map generated after the segmentation processing in the step S4 through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
Specifically, in step S2, the preprocessing specifically includes magnetic field offset correction, image registration, generation of partial volume images, tissue segmentation of gray matter white matter cerebrospinal fluid, and transformation of the images from standard space to original space.
Further, the generating of the partial volume image comprises generating partial volume image labels using the T1 weighted image, wherein the label values are between 1-3, 1 represents cerebrospinal fluid, 2 represents gray matter, 3 represents white matter, assigning a discrete label to each voxel, assigning a discrete tissue label z to each voxeliComprises the following steps:
Figure BDA0003085484470000021
where i represents each voxel, xiRepresenting voxel intensity values labeled by the partial volume image.
Specifically, in step S3, tissue signal intensity distribution confidence maps of cerebrospinal fluid, gray matter, and white matter are obtained together according to the preprocessed FLAIR image, partial volume image of T1, and prior probability result of white matter, a complete signal intensity distribution confidence map B is obtained by superimposing the tissue signal intensity distribution confidence maps, and a complete signal intensity distribution confidence map B is obtained according to BGMAnd taking a threshold value kappa to initialize a lesion point.
Further, the signal intensity distribution confidence map B is:
B＝BCSF+BGM+BWM
wherein, BCSF、BGM、BWMGroups of cerebrospinal fluid, gray matter and white matter, respectivelyAnd (4) weaving signal intensity distribution confidence maps.
Further, the threshold value kappa is 0.15 to 0.5.
Specifically, step S4 specifically includes:
starting point LinitDividing the nearby voxels into lesion and other voxels, judging the initial point L of the lesion by using the other voxels to be cerebrospinal fluid, gray matter or white matterinitWhether the nearby point of (a) is a lesion; if the nearby point is a lesion, judging whether the nearby point of the corresponding point is the lesion, setting the voxel sharing a common boundary with the lesion voxel as the lesion voxel by an iterative growth algorithm until the lesion stops diffusing, and obtaining a voxel lesion estimation value
Figure BDA0003085484470000031
0 to 1, from the initial point
Figure BDA0003085484470000032
The value is set to 1, and when updating the lesion distribution parameters alpha, beta, consideration is given to
Figure BDA0003085484470000033
While only the voxel distribution parameter theta is taken into account when updating the other voxel distribution parameters theta
Figure BDA0003085484470000034
The lesion low probability voxels of (a); when there is no new one
Figure BDA0003085484470000035
The iteration of the voxels is stopped to obtain a lesion probability map, and the probability of the voxel lesion increase around the lesion voxels represented by the Markov random field theory is substituted into the lesion probability map to obtain an optimized high-signal probability map of the white matter of the brain.
Further, the iteration through the iterative growth algorithm specifically includes:
Figure BDA0003085484470000036
wherein i represents at leastThere is one neighboring voxel j satisfying
Figure BDA0003085484470000041
biFor the signal intensity distribution confidence map of voxel i,
Figure BDA0003085484470000042
respectively, an estimated value of the distribution parameter, P, at the previous momentOtherProbability that a voxel does not belong to a lesion, PLesIs the probability that a voxel belongs to a lesion.
Further, the optimized high signal probability map of white matter of brain is as follows:
Figure BDA0003085484470000043
wherein, PLesIs the probability of a voxel belonging to a lesion, yiFor the FLAIR signal intensity of voxel i,
Figure BDA0003085484470000044
Figure BDA0003085484470000045
respectively, an estimate of the distribution parameter at the previous moment, biFor the signal intensity distribution confidence map of voxel i,
Figure BDA0003085484470000046
is a lesion probability map, P, of voxels j adjacent to voxel iOtherIs the probability that a voxel does not belong to a lesion.
Another technical solution of the present invention is a white matter high signal extraction system based on a lesion diffusion algorithm, comprising:
an input module which inputs an original image;
the preprocessing module is used for preprocessing the original image input by the input module;
the initialization module is used for performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map;
the lesion extraction module is used for initializing the obtained tissue signal intensity distribution confidence map and segmenting white matter high signals;
and the post-processing module is used for carrying out image binarization on the white matter high signal probability map generated after the segmentation processing of the lesion extraction module through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
Compared with the prior art, the invention has at least the following beneficial effects:
the invention relates to a white matter high signal extraction method based on a lesion diffusion algorithm, which comprises the steps of preprocessing FLAIR data, generating partial volume map image labels for T1 data, converting a white matter tissue prior probability map into a standard space, then generating a signal intensity distribution confidence map, then performing white matter high signal segmentation with the lesion diffusion algorithm as a core, and finally generating a white matter high signal binary image. The method has complete flow, can remove the instability of the original image, and clearly and completely calculates the white matter high signal. Moreover, for a user, the method is convenient to use, has few parameters needing to be set, has strong generalization and can be suitable for data in different types of testees.
Furthermore, due to the fact that the difference between different data individuals, the difference of factors such as the environment of MRI image data of the same individual in different modalities, the influence caused by head movement of the individual and an offset field and the like cannot be unified, the original data needs to be calibrated and unified to a certain extent through preprocessing, the method can enable MRI images of two modalities, namely an FLAIR image and a T1 image, to correspond to each other in the space position one by one, the purpose of information fusion is achieved, the offset field estimation image can be superposed on the original image through an N4 offset field correction method, the gray value homogenization of the same tissue voxel value is achieved, and voxels belonging to the same tissue are prevented from being divided into different tissues due to uneven gray value.
Further, each voxel is classified by T1 as the original image, a partial volume image with a label value between 1 and 3 is generated, and a discrete label is assigned to each voxel. (wherein 1 represents cerebrospinal fluid, 2 represents gray matter, and 3 represents white matter). The partial volume image is a prerequisite for subsequent calculation of the FLAIR signal intensity distribution confidence map for each tissue by probabilistic methods.
Further, a product calculation is carried out on a FLAIR image obtained after the original FLAIR image is preprocessed, a partial volume image obtained after the original T1 is processed and a tissue prior probability map transformed to an original space, the difference value of the FLAIR image value and the intensity mean value of each voxel is multiplied by the label value of the partial volume image and the probability that the voxel position in the tissue prior probability map belongs to white matter, signal intensity distribution confidence maps of three tissues are obtained, and the range of lesion spread can be limited through the signal intensity distribution confidence maps, and the position of a seed point for lesion initialization can be provided.
Further, the signal intensity distribution confidence maps of the three tissues are superposed to obtain a signal intensity distribution confidence map, and the superposed confidence maps can limit the spread range of the lesion and prevent the lesion from spreading to an unreasonable position. Wherein the gray matter signal intensity distribution confidence map can initialize the starting seed point of the lesion.
Further, a threshold needs to be set for the gray matter signal intensity distribution confidence map when the lesion is initialized, and if the threshold is too high, the position which should belong to the lesion may be classified as other tissues. If the threshold is too low, other tissues not belonging to the lesion can be classified as the lesion, so the value of the threshold should not be too large or too small, and is guaranteed to be between 0.15 and 0.5.
Furthermore, the diffusion of the lesion is carried out by judging whether the neighborhood voxel at the initial lesion position belongs to the lesion, then whether the neighborhood of the newly added lesion is the lesion is continuously judged, and the diffusion position of the lesion, namely the white matter high signal probability map, is finally determined by repeated iteration. The influence of the neighborhood on the voxel can be strengthened by improvement based on the Markov random field theory, so that the accuracy of a calculation result is improved.
Further, iterative setting is carried out on the iterative growth algorithm of lesion diffusion. And dividing the probability density distribution of the lesion with the probability density distribution of other tissues, multiplying the result by the signal intensity distribution confidence map of the voxel to ensure that the lesion grows within an assumed range, and then taking the value of 1 to obtain a basic lesion probability map under the algorithm idea.
Furthermore, the probability graph of the pathological changes in the basic algorithm is multiplied by the probability result of the conditional distribution of the voxel of the pathological changes based on the energy function obtained by the Markov random field theory, and the result is reduced by 1, so that a more accurate probability graph of the pathological changes after optimization, namely a high signal probability graph of the white matter of the brain, is obtained.
In conclusion, the method can clearly and effectively extract the white matter high signal through an algorithm based on lesion diffusion, and is easy to use and high in generalization.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic block diagram of the present invention;
fig. 3 is an exemplary schematic diagram of the process of the method of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
Various structural schematics according to the disclosed embodiments of the invention are shown in the drawings. The figures are not drawn to scale, wherein certain details are exaggerated and possibly omitted for clarity of presentation. The shapes of various regions, layers and their relative sizes and positional relationships shown in the drawings are merely exemplary, and deviations may occur in practice due to manufacturing tolerances or technical limitations, and a person skilled in the art may additionally design regions/layers having different shapes, sizes, relative positions, according to actual needs.
The invention provides a white matter high signal extraction method and a white matter high signal extraction system based on a lesion diffusion algorithm, which are suitable for the segmentation of a brain white matter high signal, and are used for preprocessing an original input image so as to meet the input requirement of the lesion diffusion algorithm; and initializing the lesion position of the image through a lesion diffusion algorithm, diffusing and iterating the lesion position until the lesion position is not changed any more to obtain a white matter high signal probability map, and superposing the white matter high signal extraction maps to obtain a white matter high signal location map. The method is simple to operate, accurate in segmentation result, strong in generalization performance to data and strong in practicability.
Referring to fig. 1 and fig. 3, the white matter high signal extraction method based on the lesion diffusion algorithm of the present invention includes the following steps:
s1, inputting an original image;
the original image includes: FLAIR images, T1 weighted images and tissue prior probability maps of white matter.
S2, carrying out preprocessing operation on the original image input in the step S1;
preprocessing operations including magnetic field offset correction, image registration, generation of partial volume images, tissue segmentation of gray matter white matter cerebrospinal fluid and transformation of the images from standard space to original space;
the image registration specifically comprises: registration is performed with the FLAIR image as the reference image and the T1 weighted image as the mating source image.
The generation of the partial volume image includes generating partial volume image labels using T1 weighted images, where the label values are between 1-3, where 1 represents cerebrospinal fluid, 2 represents gray matter, and 3 represents white matter, and assigning a discrete label to each voxel as shown in equation 1.
Figure BDA0003085484470000081
Where i represents each voxel, xiRepresenting voxel intensity values, z, labeled by a partial volume imageiRepresenting a discrete tissue label assigned to each voxel.
The transformation of the image from the canonical space to the original space comprises the transformation of the tissue prior probability map from the canonical space to the original space. Since the FLAIR preprocessed image and the partial volume label image obtained from the T1 original image both exist in the individual space, and the prior probability map of white matter tissue exists in the standard space, the prior probability map of white matter tissue needs to be transformed from the standard space to the original space.
S3, performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map;
calculating and obtaining signal intensity distribution confidence maps of the three tissues and a complete signal intensity distribution confidence map, setting a threshold value for the image preprocessed in the step S2, and initializing a lesion point;
the signal intensity distribution confidence maps of the three tissues are obtained by the preprocessed FLAIR image, the partial volume image of T1 and the prior probability result of white matter, and the calculation method is shown as formula 2.
Figure BDA0003085484470000091
Where i is 1,2, … …, n, n denotes the number of voxels, i denotes each voxel, k denotes the type of tissue, and there are three cases including cerebrospinal fluid, gray matter and white matter. y ═ y1,y2,……,yn) Signal intensities of n voxels in the FLAIR pre-processed image are indicated.
Figure BDA0003085484470000092
Represents the average of y along the three tissue directions.
When in use
Figure BDA0003085484470000093
At that time, it is recorded that there is an anomaly in the FLAIR signal intensity at that voxel, at which time,
Figure BDA0003085484470000094
otherwise, the FLAIR signal intensity at that voxel is recorded as normal, at which time,
Figure BDA0003085484470000095
selecting three different k, respectively obtaining signal intensity distribution confidence maps b of voxels i of three tissues of cerebrospinal fluid, gray matter and white matterCSF,i、bGM,i、bWM,i。
Next, signal intensity distribution confidence maps of the three tissues are obtained, as shown in equation 3.
Figure BDA0003085484470000096
Wherein, BCSF、BGM、BWMConfidence maps of tissue signal distribution intensity for cerebrospinal fluid, gray matter and white matter, respectively.
When the white matter prior probability is high, FLAIR signal is high, T1 signal xi<1.5 times, BCSFThe location is in cerebrospinal fluid;
when the prior probability of white matter is high, the FLAIR signal is high, and the T1 signal is 1.5 ≦ xiWhen < 2.5, BGMBelonging to gray matter; when the white matter prior probability is high, FLAIR signal is high, T1 signal xiWhen not less than 2.5, BWMThe location of the white matter where the tissue signal distribution intensity is high indicates the location of the lesion.
And (3) obtaining a complete signal intensity distribution confidence map B by superposing the tissue signal intensity distribution confidence maps, as shown in formula 4:
B＝BCSF+BGM+BWM (4)
since lesions are associated with gray matter labels of all partial volume images, it is necessary to be in accordance with BGMThe threshold k is taken to initialize the lesion point, as shown in equation 5.
Linit＝{linit,1,linit,2,……,linit,n} (5)
When b isGMWhen κ is greater, linit,i＝1，LinitThen the seed point for the initialized lesion.
The threshold k cannot be too large or too small, lesions are not divided due to too large threshold, more common tissues are divided into lesions due to too small threshold, and the application threshold k is 0.15-0.5 and the optimal threshold k is 0.3 through tests.
S4, initializing the tissue signal intensity distribution confidence map obtained in the step S3 and carrying out white matter high signal segmentation;
segmenting white matter high signal by lesion growth algorithm to enable lesions to be from initial point LinitExpands to a position within the range of the signal intensity distribution confidence map B.
Specifically, the lesion initiation point L needs to be determinedinitWhether the nearby point of (a) is a lesion; if the nearby point is a lesion, judging whether the nearby point of the point is the lesion, and diffusing the lesion point through iteration until the lesion stops diffusing, so that a high-signal probability map of the white matter can be obtained.
More specifically, the initial point LinitThe nearby voxels are classified into lesions and other two kinds, wherein the other voxels are cerebrospinal fluid, gray matter or white matter, and the three follow Gaussian distribution, so that the other voxels are classified into lesions and other voxelsThe voxel as a whole conforms to a gaussian mixture distribution as shown in equation 6.
Figure BDA0003085484470000111
Where φ is the probability density function of the normal distribution, μk，
Figure BDA0003085484470000112
Respectively mean and variance, pikThe proportion of the k-th class is assigned to the voxel, theta is the parameter mu of all tissue classesk，
Figure BDA0003085484470000113
πk. Since the tissue class of voxel i is known in the preprocessing, the gaussian mixture distribution is simplified into separate individual gaussian distributions, and the parameter estimation results of other voxels are shown in equation 7 according to the maximum likelihood estimation of the gaussian distributions.
Figure BDA0003085484470000114
Wherein n iskIndicates the number of voxels belonging to the kth class,
Figure BDA0003085484470000115
respectively, are the result of the estimated parameters.
The process of the initial lesion progressing into the lesion map is described by iteratively growing the algorithm such that voxels sharing a common boundary with the lesion voxels are also set as lesion voxels, the iterative process being shown in equation 8.
Figure BDA0003085484470000116
Wherein i indicates that at least one neighboring voxel j satisfies
Figure BDA0003085484470000117
biEnsuring that the lesion grows within the hypothesis.
Because the distribution of the pathological changes meets the gamma distribution, alpha and beta are parameters of the gamma distribution,
Figure BDA0003085484470000118
representing the ratio of the lesion voxel distribution to other voxel distributions.
Figure BDA0003085484470000121
Representing the distribution parameter estimate at the previous time. Voxel lesion estimation
Figure BDA0003085484470000122
Is a number between 0 and 1 and,
Figure BDA0003085484470000123
larger indicates that the voxel is more likely to be a lesion voxel. Starting point of
Figure BDA0003085484470000124
The value is set to 1. When updating the lesion distribution parameters α, β, only consideration is given to
Figure BDA0003085484470000125
While only the voxel distribution parameter theta is taken into account when updating the other voxel distribution parameters theta
Figure BDA0003085484470000126
The lesion low probability voxel of (2). When there is no new one
Figure BDA0003085484470000127
The iteration is stopped to obtain a lesion probability map.
Further, by combining the information of the adjacent voxels with an expansion algorithm, the assumption that the voxels surrounded by the lesion voxels are more likely to be lesions is consistent with the markov random field theory. Let random variable Z ═ Z1,Z2,……,Zn) Each Z field L ═ { Les, Oth, ZiVoxel z of nearby NiNi＝{z'iI' is belonged to Ni }. Since Hammersley Clifford Theorem theory holds that the Markov random field and Gibbs distribution are consistent, ziThe conditional distribution of (a) can be regarded as a gibbs distribution as shown in equation 9.
Figure BDA0003085484470000128
Wherein, U is an energy function, and satisfies the Ising model, as shown in equation 10.
Figure BDA0003085484470000129
Where I is an indicator function, i.e. when a sentence is true, I is 1, when a sentence is false, I is 0,
Figure BDA00030854844700001210
take 1 by default. If more voxels in Ni are labeled as lesions, the energy function U (k | z)Ni) It is advantageous to segment the voxels into lesions.
Since the algorithm is continuous rather than discrete, the energy of the lesion, U (z), is a functioni|zNi) The change is to a probabilistic form as shown in equation 11.
Figure BDA00030854844700001211
And substituting the probability of the voxel lesion increase around the lesion voxel represented by the Markov random field theory into the lesion probability map to obtain an optimized lesion probability map, as shown in formula 12.
Figure BDA0003085484470000131
Wherein the content of the first and second substances,
Figure BDA0003085484470000132
is a probability map of lesions, P, of voxel iLesIs the probability of a voxel being a lesion, yiIs the signal intensity of the voxel i and,
Figure BDA0003085484470000133
respectively, an estimate of the distribution parameter at the previous moment, biFor the signal intensity distribution confidence map of voxel i,
Figure BDA0003085484470000134
is a lesion probability map, P, of voxels j adjacent to voxel iOtherIs the probability that a voxel is not a lesion.
S5, performing image binarization on the white matter high signal probability map generated after the segmentation processing through a threshold value to obtain a white matter high signal extraction map; and superposing the white matter high signal extraction image on the normalized original image to obtain a white matter high signal localization image, and determining the size and the position of white matter high signal lesion.
And (3) counting the distribution of all voxels with the lesion probability greater than 0 in the subject, and selecting 1 as a threshold value to obtain a white matter signal extraction map.
Referring to fig. 2, in another embodiment of the present invention, a white matter high signal extraction system based on a lesion diffusion algorithm is provided, which can be used for implementing the white matter high signal extraction based on the lesion diffusion algorithm, and specifically, the white matter high signal extraction system based on the lesion diffusion algorithm includes an input module, a preprocessing module, an initialization module, a lesion extraction module, and a post-processing module.
The input module inputs an original image;
the preprocessing module is used for preprocessing the original image input by the input module;
the initialization module is used for performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map;
the lesion extraction module is used for initializing the obtained tissue signal intensity distribution confidence map and segmenting white matter high signals;
and the post-processing module is used for carrying out image binarization on the white matter high signal probability map generated after the segmentation processing of the lesion extraction module through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
In yet another embodiment of the present invention, a terminal device is provided that includes a processor and a memory for storing a computer program comprising program instructions, the processor being configured to execute the program instructions stored by the computer storage medium. The Processor may be a Central Processing Unit (CPU), or may be other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable gate array (FPGA) or other Programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc., which is a computing core and a control core of the terminal, and is adapted to implement one or more instructions, and is specifically adapted to load and execute one or more instructions to implement a corresponding method flow or a corresponding function; the processor provided by the embodiment of the invention can be used for the operation of the white matter high signal extraction method and system based on the lesion diffusion algorithm, and comprises the following steps:
inputting an original image comprising a FLAIR image, a T1 weighted image and a tissue prior probability map of white matter; preprocessing an input original image; performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map, and performing initialization and white matter high signal segmentation on the obtained tissue signal intensity distribution confidence map to obtain a white matter high signal probability map; and performing image binarization on the white matter high signal probability map generated after the segmentation processing through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
In still another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a Memory device in a terminal device and is used for storing programs and data. It is understood that the computer readable storage medium herein may include a built-in storage medium in the terminal device, and may also include an extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also, one or more instructions, which may be one or more computer programs (including program code), are stored in the memory space and are adapted to be loaded and executed by the processor. It should be noted that the computer-readable storage medium may be a high-speed RAM memory, or may be a non-volatile memory (non-volatile memory), such as at least one disk memory.
The processor can load and execute one or more instructions stored in the computer-readable storage medium to realize the corresponding steps of the white matter high signal extraction method and system based on the lesion diffusion algorithm in the above embodiments; one or more instructions in the computer-readable storage medium are loaded by the processor and perform the steps of:
inputting an original image comprising a FLAIR image, a T1 weighted image and a tissue prior probability map of white matter; preprocessing an input original image; performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map, and performing initialization and white matter high signal segmentation on the obtained tissue signal intensity distribution confidence map to obtain a white matter high signal probability map; and performing image binarization on the white matter high signal probability map generated after the segmentation processing through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
In summary, according to the white matter high signal extraction method and system based on the lesion diffusion algorithm, the complete lesion position in the white matter is found by locating the initial seed point of the lesion signal and diffusing the initial seed point through the algorithm, so that a white matter high signal location map is obtained. The method can clearly and accurately position the position and the volume of the white matter high signal, is easy to operate, has strong generalization performance, stable effect and strong practicability, and is a very effective white matter high signal extraction method.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above-mentioned contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modification made on the basis of the technical idea of the present invention falls within the protection scope of the claims of the present invention.

Claims (10)

1. A white matter high signal extraction method based on a lesion diffusion algorithm is characterized by comprising the following steps:
s1, inputting an original image comprising a FLAIR image, a T1 weighted image and a tissue prior probability map of white matter;
s2, carrying out preprocessing operation on the original image input in the step S1;
s3, performing abnormal signal product operation on the image preprocessed in the step S2 to obtain a tissue signal intensity distribution confidence map;
s4, initializing the tissue signal intensity distribution confidence map obtained in the step S3 and carrying out white matter high signal segmentation to obtain a white matter high signal probability map;
and S5, performing image binarization on the white matter high signal probability map generated after the segmentation processing in the step S4 through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
2. The method according to claim 1, characterized in that in step S2, the preprocessing includes magnetic field offset correction, image registration, generation of partial volume images, tissue segmentation of gray matter white matter cerebrospinal fluid and transformation of the images from standard space to original space.
3. The method of claim 2, wherein generating the partial volume image comprises generating a partial volume image tag using a T1 weighted image, wherein the tag value is between 1 and 3, 1 represents cerebrospinal fluid, and 2 generationsThe grey matter, 3 represents white matter, a discrete label is assigned to each voxel, a discrete tissue label z is assigned to each voxeliComprises the following steps:
Figure FDA0003085484460000011
where i represents each voxel, xiRepresenting voxel intensity values labeled by the partial volume image.
4. The method as claimed in claim 1, wherein in step S3, tissue signal intensity distribution confidence maps of cerebrospinal fluid, gray matter and white matter are obtained together according to the pre-processed FLAIR image, partial volume image of T1 and prior probability result of white matter, the tissue signal intensity distribution confidence maps are superposed to obtain a complete signal intensity distribution confidence map B, and the signal intensity distribution confidence map B is obtained according to the resultGMAnd taking a threshold value kappa to initialize a lesion point.
5. The method of claim 4, wherein the signal strength distribution confidence map B is:
B＝BCSF+BGM+BWM
wherein, BCSF、BGM、BWMConfidence maps of tissue signal intensity distributions for cerebrospinal fluid, gray matter and white matter, respectively.
6. The method of claim 4, wherein the threshold k is 0.15-0.5.
7. The method according to claim 1, wherein step S4 is specifically:
starting point LinitDividing the nearby voxels into lesion and other voxels, judging the initial point L of the lesion by using the other voxels to be cerebrospinal fluid, gray matter or white matterinitWhether the nearby point of (a) is a lesion; if the nearby point is a lesion, judging whether the nearby point of the corresponding point is a lesion, and sharing the lesion with the voxel by an iterative growth algorithmVoxels of the common boundary are also set as lesion voxels until the lesion stops spreading, voxel lesion estimate
Figure FDA0003085484460000021
0 to 1, from the initial point
Figure FDA0003085484460000022
The value is set to 1, and when updating the lesion distribution parameters alpha, beta, consideration is given to
Figure FDA0003085484460000023
While only the voxel distribution parameter theta is taken into account when updating the other voxel distribution parameters theta
Figure FDA0003085484460000024
The lesion low probability voxels of (a); when there is no new one
Figure FDA0003085484460000025
The iteration of the voxels is stopped to obtain a lesion probability map, and the probability of the voxel lesion increase around the lesion voxels represented by the Markov random field theory is substituted into the lesion probability map to obtain an optimized high-signal probability map of the white matter of the brain.
8. The method according to claim 7, wherein iterating through the iterative growth algorithm is specifically:
Figure FDA0003085484460000031
wherein i indicates that at least one neighboring voxel j satisfies
Figure FDA0003085484460000032
biFor the signal intensity distribution confidence map of voxel i,
Figure FDA0003085484460000033
respectively, an estimated value of the distribution parameter, P, at the previous momentOtherProbability that a voxel does not belong to a lesion, PLesIs the probability that a voxel belongs to a lesion.
9. The method of claim 7, wherein the optimized high signal probability map of white matter is as follows:
Figure FDA0003085484460000034
wherein, PLesIs the probability of a voxel belonging to a lesion, yiFor the FLAIR signal intensity of voxel i,
Figure FDA0003085484460000035
Figure FDA0003085484460000036
respectively, an estimate of the distribution parameter at the previous moment, biFor the signal intensity distribution confidence map of voxel i,
Figure FDA0003085484460000037
is a lesion probability map, P, of voxels j adjacent to voxel iOtherIs the probability that a voxel does not belong to a lesion.
10. A white matter high signal extraction system based on a lesion diffusion algorithm is characterized by comprising:
an input module which inputs an original image;
the preprocessing module is used for preprocessing the original image input by the input module;
the initialization module is used for performing abnormal signal product operation on the preprocessed image to obtain a tissue signal intensity distribution confidence map;
the lesion extraction module is used for initializing the obtained tissue signal intensity distribution confidence map and segmenting white matter high signals;
and the post-processing module is used for carrying out image binarization on the white matter high signal probability map generated after the segmentation processing of the lesion extraction module through a threshold value to obtain a white matter high signal extraction map, and superposing the white matter high signal extraction map on the normalized original image to obtain a white matter high signal localization map.
CN202110579377.7A 2021-05-26 2021-05-26 White matter high signal extraction method and system based on lesion diffusion algorithm Active CN113256654B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110579377.7A CN113256654B (en) 2021-05-26 2021-05-26 White matter high signal extraction method and system based on lesion diffusion algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110579377.7A CN113256654B (en) 2021-05-26 2021-05-26 White matter high signal extraction method and system based on lesion diffusion algorithm

Publications (2)

Publication Number Publication Date
CN113256654A true CN113256654A (en) 2021-08-13
CN113256654B CN113256654B (en) 2023-09-12

Family

ID=77184827

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110579377.7A Active CN113256654B (en) 2021-05-26 2021-05-26 White matter high signal extraction method and system based on lesion diffusion algorithm

Country Status (1)

Country Link
CN (1) CN113256654B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113777546A (en) * 2021-09-13 2021-12-10 西安交通大学医学院第一附属医院 Magnetic resonance multi-parameter white matter high-signal quantification method based on three-dimensional mapping
CN117934689A (en) * 2024-03-25 2024-04-26 四川省医学科学院·四川省人民医院 Multi-tissue segmentation and three-dimensional rendering method for fracture CT image

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030009098A1 (en) * 2001-04-05 2003-01-09 Jack Clifford R. Histogram segmentation of FLAIR images
EP2996085A1 (en) * 2014-09-09 2016-03-16 icoMetrix NV Method and system for analyzing image data
CN108171697A (en) * 2018-01-05 2018-06-15 北京航空航天大学 A kind of WMH automatic extracting systems based on cluster
CN110046646A (en) * 2019-03-07 2019-07-23 深圳先进技术研究院 Image processing method, calculates equipment and storage medium at system
CN110991408A (en) * 2019-12-19 2020-04-10 北京航空航天大学 Method and device for segmenting white matter high signal based on deep learning method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030009098A1 (en) * 2001-04-05 2003-01-09 Jack Clifford R. Histogram segmentation of FLAIR images
EP2996085A1 (en) * 2014-09-09 2016-03-16 icoMetrix NV Method and system for analyzing image data
CN108171697A (en) * 2018-01-05 2018-06-15 北京航空航天大学 A kind of WMH automatic extracting systems based on cluster
CN110046646A (en) * 2019-03-07 2019-07-23 深圳先进技术研究院 Image processing method, calculates equipment and storage medium at system
CN110991408A (en) * 2019-12-19 2020-04-10 北京航空航天大学 Method and device for segmenting white matter high signal based on deep learning method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
周震;马斌荣;: "结合水平集和区域生长的脑MR图像分割", 北京生物医学工程, no. 01 *
杨俊;李娜;李迟迟;杨泽鹏;周寿军;: "基于高斯混合模型和马尔科夫随机场的脑MR图像分割", 解剖学研究, no. 05 *
王啸;余永强;钱银锋;潘义广;彭传勇;: "苯丙酮尿症脑部病变扩散加权成像", 临床放射学杂志, no. 06 *
赵波;郑兴华;白璐;杨勇;: "脑白质疏松症MR图像病变区域的量化分析", 杭州电子科技大学学报, no. 06 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113777546A (en) * 2021-09-13 2021-12-10 西安交通大学医学院第一附属医院 Magnetic resonance multi-parameter white matter high-signal quantification method based on three-dimensional mapping
CN117934689A (en) * 2024-03-25 2024-04-26 四川省医学科学院·四川省人民医院 Multi-tissue segmentation and three-dimensional rendering method for fracture CT image

Also Published As

Publication number Publication date
CN113256654B (en) 2023-09-12

Similar Documents

Publication Publication Date Title
US11379985B2 (en) System and computer-implemented method for segmenting an image
Almajalid et al. Development of a deep-learning-based method for breast ultrasound image segmentation
Chen et al. Automatic segmentation of individual tooth in dental CBCT images from tooth surface map by a multi-task FCN
Nain et al. Vessel segmentation using a shape driven flow
Banerjee et al. Automated 3D segmentation of brain tumor using visual saliency
Alilou et al. A comprehensive framework for automatic detection of pulmonary nodules in lung CT images
Passat et al. Magnetic resonance angiography: From anatomical knowledge modeling to vessel segmentation
Chi et al. X-Net: Multi-branch UNet-like network for liver and tumor segmentation from 3D abdominal CT scans
EP3213296A1 (en) A method and system for image processing
CN110910405A (en) Brain tumor segmentation method and system based on multi-scale cavity convolutional neural network
CN110036408A (en) The automatic ct of active hemorrhage and blood extravasation detection and visualization
CN113256654A (en) White matter high signal extraction method and system based on lesion diffusion algorithm
Jaffar et al. Anisotropic diffusion based brain MRI segmentation and 3D reconstruction
EP4118617A1 (en) Automated detection of tumors based on image processing
Vandermeulen et al. Local filtering and global optimization methods for 3-D magnetic-resonance angiography image enhancement
Zhao et al. 3D Brain Tumor Image Segmentation Integrating Cascaded Anisotropic Fully Convolutional Neural Network and Hybrid Level SetMethod.
Ananth Geo-cutting Liver Tumor
Lopes et al. A methodical approach for detection and 3-D reconstruction of brain tumor in MRI
CN112862785B (en) CTA image data identification method, device and storage medium
Patil et al. A robust system for segmentation of primary liver tumor in CT images
Reska et al. Fast 3D segmentation of hepatic images combining region and boundary criteria
CN112862786A (en) CTA image data processing method, device and storage medium
Ali et al. Automatic technique to produce 3D image for brain tumor of MRI images
Badura et al. Automatic 3D segmentation of renal cysts in CT
Li et al. A new efficient 2D combined with 3D CAD system for solitary pulmonary nodule detection in CT images

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