CN109685814B - Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN - Google Patents

Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN Download PDF

Info

Publication number
CN109685814B
CN109685814B CN201910001478.9A CN201910001478A CN109685814B CN 109685814 B CN109685814 B CN 109685814B CN 201910001478 A CN201910001478 A CN 201910001478A CN 109685814 B CN109685814 B CN 109685814B
Authority
CN
China
Prior art keywords
segmentation
gallbladder
calculus
region
binary image
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.)
Active
Application number
CN201910001478.9A
Other languages
Chinese (zh)
Other versions
CN109685814A (en
Inventor
廉敬
石斌
杨臻
马义德
刘冀钊
孙文灏
杜晓刚
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.)
Lanzhou Jiaotong University
Original Assignee
Lanzhou 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 Lanzhou Jiaotong University filed Critical Lanzhou Jiaotong University
Priority to CN201910001478.9A priority Critical patent/CN109685814B/en
Publication of CN109685814A publication Critical patent/CN109685814A/en
Application granted granted Critical
Publication of CN109685814B publication Critical patent/CN109685814B/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
    • 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
    • G06T2207/10132—Ultrasound image
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)

Abstract

The invention discloses a full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN, which comprises the following steps: segmenting the ultrasonic image by adopting an MSPCNN algorithm to obtain a gallbladder rough segmentation binary image; segmenting the gallbladder rough segmentation binary image by adopting a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image; and respectively carrying out post-processing on the gallbladder accurate segmentation binary image and the calculus accurate segmentation binary image by adopting a local weighted linear regression algorithm to smoothen the edge contour of the gallbladder calculus and finally obtain a gallbladder region segmentation result and a calculus region segmentation result. The method has the advantages of reducing the calculation complexity, reducing the segmentation steps and improving the image segmentation speed and precision.

Description

Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN
Technical Field
The invention relates to the field of image processing, in particular to a full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN.
Background
Gallstone is a common gallbladder disease, for example, the stone incidence rate of natural population in B-mode ultrasonic examination in northwest of China is about 15%. In recent years, medical imaging technologies such as CT, magnetic resonance MRI, and ultrasound have been rapidly developed. Because the ultrasonic image imaging examination has the advantages of low cost, strong real-time performance, no damage and no electromagnetic radiation, the ultrasonic image examination is generally used as an initial diagnosis means for the cholecystolithiasis.
The purpose of gallbladder calculus ultrasonic image segmentation is to rapidly and accurately segment a gallbladder region and a calculus region, so that a doctor can obtain valuable complete diagnostic information at the initial stage of image examination. At present, the gallbladder stone ultrasonic image segmentation mainly adopts manual and semi-automatic segmentation methods of doctors, the time is long, and the method depends on the subjective experience and knowledge of operators to different degrees, and the diagnosis conclusion obtained by the doctors is inevitably mixed with subjective factors. The requirements of clinical practice on the speed and accuracy of segmentation cannot be fully met. Therefore, the full-automatic segmentation of the cholecystolithiasis image becomes a new subject focused on and researched by the current academic community, and a computer-aided diagnosis system supported by the image segmentation technology with strong robustness is established and is the focused research direction.
In the prior art, a full-automatic segmentation method based on Pulse Coupled Neural Network (PCNN) realizes computer full-automatic segmentation, and segments gallbladder regions and stone regions with clear edge contour lines as shown in (a) and (b) in fig. 1. In fig. 1, the gallbladder region is shown in the contour line 2, and the calculus region is shown in the contour line 3. The segmentation method divides the cholecystolithiasis ultrasonic image into four steps of preprocessing, gallbladder region segmentation, calculus region segmentation and post-processing, and has six segmentation algorithms in total. Wherein, the preprocessing step adopts an improved Otsu threshold method and an improved anisotropic diffusion algorithm; the gallbladder region segmentation step adopts a morphological algorithm; the calculus region segmentation step adopts a Simplified Pulse Coupled Neural Network (SPCNN) and an improved region growing algorithm; the post-processing step adopts a local weighted linear regression algorithm. The overall flow path is shown in fig. 2.
However, the above method still has the following defects and shortcomings:
the method has the advantages that four segmentation steps and six segmentation algorithms are provided, the step algorithms are complex, the calculation amount is large, and the steps and the algorithm types need to be merged and reduced, so that the calculation complexity is reduced, and the segmentation links are reduced.
Secondly, parameters of the full-automatic segmentation algorithm are obtained in a self-adaptive mode according to image attribute values, compared with the semi-automatic segmentation method that parameters are set according to empirical values, the method is obviously improved, the set parameters are still more, the calculation formula and the calculation process are complicated and redundant, the parameters need to be simplified, and the algorithm simplicity is improved.
In the step of dividing the calculus area, three iterations are needed to obtain the calculus area dividing result, and if the iteration times can be reduced, the dividing process can be converged, and the image dividing efficiency is improved.
And fourthly, a pulse coupled neural network model (PCNN) introduced in the calculus region segmentation step is an image processing tool which is more in line with human eye visual features, the early-stage research is not deep enough, the advantages of the PCNN are not fully developed and displayed, and the full-automatic segmentation speed and precision still have improved space.
Disclosure of Invention
The invention aims to provide a full-automatic gallbladder calculus ultrasonic image segmentation method based on MSPCNN, aiming at the problems, so as to achieve the advantages of reducing the computational complexity, reducing the segmentation steps and improving the image segmentation speed and precision.
In order to achieve the purpose, the invention adopts the technical scheme that:
a full-automatic segmentation method of a cholecystolithiasis ultrasonic image based on MSPCNN comprises the following steps:
segmenting the ultrasonic image by adopting an MSPCNN algorithm to obtain a gallbladder rough segmentation binary image;
segmenting the gallbladder rough segmentation binary image by adopting a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image;
and respectively processing the gallbladder accurate segmentation binary image and the calculus accurate segmentation binary image by adopting a local weighted linear regression algorithm so as to obtain a gallbladder region segmentation result and a calculus region segmentation result.
Optionally, the segmenting the ultrasound image by using the MSPCNN algorithm to obtain the gallbladder rough segmentation binary image includes:
determining variable threshold parameter S in MSPCNN algorithmnThe value of the number k;
deriving k parameters S based on the value of the number knThe value of (c).
Optionally, the calculation formula of the number k is as follows:
Figure BDA0001933743970000031
and S' is the normalized Otsu threshold of the initial gallstone image.
Optionally, the value based on the number k is used to obtain k parameters SnThe values of (a) are specifically:
Figure BDA0001933743970000032
and S' is the normalized Otsu threshold of the initial gallstone image.
Optionally, the value based on the number k is used to obtain k parameters SnAfter the step of calculating, further comprising:
the k parameters S will be obtainednRespectively substituting the values into an MSPCNN algorithm to obtain k candidate gallbladder region rough segmentation binary images;
removing regions connected with the boundary line of the whole image from the k candidate gallbladder region coarse segmentation binary images to obtain k residual regions, calculating the area of each residual region, and selecting a residual region with the largest area, wherein the residual region with the largest area is the gallbladder coarse segmentation region, and the binary image comprising the region is the gallbladder coarse segmentation binary image.
Optionally, the segmenting the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image, includes:
the gallbladder rough segmentation region is of a disk type with a radius of RmaxPerforming morphological closure operation on the structural elements to form a calculus-containing regionContained in the gallbladder rough segmentation region; the thick segmentation region of the gallbladder with calculus region adopts a disk type with a radius of RminAnd performing morphological opening operation on the structural elements to obtain a gallbladder accurate segmentation binary image.
Optionally, the segmenting the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image, includes:
in the gallbladder accurate segmentation binary image, radius values R in structural elements are dividedmaxDetermining the distance from the precise gallbladder partition region to each side, and cutting the image to obtain the cholecystolithiasis region.
Optionally, the segmenting the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image, includes:
removing the gallbladder rough segmentation binary image from the gallbladder precise segmentation binary image to obtain a plurality of remaining binary regions, and selecting a region with the largest area from the remaining binary regions to obtain a calculus rough segmentation region;
the calculus rough segmentation region is of a disk type with a radius of RminPerforming morphological closing operation on the structural elements, and including holes containing a calculus area in the calculus rough segmentation area; and performing morphological opening operation on the calculus rough segmentation region containing the holes of the calculus region to obtain an accurate calculus segmentation binary image.
Optionally, after obtaining the calculus accurate segmentation binary image, the method further includes:
adopts a disk type with a radius of RmidAnd performing morphological closure operation on the structural elements, and combining the gallbladder accurate segmentation region and the calculus accurate segmentation region into a region to obtain a combined region completely containing the calculus accurate segmentation region.
The technical scheme of the invention has the following beneficial effects:
the technical scheme of the invention mainly adopts the MSPCNN algorithm, improves the existing SPCNN algorithm, further shows the variable threshold characteristic, the nonlinear modulation characteristic, the synchronous pulse release phenomenon, the capture characteristic, the dynamic pulse release phenomenon, the automatic wave characteristic and the comprehensive space-time characteristic of the PCNN, and is combined with other algorithms to finish the full-automatic segmentation of the cholecystolithiasis ultrasonic image. According to the experimental data and the analysis and evaluation, the technical effects of the invention are expressed in the following aspects:
1. the MSPCNN algorithm combines two preprocessing algorithms and a gallbladder region segmentation part algorithm in the former method, simplifies the original four segmentation steps of preprocessing, gallbladder region segmentation, calculus region segmentation and postprocessing into three segmentation steps of gallbladder region segmentation, calculus region segmentation and postprocessing, and simplifies the original six algorithm types into four, thereby reducing segmentation steps and calculation links and reducing calculation complexity.
2. The MSPCNN algorithm only needs to calculate the value of one variable threshold parameter Sn, three parameters set by the SPCNN are simplified into one parameter, the algorithm simplification degree is improved, and the calculation formula and the operation process are simplified.
3. The MSPCNN algorithm changes the iteration times of automatic segmentation from three times to two times, the segmentation process is converged, and the image segmentation efficiency is improved.
4. More importantly, the MSPCNN algorithm further exerts the advantages of high accuracy and low complexity of the PCNN model compared with the traditional model, not only retains the characteristics of PCNN capture, synchronous ignition and the like, but also improves the description capacity of the model on the pixel space, so that the network model is more in line with the characteristics of human eyes, and the segmentation speed and the segmentation accuracy are further improved.
The technology of the invention better meets the clinical requirement of quickly and accurately segmenting the cholecystolithiasis image, the full-automatic segmentation of the cholecystolithiasis image is also a precondition for automatically identifying and classifying the cholecystolithiasis, and the automatic classification of the types of gallstones such as full moon shape, crescent moon shape, sediment shape and the like can be realized by depending on the support of the full moon shape, the crescent moon shape, the sediment shape and the like, thereby providing a basis for a doctor to formulate an individualized accurate diagnosis and treatment scheme and finally realizing the aim of automatically locking the alarm prompt of a calculus area and automatically generating the diagnosis and treatment scheme by the doctor using an ultrasonic diagnostic apparatus.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a diagram illustrating a segmentation result of a cholecystolithiasis;
FIG. 2 is a flowchart of ultrasonic image segmentation of cholecystolithiasis'
FIG. 3 is a flowchart of a MSPCNN-based gallbladder stone ultrasound image full-automatic segmentation method according to an embodiment of the present invention;
fig. 4 is a characteristic diagram of the MSPCNN-based gallbladder stone ultrasonic image full-automatic segmentation method in the segmentation process according to the embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in conjunction with the accompanying drawings, and it will be understood that they are described herein for the purpose of illustration and explanation and not limitation.
PCNN is an abbreviation for Pulse-coupled neural network in english, and the chinese table is called Pulse-coupled neural network. At present, pulse coupled neural networks are commonly known by the academic community as PCNN. S is an abbreviation for simplifed and M is a modified abbreviation. MSPCNN refers to an improved simplified pulse-coupled neural network model.
Pulse Coupled Neural Network (PCNN) characteristics and advantages:
in 1989, Eckhorn simulates the dynamic synchronous activity characteristics of cortical neurons in the cat brain, and proposes a link domain model, and in 1993, Johnson et al propose an improved neuron model named Pulse Coupled Neural Network (PCNN). The PCNN neuron better simulates biological neurons than traditional artificial neurons, transverse connection among all pulse coupling neurons does not need any training, and domestic and foreign researches prove that in medical image processing, due to natural connection of the PCNN and a visual nervous system and finer simulation of visual cortex of mammals, the threshold value of each neuron in a model is attenuated and changed according to time according to an exponential law, the nonlinear characteristic of human vision response to brightness intensity is met, and segmented images have better visual effect. Meanwhile, the PCNN has more setting parameters and larger calculation amount, and the parameters need to be simplified to reduce the calculation complexity. The PCNN theory is still not mature at present, the relationship between the model parameters and the image processing mechanism is not clear, and the PCNN hopes to naturally process images in real time like the human eyes, so that a plurality of problems need to be solved.
A full-automatic segmentation method of a cholecystolithiasis ultrasonic image based on MSPCNN comprises the following steps:
segmenting the ultrasonic image by adopting an MSPCNN algorithm to obtain a gallbladder rough segmentation binary image;
segmenting the gallbladder rough segmentation binary image by adopting a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image;
and respectively carrying out post-processing on the gallbladder accurate segmentation binary image and the calculus accurate segmentation binary image by adopting a local weighted linear regression algorithm so as to obtain a gallbladder region segmentation result and a calculus region segmentation result.
Optionally, the segmenting the ultrasound image by using the MSPCNN algorithm to obtain the gallbladder rough segmentation binary image includes:
determining variable threshold parameter S in MSPCNN algorithmnThe value of the number k;
deriving k parameters S based on the value of the number knThe value of (c).
Optionally, the calculation formula of the number k is as follows:
Figure BDA0001933743970000071
and S' is the normalized Otsu threshold of the initial gallstone image.
Optionally, the value based on the number k is used to obtain k parameters SnThe values of (a) are specifically:
Figure BDA0001933743970000072
and S' is the normalized Otsu threshold of the initial gallstone image.
Optionally, the value based on the number k is used to obtain k parameters SnAfter the step of calculating, further comprising:
the k parameters S will be obtainednRespectively substituting the values into an MSPCNN algorithm to obtain k candidate gallbladder region segmentation binary images;
removing regions connected with the boundary line of the whole image in the k candidate gallbladder region segmentation binary images to obtain k residual regions, calculating the area of each residual region, and selecting a residual region with the largest area, wherein the residual region with the largest area is a gallbladder rough segmentation region, and the binary image comprising the region is the gallbladder rough segmentation binary image.
Optionally, the segmenting the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image, includes:
the gallbladder rough segmentation region is of a disk type with a radius of RmaxPerforming a morphological closure operation on the structural elements so as to contain the pores of the calculus-containing region in the thick gallbladder segmentation region; the thick segmentation region of the gallbladder with calculus region adopts a disk type with a radius of RminAnd performing morphological opening operation on the structural elements to obtain a gallbladder accurate segmentation binary image.
Optionally, the segmenting the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image, includes:
in the gallbladder accurate segmentation binary image, radius values R in structural elements are dividedmaxDetermining the distance from the precise gallbladder partition region to each side, and cutting the image to obtain the cholecystolithiasis region.
Optionally, the segmenting the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image, includes:
removing the gallbladder rough segmentation binary image from the gallbladder precise segmentation binary image to obtain a plurality of remaining binary regions, and selecting a region with the largest area from the remaining binary regions to obtain a calculus rough segmentation region;
the calculus rough segmentation region is of a disk type with a radius of RminPerforming morphological closing operation on the structural elements, and including holes containing a calculus area in the calculus rough segmentation area; and performing morphological opening operation on the calculus rough segmentation region containing the holes of the calculus region to obtain an accurate calculus segmentation binary image.
Optionally, after obtaining the calculus accurate segmentation binary image, the method further includes:
adopts a disk type with a radius of RmidAnd performing morphological closure operation on the structural elements, and combining the gallbladder accurate segmentation region and the calculus accurate segmentation region into a region to obtain a combined region completely containing the calculus accurate segmentation region.
In one particular application scenario, as shown in figure 3,
the method comprises the following steps of;
the following algorithm is adopted in the gallbladder region segmentation step:
determining variable threshold parameter S in MSPCNN algorithmnThe calculation formula of the number k, k is as follows:
Figure BDA0001933743970000081
in the formula (1), the parameter S ' is the normalized greater saliva threshold of the initial cholecystolithiasis image, the value of the power of 1+ S ' of the parameter S ' is calculated first, the reciprocal is taken, the value is rounded off to obtain a new value, the value is determined as the value of the number k, and then the value is substituted into the formula (2) to obtain the parameter SnK values of
Figure BDA0001933743970000091
Secondly, the MSPCNN algorithm is adopted to realize the segmentation pretreatment of the ultrasonic image and the initial segmentation of the gallbladder region, and the result obtained in the step one isK parameters S ofnThe obtained values are respectively substituted into an MSPCNN algorithm to obtain k candidate gallbladder region segmentation binary images. The MSPCNN algorithm is as follows:
Fij[n]＝Sij(3)，
Figure BDA0001933743970000092
Figure BDA0001933743970000093
Figure BDA0001933743970000094
Eij[n]＝e-αEij[n-1]+VYij[n](7)，
wherein
Figure BDA0001933743970000095
Figure BDA0001933743970000096
Figure BDA0001933743970000097
V＝1+Sn 2-Sn 3(11)。
In formulas (3) to (11), neuron NijTwo inputs are included at location (i, j): the first input is the feed input Fij[n]It is excited by external excitation SijRepresenting, by the action of external factors; the second input is the link input Lij[n]It is composed of a synaptic weight matrix WijklAnd the output Y of the neighboring neurons of the previous iterationkl[n-1]The product of (a), results from peripheral neuronal action. U shapeij[n]Is neuron NijThe internal activity item of (2) is composed of two parts: part of previous iteration of internal activity itemResult Uij[n-1]And an exponential decay factor e-αIs determined by the state of the previous iteration of the neuron; the other part is input L by simplified linkij[n]And a feed input Fij[n]The result of the modulation is determined by the peripheral neurons and by external stimuli. Beta represents the strength of the connection of a neuron with its surrounding neurons, and the larger the value of beta, the more closely the neurons are connected with each other. Eij[n]For dynamic threshold values, V and e-αRespectively representing the magnitude and exponential decay coefficient of the dynamic threshold. SnThe MSPCNN model neurons and the image pixels have a one-to-one correspondence relationship, which is an output value obtained by calculation in formula (2). In each iteration, when the internal activity item Uij[n]Greater than dynamic threshold Eij[n-1]When it is time, the neuron fires, otherwise it does not fire.
And thirdly, after k binary images are obtained, removing regions connected with the boundary line of the whole image, calculating the area of each residual region, and selecting a region with the largest area from the k images, wherein the region is a gallbladder rough segmentation region, and the binary image comprising the region is the gallbladder rough segmentation binary image.
Fourthly, adopting the type of disk and the radius of R for the coarse segmentation area of the gall bladdermaxIs subjected to a morphological closing operation (R)maxAnd R mentioned belowmidAnd RminAll from the documents Jing Lian, Yide Ma, Yurun Ma, Bin Shi, Jizhao Liu, Zhen Yang, Yanan Guo]International journal of calculated radio and surgery,12(4): 553-; adopts a disk type with a radius of RmidThe structural elements of the gallbladder are subjected to morphological opening operation, so that the spines of the edge outline can be removed, and a gallbladder accurate segmentation binary image is obtained.
Fifthly, in the accurate gallbladder segmentation binary image, dividing the radius value R in the structural elementsmaxDetermining the distance from the precise gallbladder partition region to each side, cutting the image, and locating the cholecystolithiasis region.
The following algorithm is adopted in the calculus region segmentation step:
1. since the calculus is usually located in the gallbladder, the gallbladder obtained in the gallbladder region segmentation step (c) is accurately segmented into binary images, the gallbladder obtained in the gallbladder region segmentation step (c) is subtracted from the gallbladder roughly segmented binary images, and the region with the largest area is selected from the remaining binary images and determined as the roughly segmented region of the calculus.
2. The calculus rough segmentation region is of a disk type with a radius of RminPerforming morphological closing operation on the structural elements to contain the holes of the calculus-containing area in the area; and performing morphological opening operation to remove sharp thorns on the edge contour and obtain the calculus accurate segmentation binary image.
The following algorithm is adopted in the post-processing segmentation step:
firstly, the adopted type is 'disk' type, and the radius is RmidAnd performing morphological closure operation on the structural elements, and combining the gallbladder accurate segmentation region and the calculus accurate segmentation region into a region, so that the obtained combined region completely contains the calculus accurate segmentation region.
Secondly, a local weighted linear regression algorithm (LOESS) is adopted for the precise gallbladder segmentation region to obtain a finally segmented gallbladder region.
Thirdly, obtaining the finally segmented calculus area by adopting a local weighted linear regression algorithm (LOESS) for the calculus accurate segmentation area.
The key algorithm of the technical scheme of the invention is MSPCNN algorithm. The PCNN algorithm is obviously improved compared with the popular PCNN algorithm and SPCNN algorithm, so that the full-automatic segmentation of the target image aimed at by the technical scheme of the invention is realized by the combined application of the PCNN algorithm and other algorithms, the unique advantages of the PCNN in image processing are further shown, the computational complexity of image segmentation is reduced, the process is converged, and the speed and the precision are improved. The MSPCNN algorithm plays a key role in a full-automatic segmentation algorithm system, and the optimization of the existing method and the upgrading of the existing technology are realized.
The beneficial effects of the invention can be obtained by comparing specific experimental data, which are as follows:
the technology of the invention compares the full-automatic segmentation method based on the Pulse Coupled Neural Network (PCNN) with popular manual and semi-automatic technologies to verify the effect.
The ultrasonic images of the cholecystolithiasis of the experiment are all from clinical diagnosis cases in national hospitals of Gansu province. Comparative evaluation OF the segmentation was carried out using five evaluation indices OF, OV, DSI, PE and T, the first four OF which were derived from the respective petronella Anbeek, Koen L Vincken, Matthias J P Van Osch, Robertus H C Bisschops, Jeronen Vander Grond. Probalistic segmentation OF white matter segmentation in MRimaging [ J ] Neuroidea, 21(3) 1037-. Wherein, the evaluation indexes OF, OV, and DSI are from the literature Petronella Anbeek, Koen L Vincken, Matthias J P Van Osch, Robertus H C Bisschops, Jeroen Vander connected, Probalistic segmentation OF white matter segmentation in MRimagining [ J ] Neuromage, 21(3): 1037-1044, 2004, which are the evaluation (ratio) OF the similarity between the segmentation region obtained by the physician's manual method and the segmentation region obtained by the full-automatic or semi-automatic segmentation algorithm, and the value ranges from 0 to 1, the closer the value is to the segmentation region obtained by the full-automatic or semi-automatic segmentation algorithm and the physician's manual method, the better the segmentation effect, and vice versa. The evaluation index PE is from the documents C Alberola-Lopez, M Martin-Femandez, J Ruiz-Alzola. A method for evaluating the approximation degree of boundary detection on the median images [ J ] IEEE, is an evaluation (in mm) of the closeness degree of the edge contour obtained by the manual method of a doctor and the edge contour obtained by the full-automatic or semi-automatic segmentation algorithm, and the numerical value is closer to 0, which indicates that the closer the edge contour obtained by the manual method of the doctor is to the full-automatic or semi-automatic segmentation algorithm, the better the segmentation effect is, and vice versa. The last evaluation index T is the evaluation (in seconds) of the time complexity of the segmentation algorithm, and the smaller the numerical value is, the smaller the time complexity of the segmentation algorithm is, and vice versa.
The effectiveness of the segmentation method is verified by comparing the segmentation method, the PCNN, the Snake-GVF algorithm (GVF), the Snake-Distance algorithm (Distance) and the Snake-Balloon algorithm (Balloon)' with manual segmentation results obtained by two provincial professionals from radiology departments of people hospitals in Gansu province respectively, and experimental data are shown in tables 1 and 2.
As can be seen from the gallbladder region segmentation results in Table 1, the average values OF the evaluation indexes OF, OV and DSI OF the segmentation method are the highest in the five algorithms, the average value OF the evaluation index PE is the lowest, and the segmentation effect is the best. In the aspect of operation time T, the operation time of the segmentation method is only longer than that of the Balloon algorithm, and the segmentation method has obvious advantages compared with other algorithms.
Figure BDA0001933743970000141
Table 1: comparison graph of gallbladder region segmentation results (provincial expert 1 and provincial expert 2).
Figure BDA0001933743970000151
Table 2: calculus region segmentation result comparison graph (provincial expert 1 and provincial expert 2).
As can be seen from the calculus region segmentation results in Table 2, the average values OF the evaluation indexes OF, DSI and T OF the segmentation method are the highest in the five algorithms, and the evaluation indexes OV and PE are equivalent to the evaluation value results OF the segmentation method proposed by the previous team.
In order to obtain the final evaluation result OF the effectiveness OF the segmentation method, the Mean value OF each evaluation index given by the two provincial experts in the experiment is added to obtain the average value, then the average value OF the evaluation indexes OF, OV and DSI is calculated to obtain a comprehensive evaluation index EVA, the comprehensive evaluation index and other two evaluation indexes PE and T are jointly used as the evaluation indexes OF the final result, and the final evaluation results OF the gall bladder area and the calculus area are respectively shown in tables 3 and 4.
Experimental methods EVA PE T
The method of the invention 0.8851 4.3149 2.3905
PCNN 0.8611 5.5183 2.9619
GVF 0.6538 15.0764 32.6468
Distance 0.7209 11.0328 3.4529
Balloon 0.6634 19.4727 0.7535
Table 3: and finally evaluating the result of gallbladder region segmentation.
Experimental methods EVA PE T
The method of the invention 0.8363 1.7157 0.1646
PCNN 0.8021 1.6972 0.4561
GVF 0.7073 2.5036 16.9515
Distance 0.6909 2.4565 2.0835
Balloon 0.6797 3.9722 0.5777
Table 4: and (5) dividing the calculus area to finally evaluate the result.
FIG. 4 is a diagram of the segmentation process of the method of the present invention, wherein sub-images (a) and (i) are original images of ultrasound images of cholecystolithiasis; sub-images (b) and (j) are gallbladder rough segmentation binary images obtained by the gallbladder region segmentation step (the third step in the gallbladder segmentation step); sub-graph (c) and sub-graph (k) are gallbladder accurate segmentation binary images obtained by the gallbladder region segmentation step (the fourth step in the gallbladder segmentation step); sub-graph (d) and sub-graph (l) are positioned gallbladder calculus binary images obtained by the gallbladder region segmentation step (the fifth step in the gallbladder segmentation step); subgraph (e) and subgraph (m) are accurate calculus segmentation binary images obtained by the calculus region segmentation step (the second step in the calculus segmentation step); sub-graph (f) and sub-graph (n) are gallbladder region segmentation results (gallbladder region determined by physician in No. 1 contour line) obtained by provincial expert 1 of Min Hospital of Gansu province through a manual method; sub-graph (g) and sub-graph (o) are calculus region segmentation results (calculus region determined by physician in No. 1 contour line) obtained by provincial expert 1 of Min Hospital of Gansu province through a manual method; the subgraph (h) and the subgraph (p) are the segmentation results obtained by the segmentation method of the technology (the gallbladder region is in the No. 2 contour line, and the calculus region is in the No. 3 contour line). In the figure, 1 is the contour line No. 1, 2 is the contour line No. 2, and 3 is the contour line No. 3.
The images used for the test of the experiment are all from people hospitals in Gansu province, and 235 gallbladder calculus ultrasonic images from clinical cases form an image test library. From the view of the shape of the gallbladder, 176 images of 235 images have regular gallbladder shapes, and 59 images have irregular gallbladder shapes; from the shape of the calculus, 188 out of 235 images only contain one large calculus, 26 images only contain one small calculus, 18 calculus have the width more than one half of the width of the gallbladder, 13 calculus areas occupy more than one half of the area of the gallbladder, and the other 12 calculus are accompanied by heavy sound shadow. Each image has a resolution of 512 x 512 and each pixel has 256 gray levels.
The computer model used in the experiment was the associative rescuer R720-15, the Intel (R) core (TM) i7-7700HQ CPU @2.80GHz, version MATLAB 2014(a) were used in the processor, and the hospital used the Siemens Acuson X300 imaging device in a unified manner.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (6)

1. A full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN is characterized by comprising the following steps:
segmenting the ultrasonic image by adopting an MSPCNN algorithm to obtain a gallbladder rough segmentation binary image;
segmenting the gallbladder rough segmentation binary image by adopting a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image;
respectively carrying out post-processing on the gallbladder accurate segmentation binary image and the calculus accurate segmentation binary image by adopting a local weighted linear regression algorithm so as to obtain a gallbladder region segmentation result and a calculus region segmentation result;
the method for segmenting the ultrasonic image by adopting the MSPCNN algorithm to obtain the gallbladder rough segmentation binary image comprises the following steps:
determining variable threshold parameter S in MSPCNN algorithmnThe value of the number k;
deriving k parameters S based on the value of the number knA value of (d);
the calculation formula of the number k is as follows:
Figure FDA0002641564400000011
s' is the normalized Otsu threshold of the initial gallstone image;
the value based on the number k obtains k parameters SnThe values of (a) are specifically:
Figure FDA0002641564400000012
and S' is the normalized Otsu threshold of the initial gallstone image.
2. The MSPCNN-based gallbladder calculus ultrasound image full-automatic segmentation method of claim 1, wherein the k parameters S are derived based on k valuesnAfter the step of calculating, further comprising:
the k parameters S will be obtainednRespectively substituting the values into an MSPCNN algorithm to obtain k candidate gallbladder region rough segmentation binary images;
removing regions connected with the boundary line of the whole image from the k candidate gallbladder region coarse segmentation binary images to obtain k residual regions, calculating the area of each residual region, and selecting a residual region with the largest area, wherein the residual region with the largest area is the gallbladder coarse segmentation region, and the binary image comprising the region is the gallbladder coarse segmentation binary image.
3. The MSPCNN-based gallbladder calculus ultrasound image full-automatic segmentation method of claim 2, wherein the segmentation of the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image comprises:
the gallbladder rough segmentation region is of a disk type with a radius of RmaxPerforming a morphological closure operation on the structural elements so as to contain the pores of the calculus-containing region in the thick gallbladder segmentation region; the thick segmentation area of the gall bladder containing the calculus area adopts a shape of' diskRadius RminAnd performing morphological opening operation on the structural elements to obtain a gallbladder accurate segmentation binary image.
4. The MSPCNN-based gallbladder calculus ultrasound image full-automatic segmentation method of claim 3, wherein the segmentation of the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image comprises:
in the gallbladder accurate segmentation binary image, radius values R in structural elements are dividedmaxDetermining the distance from the precise gallbladder partition region to each side, and cutting the image to obtain the cholecystolithiasis region.
5. The MSPCNN-based gallbladder calculus ultrasound image full-automatic segmentation method of claim 3, wherein the segmentation of the gallbladder rough segmentation binary image by using a morphological algorithm to obtain a gallbladder accurate segmentation binary image and a calculus accurate segmentation binary image comprises:
removing the gallbladder rough segmentation binary image from the gallbladder precise segmentation binary image to obtain a plurality of remaining binary regions, and selecting a region with the largest area from the remaining binary regions to obtain a calculus rough segmentation region;
the calculus rough segmentation region is of a disk type with a radius of RminPerforming morphological closing operation on the structural elements, and including holes containing a calculus area in the calculus rough segmentation area; and performing morphological opening operation on the calculus rough segmentation region containing the holes of the calculus region to obtain an accurate calculus segmentation binary image.
6. The MSPCNN-based gallbladder calculus ultrasound image fully-automatic segmentation method of claim 5, wherein after obtaining the calculus accurate segmentation binary image, further comprising:
adopts a disk type with a radius of RmidPerforming a morphological closing operation on the structural elements of (1)The gallbladder accurate segmentation area and the calculus accurate segmentation area are combined into one area, and a combined area completely containing the calculus accurate segmentation area is obtained.
CN201910001478.9A 2019-01-02 2019-01-02 Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN Active CN109685814B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910001478.9A CN109685814B (en) 2019-01-02 2019-01-02 Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910001478.9A CN109685814B (en) 2019-01-02 2019-01-02 Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN

Publications (2)

Publication Number Publication Date
CN109685814A CN109685814A (en) 2019-04-26
CN109685814B true CN109685814B (en) 2020-10-23

Family

ID=66190407

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910001478.9A Active CN109685814B (en) 2019-01-02 2019-01-02 Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN

Country Status (1)

Country Link
CN (1) CN109685814B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110310280B (en) * 2019-07-10 2021-05-11 广东工业大学 Image recognition method, system, equipment and storage medium for hepatobiliary duct and calculus
CN110570407B (en) * 2019-08-29 2022-10-28 上海联影智能医疗科技有限公司 Image processing method, storage medium, and computer device
CN112754522A (en) * 2019-11-01 2021-05-07 深圳迈瑞生物医疗电子股份有限公司 Doppler calculus imaging method and ultrasonic imaging device
CN112819712B (en) * 2021-01-22 2022-03-11 兰州交通大学 Low-illumination color image enhancement method based on PNA-MSPCNN model
CN115553753B (en) * 2022-11-17 2023-03-14 武汉楚精灵医疗科技有限公司 Risk early warning device for gallstones and related equipment
CN116363132B (en) * 2023-06-01 2023-08-22 中南大学湘雅医院 Ophthalmic image processing method and system
CN116385436B (en) * 2023-06-02 2023-08-15 胜利油田中心医院 Cholelithiasis auxiliary detection system based on CT image

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101908891A (en) * 2010-08-23 2010-12-08 南京信息工程大学 Medical image ROI (Region of Interest) compression method based on lifting wavelet and PCNN (Pulse Coupled Neural Network)
CN102306289A (en) * 2011-09-16 2012-01-04 兰州大学 Method for extracting iris features based on pulse couple neural network (PCNN)
CN105989607A (en) * 2016-04-13 2016-10-05 西华大学 Catenary image segmentation method based on SPCNN and minimum cross entropy

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN100587717C (en) * 2007-09-18 2010-02-03 湖南大学 Medical large transfusion machine vision on-line detection method
CN104835176B (en) * 2015-01-20 2018-02-27 广西机电职业技术学院 A kind of threshold segmentation method of laser vision image
US10748451B2 (en) * 2016-12-15 2020-08-18 Duke University Methods and systems for generating fluid simulation models
CN106651846B (en) * 2016-12-20 2019-07-12 中南大学湘雅医院 Segmentation method of retinal blood vessel image
CN107527028B (en) * 2017-08-18 2020-03-24 深圳乐普智能医疗器械有限公司 Target cell identification method and device and terminal

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101908891A (en) * 2010-08-23 2010-12-08 南京信息工程大学 Medical image ROI (Region of Interest) compression method based on lifting wavelet and PCNN (Pulse Coupled Neural Network)
CN102306289A (en) * 2011-09-16 2012-01-04 兰州大学 Method for extracting iris features based on pulse couple neural network (PCNN)
CN105989607A (en) * 2016-04-13 2016-10-05 西华大学 Catenary image segmentation method based on SPCNN and minimum cross entropy

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
An image segmentation method of a modified SPCNN based on human visual system in medical images;Jing Lian，Zhen Yang， Wenhao Sun，Yanan Guo，Li Zheng;《Neurocomputing》;20181229;第292-306页 *
Automatic gallbladder and gallstone regions segmentation in ultrasound image;Jing Lian，Yide Ma1，Yurun Ma，Bin Shi，Jizhao Liu，Zhen Yang;《Int J CARS》;20170106;第553-568页 *
PCNN 参数自适应设定及其模型的改进;邓翔宇，马义德;《电子学报》;20120531;第955-964页 *
改进型脉冲耦合神经网络检测乳腺肿瘤超声图像感兴趣区域;汪源源，焦静;《光学精密工程》;20110630;第1398-1404页 *

Also Published As

Publication number Publication date
CN109685814A (en) 2019-04-26

Similar Documents

Publication Publication Date Title
CN109685814B (en) Full-automatic cholecystolithiasis ultrasonic image segmentation method based on MSPCNN
CN112700390B (en) Cataract OCT image repairing method and system based on machine learning
Karim et al. DeepKneeExplainer: explainable knee osteoarthritis diagnosis from radiographs and magnetic resonance imaging
Saranyaraj et al. A deep convolutional neural network for the early detection of breast carcinoma with respect to hyper-parameter tuning
Iranpoor et al. Skin lesion segmentation using convolutional neural networks with improved U-Net architecture
Panigrahy et al. Parameter adaptive unit-linking pulse coupled neural network based MRI–PET/SPECT image fusion
CN113538260B (en) LDCT image denoising and classifying method for self-supervision and supervised combined training
Shobana et al. Brain tumor diagnosis from MRI feature analysis-A comparative study
Senthilkumaran et al. Brain image segmentation
Adegoke et al. Review of feature selection methods in medical image processing
Wu et al. DA-U-Net: densely connected convolutional networks and decoder with attention gate for retinal vessel segmentation
Fu et al. MDRANet: A multiscale dense residual attention network for magnetic resonance and nuclear medicine image fusion
Subasree et al. Selection of optimal thresholds in multi-level thresholding using multi-objective emperor penguin optimization for precise segmentation of mammogram images
CN114332910A (en) Human body part segmentation method for similar feature calculation of far infrared image
Huang et al. Breast cancer diagnosis based on hybrid SqueezeNet and improved chef-based optimizer
Ogiela et al. Computer analysis of gallbladder ultrasonic images towards recognition of pathological lesions
Guo et al. Semi-supervised medical image classification based on CamMix
Wang et al. Medical image fusion based on modified pulse coupled neural network model and kirsch operator
CN114049357A (en) Breast ultrasonic segmentation method based on feature set association degree
Wu et al. Urine red blood cells generation using StyleGAN2 network
P. S et al. Microaneurysms detection using fundus images based on deep convolutional neural network enabled fractional hunter osprey optimization
Zhao et al. Intracranial vascular structure extraction: a machine learning approach
Wibisono et al. Segmentation-based knowledge extraction from chest X-ray images
Parvathy et al. Multimodality medical image fusion based on non-sub-sampled contourlet transform
Thepade et al. Haar wavelet pyramid-based melanoma skin cancer identification with ensemble of machine learning algorithms

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