WO2021129324A1 - 超声图像病灶的分割方法、装置和计算机设备 - Google Patents

超声图像病灶的分割方法、装置和计算机设备 Download PDF

Info

Publication number
WO2021129324A1
WO2021129324A1 PCT/CN2020/133028 CN2020133028W WO2021129324A1 WO 2021129324 A1 WO2021129324 A1 WO 2021129324A1 CN 2020133028 W CN2020133028 W CN 2020133028W WO 2021129324 A1 WO2021129324 A1 WO 2021129324A1
Authority
WO
WIPO (PCT)
Prior art keywords
level set
target
lesion
zero level
area
Prior art date
Application number
PCT/CN2020/133028
Other languages
English (en)
French (fr)
Inventor
李焰驹
Original Assignee
飞依诺科技(苏州)有限公司
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 飞依诺科技(苏州)有限公司 filed Critical 飞依诺科技(苏州)有限公司
Publication of WO2021129324A1 publication Critical patent/WO2021129324A1/zh

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
    • 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
    • G06T2207/30068—Mammography; Breast
    • 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

Definitions

  • This application relates to the technical field of medical image processing, and in particular to a method, device and computer equipment for segmenting ultrasound image lesions.
  • Breast cancer is a common malignant tumor in female diseases and has become one of the diseases that seriously threaten women's health.
  • Early detection, early diagnosis, and early treatment are the basic principles currently adopted in medicine for the prevention and treatment of breast cancer.
  • Ultrasound imaging has become one of the main methods of clinical diagnosis of breast tumors due to its advantages of non-invasiveness, non-radiation, and low cost.
  • ultrasound images often have large noise, low contrast, uneven grayscale, varying degrees of attenuation and infiltration effects, etc., making the surface of breast tumors more similar to the surrounding normal tissues, that is, ultrasound
  • the image has a weak ability to express the morphology of human organs, or the presentation of organs in the image is blurry and abstract; in addition, breast tumors vary greatly among individuals. Therefore, the judgment and reading of the lesion area in breast ultrasound images requires clinicians to have a higher professional level and rich experience. It is difficult for general doctors to accurately and quickly compare the breast tumor area with the normal surroundings in the ultrasound image. Distinguish the tissues and describe the results of the lesion.
  • an embodiment of the present application provides a method for segmenting an ultrasound image lesion, and the method includes:
  • an initial zero level set function where the initial zero level set function represents the initial contour of the lesion area
  • the energy functional is defined by the local binary fitting evolution algorithm based on the initial zero level set function
  • the corresponding target zero level set binary image is obtained, and the target zero level set binary image is post-processed to obtain the target contour of the lesion area.
  • constructing the initial zero level set function according to the region of interest includes: determining the coordinates of each vertex of the region of interest as a reference point, and shifting the reference point according to the set shift amount to obtain the relative The new vertex coordinates of each reference point; the initial contour of the lesion area is obtained according to the new vertex coordinates; the initial zero level set function is obtained based on the area of interest and the initial contour of the lesion area.
  • the minimum value of the energy functional is solved by the gradient descent method to obtain the target level set function at the end of the evolution, including: minimizing the energy functional by the gradient descent method to obtain the level set active contour evolution equation ; Based on the level set active contour evolution equation, iteratively calculates with the set step length and the number of iterations to obtain the target level set function at the end of evolution.
  • the method for setting the step size and the number of iterations includes: estimating the echo type of the lesion based on the histogram distribution of the region of interest; and determining the corresponding step size and the number of iterations based on the echo type.
  • the echo type includes anechoic and hypoechoic
  • the corresponding step size and number of iterations are determined based on the echo type, including: if the echo type is anechoic, the corresponding number of iterations is 80-260, and the time The step size is 0.1 to 1; if the echo type is low echo, the corresponding iteration number is 280 to 320, and the time step is 0.1 to 1.
  • post-processing the binary image of the target zero level set to obtain the target contour of the lesion area includes: performing inverse color processing on the binary image of the target zero level set to obtain multiple foregrounds to be filtered Region; fill in the multiple foreground regions to be screened to obtain the new binary image of the zero level set of the target after filling; extract each connected component from the binary image of the new target zero level set according to the neighborhood connectivity criterion, and Calculate the area of the area where each connected component is located; determine the area where the connected component with the largest area in the area where each connected component is located is the lesion area, and then the boundary pixels of the lesion area are the corresponding target contours.
  • the method further includes: processing the lesion area based on a morphological operation, and determining the boundary pixels of the processed lesion area as the corresponding target contour.
  • an ultrasound image lesion segmentation device the device includes:
  • the region of interest recognition module is used to identify the lesion in the ultrasound image to obtain the corresponding region of interest;
  • the initial zero level set function construction module is used to construct the initial zero level set function according to the region of interest, where the initial zero level set function represents the initial contour of the lesion area;
  • the energy functional definition module is used to define the energy functional based on the initial zero level set function using the local binary fitting evolution algorithm
  • the evolution module is used to solve the minimum value of the energy functional through the gradient descent method to obtain the target level set function at the end of the evolution;
  • the lesion segmentation result determination module is used to obtain the corresponding target zero level set binary image based on the target level set function, and perform post-processing on the target zero level set binary image to obtain the target contour of the lesion area.
  • an embodiment of the present application also provides a computer device, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method when the computer program is executed.
  • the embodiments of the present application also provide a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described above are implemented.
  • the above-mentioned ultrasound image lesion segmentation method, device, computer equipment and storage medium identify the lesion in the ultrasound image to locate the corresponding region of interest, and then construct the corresponding region of interest based on the region of interest and based on the local binary fitting evolutionary algorithm.
  • Initial zero level set function and energy functional, the minimum value of the energy functional is solved by the gradient descent method to obtain the target level set function at the end of the evolution, and the corresponding target zero level set binary image is obtained based on the target level set function.
  • the binary image of the target zero level set is post-processed to obtain the target contour of the lesion area, thereby facilitating the subsequent diagnosis work, which can not only effectively reduce the workload of the doctor, but also improve the diagnosis efficiency.
  • FIG. 1 is an application environment diagram of a method for segmentation of an ultrasound image lesion in an embodiment
  • FIG. 2 is a schematic flowchart of a method for segmenting an ultrasound image lesion in an embodiment
  • Figure 3 is a schematic diagram of an ultrasound image of the original breast
  • Fig. 4 is a schematic diagram of a region of interest obtained by performing target detection on Fig. 3;
  • FIG. 5 is a schematic flowchart of the steps of constructing an initial zero level set function in an embodiment
  • Fig. 6 is a schematic diagram of a binarized image corresponding to an initial zero level set function
  • FIG. 7 is a schematic diagram of pixel point x and its neighboring pixel y
  • FIG. 8 is a schematic flowchart of the step of solving the target level set function in an embodiment
  • FIG. 9 is a schematic diagram of a binary image of the target zero level set corresponding to the target level set function
  • Fig. 10A is a schematic diagram of a low-echo grayscale histogram
  • Fig. 10B is a schematic diagram of an echoless grayscale histogram
  • FIG. 11 is a schematic flowchart of post-processing steps for a binary image of a target zero level set in an embodiment
  • Fig. 12 is a schematic diagram of a new target zero level set binary image obtained after processing Fig. 9;
  • Figure 13 is a schematic diagram of the lesion area determined after analyzing Figure 12;
  • Figure 14 is a schematic diagram obtained after performing morphological operations on Figure 13;
  • Figure 15 is a schematic diagram showing the outline of the lesion area on the original image
  • Fig. 16 is a structural block diagram of an ultrasound image lesion segmentation device in an embodiment
  • Fig. 17 is a diagram of the internal structure of a computer device in an embodiment.
  • the method for segmenting an ultrasound image lesion provided in this application can be applied to the application environment as shown in FIG. 1.
  • the terminal 102 and the server 104 communicate through a network.
  • the terminal 102 may be a device with an ultrasound image collection function, or a device that stores the collected ultrasound images
  • the server 104 can be an independent server. Or it can be realized by a server cluster composed of multiple servers.
  • the terminal 102 is used to collect or store ultrasound images, and send the collected or stored ultrasound images to the server 104 through the network, and the server 104 recognizes the lesions in the ultrasound images to locate the corresponding region of interest, and then according to Region of interest, and based on the Local Binary Fitting (LBF) evolution algorithm to construct the initial zero level set function and the energy functional, the minimum value of the energy functional is solved by the gradient descent method to obtain the evolution termination time
  • LBF Local Binary Fitting
  • the target level set function based on the target level set function to obtain the corresponding target zero level set binary image, post-process the target zero level set binary image to obtain the target contour of the lesion area, so as to facilitate the subsequent diagnosis work, not only It can effectively reduce the workload of doctors and improve diagnosis efficiency.
  • a method for segmenting ultrasound image lesions is provided. Taking the method applied to the server in FIG. 1 as an example for description, the method includes the following steps:
  • Step 202 Identify the lesion in the ultrasound image to obtain a corresponding region of interest.
  • ultrasound images are clinically ultrasound images of breast nodules, breast tumors, thyroid nodules, thyroid tumors, liver cysts, kidney cysts, spleen cysts, etc.
  • the lesions refer to breast nodules, breast tumors, thyroid nodules, Thyroid tumors, liver cysts, kidney cysts, spleen cysts and other diseased parts.
  • a region of interest region of interest, ROI for short
  • the lesion in the ultrasound image can be automatically identified and located through target detection, so that the corresponding region of interest can be obtained in the ultrasound image.
  • the corresponding ultrasound image may be a breast ultrasound image (as shown in FIG. 3), so that the corresponding region of interest is obtained in the breast ultrasound image (as shown in FIG. 4).
  • Step 204 Construct an initial zero level set function according to the region of interest.
  • a digital image can be understood as a binary function
  • the initial contour of the lesion area is obtained, and then the corresponding initial contour is determined by the region of interest and the initial contour of the lesion area.
  • the zero level set function is used to represent the initial contour of the lesion area, which is also the initial object of the evolution algorithm.
  • the main idea of the level set method is to embed the curve as a zero level set on a higher one-dimensional surface, and obtain the evolution equation of the function through the evolution equation of the surface.
  • Step 206 Use a local binary fitting evolution algorithm to define an energy functional based on the initial zero level set function.
  • the functional is a special function, that is, the function of the function.
  • the independent variables of functionals include ordinary variables, such as x and y, as well as functions.
  • the independent variable of the energy functional in this embodiment is the initial zero level set function constructed in the above steps, which can also be referred to as an energy equation based on local binary fitting.
  • Step 208 Solve the minimum value of the energy functional by the gradient descent method to obtain the target level set function at the end of the evolution.
  • the local binary fitting (hereinafter referred to as LBF) evolution algorithm is used to solve the minimum value of the energy functional, so that the initial zero level set evolves toward the target boundary, thereby obtaining the target level at the end of the evolution.
  • Set function is used to express a closed two-dimensional curve in a three-dimensional space.
  • Step 210 Obtain a corresponding target zero level set binary image based on the target level set function, and perform post-processing on the target zero level set binary image to obtain a target contour of the lesion area.
  • the target level set function represents the target contour of the lesion area.
  • Post-processing includes morphological operations and filtering of the image. Since a digital image can be understood as a binary function, correspondingly, a binary function can also be expressed as a digital image.
  • the corresponding target zero level set binary image can be obtained, and then the target zero level set binary image is subjected to morphological operations and filtering, so as to filter out the precise lesion area.
  • Target profile is based on the target level set function.
  • the above-mentioned ultrasound image lesion segmentation method recognizes the lesion in the ultrasound image to locate the corresponding region of interest, and then constructs the initial zero level set function and energy functional based on the LBF evolution algorithm based on the region of interest, and passes the gradient
  • the descent method solves the minimum value of the energy functional to obtain the target level set function at the end of the evolution.
  • the corresponding target zero level set binary image is obtained, and the target zero level set binary image is post-processed to Obtain a precise target contour of the lesion area, thereby facilitating the subsequent diagnosis work, which not only can effectively reduce the workload of the doctor, but also can improve the diagnosis efficiency.
  • constructing an initial zero level set function according to the region of interest may specifically include the following steps:
  • Step 502 Determine the coordinates of each vertex of the region of interest as a reference point, and translate the reference point according to the set translation amount to obtain a new vertex coordinate relative to each reference point.
  • the set translation amount can be any constant, and the constant should be less than the minimum side length of the region of interest.
  • the coordinates of each vertex of the region of interest are taken as the reference points, and the X and Y coordinates of each reference point are respectively translated to the inside of the region of interest according to the set translation amount, so as to obtain the reference point relative to each reference point.
  • the new vertex coordinates are the vertices of the rectangular area 40 in FIG. 4.
  • Step 504 Obtain the initial contour of the lesion area according to the new vertex coordinates.
  • the coordinate points are sequentially connected to obtain the corresponding closed curve (such as the rectangular area 40 in FIG. 4).
  • the area enclosed by the closed curve is the initial area of the breast tumor lesion.
  • the curve is the initial contour of the lesion area.
  • Step 506 Obtain an initial zero level set function based on the initial contours of the region of interest and the lesion area.
  • an initial image with the same size as the ROI is created according to the initial contour of the region of interest (ROI) and the lesion area, where the black area shown in Figure 6 represents the interior of the initial contour , which is the initial area of the breast tumor lesion, and its pixel value is set to -2; the white area represents the outside of the initial contour, and its pixel value is set to 2.
  • ROI region of interest
  • the initial image is a binarized image
  • the binarized image is the initial zero level
  • the set of images is also the initial object of the evolutionary algorithm. Also, because digital images can be represented by a binary function, the binary image shown in Figure 6 can be represented by the following function (ie, the initial zero level set function):
  • r is the row coordinate of any pixel in the image
  • c is the column coordinate
  • R 0 represents the initial area of the lesion.
  • the energy functional is defined based on the above-mentioned initial zero level set function. Specifically, assuming that x is any pixel in the original image, y is any pixel adjacent to pixel x (called the neighbor of x). Domain pixels), as shown in Figure 7, where x and y are two-dimensional vectors, which can be expressed as x(c,r),y(c,r). Then define the energy functional as:
  • the first term on the right side of the equation is the subject term of the energy functional
  • the P in the second term is the penalty term
  • the L in the third term is the length of the zero-level curve of the level set function
  • ⁇ , ⁇ are Normal number.
  • the minimum value of the energy functional is solved by the gradient descent method to obtain the target level set function at the end of the evolution, including the following steps:
  • step 802 the energy functional is minimized by the gradient descent method, and the level set active contour evolution equation is obtained.
  • H in equation (3) is the Heaviside function.
  • H ⁇ is the Heaviside function.
  • H ⁇ is the Gaussian kernel function with standard deviation ⁇ (this function is based on the region of interest Obtained after Gaussian processing).
  • I(y) represents the pixel gray value of the neighboring pixel y of any pixel x in the binarized image shown in Figure 6, ⁇ 1 and ⁇ 2 are positive constants, which are the weights of the corresponding integral terms.
  • ⁇ 1 1 (that is, it is always 1), and ⁇ 2 can be determined according to the echo type of the region of interest.
  • f 1 (x) and f 2 (x) are always greater than zero, where:
  • step 804 iterative calculation based on the level set active contour evolution equation using the set step size and iteration number to obtain the target level set function at the end of the evolution.
  • equation (12) is approximately transformed into a discrete finite difference form:
  • the setting of the step size and the number of iterations includes: estimating the echo type of the lesion according to the histogram distribution of the region of interest, ie, FIG. 4, and then determining the corresponding step size and the number of iterations according to the echo type.
  • the echo types include low echo and anechoic, as shown in FIG. 10A and FIG. 10B, which represent the gray histograms of low echo and anechoic, respectively.
  • the horizontal axis represents a total of 256 intervals from 0 to 255 (that is, the possible gray values), and the vertical axis represents the frequency of each gray value.
  • the frequency of the most frequent gray values in the echo-free histogram is very different from the average frequency of other gray values. Therefore, the distribution feature of the histogram can be used to distinguish the echo type of the lesion.
  • the following formula can be used to quantitatively express:
  • ratio maxFrequency/mean_num, where maxFrequency is the frequency of the gray value that appears most frequently in the histogram, and mean_num is the average frequency of other gray values.
  • maxFrequency is the frequency of the gray value that appears most frequently in the histogram
  • mean_num is the average frequency of other gray values. The larger the ratio, the more likely it is an anechoic lesion. Specifically, when the ratio is greater than 7, it can be determined as an anechoic lesion, and when the ratio is less than 7, it can be determined as a hypoechoic lesion.
  • the corresponding number of iterations is 80-260
  • the time step is 0.1-1.0
  • the value of ⁇ 2 in the corresponding formula (3) is 2.0 ⁇ 3.3
  • the value of V in the corresponding formula (1) is 0.003*255*255 ⁇ 0.008*255*255
  • the corresponding iteration number is 280 ⁇ 320
  • the time step is 0.1 ⁇ 1.0
  • the value of ⁇ 2 in the corresponding formula (3) is 1.5 to 2.2
  • the value of V in the corresponding formula (1) is 10 to 8.
  • the calculation is performed based on the echo type and the corresponding parameters are substituted to obtain the target level set function at the end of the evolution, which corresponds to the target zero level set binary image as shown in FIG. 9.
  • post-processing the binary image of the target zero level set to obtain the target contour of the lesion area may specifically include the following steps:
  • Step 1102 Perform inverse color processing on the binary image of the target zero level set to obtain multiple foreground regions to be screened.
  • the binary image of the target zero level set as shown in FIG. 9 is subjected to inverse color processing to obtain a plurality of foreground regions to be screened after the inverse color processing, wherein the pixel gray level of the foreground area after the inverted color processing is The value is 255 (that is, white), and the gray value of the background pixel is 0 (that is, black).
  • Step 1104 Filling holes in the multiple foreground regions to be screened to obtain a new binary image of the target zero level set after filling.
  • hole filling is performed on the multiple foreground regions to be screened, so as to obtain a new binary image of the target zero level set after filling, as shown in FIG. 12.
  • Step 1106 Extract each connected component from the binary image of the new target zero level set according to the neighborhood connectivity criterion, and calculate the area of the area where each connected component is located.
  • each neighborhood in Figure 12 there are 4 connected components in Figure 12.
  • the white area is not connected with other white areas, that is, 4 independent white areas. Then calculate the area of each connected component area (that is, the number of pixels occupied by the connected component area).
  • Step 1108 Determine the area of the connected component with the largest area among the areas where the connected components are located as the lesion area.
  • the area where the connected component with the largest area is located in the area where each connected component is located is determined as the lesion area.
  • the boundary of the lesion area is The pixels are the corresponding target contours, that is, the boundary pixels between the foreground and the background in FIG. 13 are the target contours of the lesion area.
  • the lesion area in order to make the target contour of the lesion area more accurate and smooth, after the lesion area is determined, that is, after Figure 13 is obtained, the lesion area can also be processed based on morphological operations, that is, Figure 13 is processed In order to obtain the processed FIG. 14, the boundary pixels between the foreground and the background in the processed FIG. 14 are the target contours of the lesion area, so as to achieve the goal of more accurate and smooth target contours of the lesion area. Further, based on the target contour, the contour of the breast tumor focus area (as shown in Fig. 15) can be displayed on the original region of interest (that is, Fig. 4).
  • a device for segmenting an ultrasound image lesion including: a region of interest recognition module 1601, an initial zero level set function construction module 1602, an energy functional definition module 1603, and an evolution module 1604 and lesion segmentation result determination module 1605, where:
  • the region of interest identification module 1601 is used to identify the lesion in the ultrasound image to obtain the corresponding region of interest;
  • the initial zero level set function construction module 1602 is used to construct an initial zero level set function according to the region of interest, where the initial zero level set function represents the initial contour of the lesion area;
  • the energy functional definition module 1603 is used to define the energy functional based on the initial zero level set function using the local binary fitting evolutionary algorithm
  • the evolution module 1604 is used to solve the minimum value of the energy functional through the gradient descent method to obtain the target level set function at the end of the evolution;
  • the lesion segmentation result determination module 1605 is configured to obtain the corresponding target zero level set binary image based on the target flat set function, and perform post-processing on the target zero level set binary image to obtain the target contour of the lesion area.
  • the initial zero-level set function construction module 1602 is specifically configured to: determine the coordinates of each vertex of the region of interest as a reference point, and translate the reference point according to the set translation amount to obtain a reference point relative to each reference point. According to the new vertex coordinates; get the initial contour of the lesion area according to the new vertex coordinates; get the initial zero level set function based on the area of interest and the initial contour of the lesion area.
  • the evolution module 1604 is specifically used for: minimizing the energy functional through the gradient descent method to obtain the level set active contour evolution equation; based on the level set active contour evolution equation using the set step size and number of iterations to iteratively calculate , In order to obtain the target level set function at the end of the evolution.
  • the echo type includes anechoic and hypoechoic. If the echo type is anechoic, the corresponding iteration number is 80-260, and the time step is 0.1-1; if the echo type is hypoechoic, Then the corresponding number of iterations is 280-320, and the time step is 0.1-1.
  • the lesion segmentation result determination module 1605 is specifically configured to: perform inverse color processing on the target zero-level set binary image to obtain multiple foreground regions to be screened; fill the multiple foreground regions to be screened, In order to obtain the new binary image of the target zero level set after filling; extract the connected components from the new binary image of the target zero level set according to the neighborhood connectivity criterion, and calculate the area of the area where each connected component is located; combine the connected components The area where the connected component with the largest area in the area is determined to be the focus area, and the boundary pixels of the focus area are the corresponding target contours.
  • the lesion segmentation result determination module 1605 is further configured to process the lesion area based on morphological operations, and determine the boundary pixels of the processed lesion area as the corresponding target contour.
  • Each module in the above-mentioned ultrasound image lesion segmentation device can be implemented in whole or in part by software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the form of hardware or independent of the processor in the computer equipment, or may be stored in the memory of the computer equipment in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.
  • a computer device is provided.
  • the computer device may be a server, and its internal structure diagram may be as shown in FIG. 17.
  • the computer equipment includes a processor, a memory, a network interface, and a database connected through a system bus.
  • the processor of the computer device is used to provide calculation and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system, a computer program, and a database.
  • the internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer equipment is used to store ultrasound image data.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • FIG. 17 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
  • the specific computer device may Including more or fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.
  • a computer device including a memory and a processor, and a computer program is stored in the memory, and the processor implements the following steps when the processor executes the computer program:
  • an initial zero level set function where the initial zero level set function represents the initial contour of the lesion area
  • the energy functional is defined by the local binary fitting evolution algorithm based on the initial zero level set function
  • the corresponding target zero level set binary image is obtained, and the target zero level set binary image is post-processed to obtain the target contour of the lesion area.
  • the processor further implements the following steps when executing the computer program: determining the coordinates of each vertex of the region of interest as the reference point, and shifting the reference point according to the set translation amount to obtain the reference point relative to each reference point.
  • the new vertex coordinates; the initial contour of the lesion area is obtained according to the new vertex coordinates; the initial zero level set function is obtained based on the area of interest and the initial contour of the lesion area.
  • the processor also implements the following steps when executing the computer program: the energy functional is minimized by the gradient descent method to obtain the level set active contour evolution equation; the level set active contour evolution equation is based on the set step size and The number of iterations is calculated iteratively to obtain the target level set function at the end of the evolution.
  • the processor further implements the following steps when executing the computer program: according to the histogram distribution of the region of interest, the echo type of the lesion is estimated; and the corresponding step size and the number of iterations are determined based on the echo type.
  • the echo types include anechoic and hypoechoic
  • the processor further implements the following steps when executing the computer program: if the echo type is anechoic, the corresponding iteration number is 80-260, and the time step is 0.1 ⁇ 1; if the echo type is low echo, the corresponding iteration number is 280-320, and the time step is 0.1-1.
  • the processor further implements the following steps when executing the computer program: performing inverse color processing on the target zero level set binary image to obtain multiple foreground regions to be screened; performing hole filling on the multiple foreground regions to be screened , To obtain the new binary image of the target zero level set after filling; extract each connected component from the new binary image of the target zero level set according to the neighborhood connectivity criterion, and calculate the area of the area where each connected component is located; The area where the connected component with the largest area in the area of the component is determined to be the focus area, and the boundary pixels of the focus area are the corresponding target contours.
  • the processor further implements the following steps when executing the computer program: after determining the lesion area, processing the lesion area based on morphological operations, and determining that the boundary pixels of the processed lesion area are the corresponding target contours.
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
  • an initial zero level set function where the initial zero level set function represents the initial contour of the lesion area
  • the energy functional is defined by the local binary fitting evolution algorithm based on the initial zero level set function
  • the corresponding target zero level set binary image is obtained, and the target zero level set binary image is post-processed to obtain the target contour of the lesion area.
  • the following steps are further implemented: the coordinates of each vertex of the region of interest are determined as the reference point, and the reference point is translated according to the set translation amount to obtain a reference point relative to each reference point. According to the new vertex coordinates; get the initial contour of the lesion area according to the new vertex coordinates; get the initial zero level set function based on the area of interest and the initial contour of the lesion area.
  • the following steps are also implemented: the energy functional is minimized by the gradient descent method to obtain the level set active contour evolution equation; the level set active contour evolution equation is based on the set step size And the number of iterations are calculated iteratively to obtain the target level set function at the end of the evolution.
  • the following steps are further implemented: according to the histogram distribution of the region of interest, the echo type of the lesion is estimated; and the corresponding step size and the number of iterations are determined based on the echo type.
  • the echo types include anechoic and hypoechoic, and when the computer program is executed by the processor, the following steps are also implemented: if the echo type is anechoic, the corresponding number of iterations is 80-260, and the time step is 0.1 ⁇ 1; if the echo type is low echo, the corresponding iteration number is 280 ⁇ 320, and the time step is 0.1 ⁇ 1.
  • the following steps are also implemented: inverting the binary image of the target zero level set to obtain multiple foreground regions to be screened; performing holes on the multiple foreground regions to be screened Fill to obtain a new binary image of the target zero level set after filling; extract each connected component from the new binary image of the target zero level set according to the neighborhood connectivity criterion, and calculate the area of each connected component; The area where the connected component with the largest area is located in the area where the connected component is located is determined as the focus area, and the boundary pixels of the focus area are the corresponding target contours.
  • the following steps are further implemented: after determining the lesion area, processing the lesion area based on morphological operations, and determining the boundary pixels of the processed lesion area as the corresponding target contour.
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

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)
  • Image Analysis (AREA)

Abstract

一种超声图像病灶的分割方法、装置和计算机设备。其中方法通过对超声图像中的病灶进行识别，以定位对应的感兴趣区域，进而根据感兴趣区域，并基于LBF演化算法构造初始零水平集函数及能量泛函，通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数，基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到精确的病灶区域的目标轮廓，从而便于后续诊断工作的进行，不仅能够有效地减少医生的工作量，且能提高诊断效率。

Description

超声图像病灶的分割方法、装置和计算机设备 技术领域
本申请涉及医学图像处理技术领域，特别是涉及一种超声图像病灶的分割方法、装置和计算机设备。
背景技术
乳腺癌是女性疾病中常见的恶性肿瘤，已成为严重威胁女性健康的病症之一。早发现、早诊断、早治疗是目前医学上对防治乳腺癌采取的基本原则。超声成像凭借其无创伤、无辐射、费用低廉等优势，已成为乳腺肿瘤临床诊断的主要手段之一。
然而，由于受成像设备的影响，超声图像常常具有较大的噪声、低对比度、灰度不均匀、不同程度的衰减以及浸润效应等，使得乳腺肿瘤在表面上与周围正常组织较为相似，即超声图像对人体器官形态的表达能力较弱，或者说器官在图像中的呈现形式较为模糊与抽象；此外，不同个体间乳腺肿瘤差异也较大。因此，乳腺超声图像中对于病灶区域的判断和阅读需要临床医生具有较高的专业水平和较丰富的经验，一般医生较难在超声图像中用肉眼准确、快速地将乳腺肿瘤区域与其周围的正常组织区分开来并作出病灶描述结果。
发明内容
基于此，有必要针对上述一般医生较难在超声图像中快速准确的定位病灶区域的问题，提供一种超声图像病灶的分割方法、装置和计算机设备。
为了实现上述目的，一方面，本申请实施例提供了一种超声图像病灶的分 割方法，所述方法包括：
识别超声图像中的病灶，以得到对应的感兴趣区域；
根据感兴趣区域，构造初始零水平集函数，其中，初始零水平集函数表示病灶区域的初始轮廓；
基于初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
在其中一个实施例中，根据感兴趣区域，构造初始零水平集函数，包括：将感兴趣区域的各顶点坐标确定为参考点，根据设定的平移量对参考点进行平移，以得到相对于各参考点的新的顶点坐标；根据新的顶点坐标得到病灶区域的初始轮廓；基于感兴趣区域以及病灶区域的初始轮廓得到初始零水平集函数。
在其中一个实施例中，通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数，包括：通过梯度下降法使得能量泛函最小化，得到水平集活动轮廓演化方程；基于水平集活动轮廓演化方程采用设定的步长和迭代次数迭代计算，以得到演化终止时刻的目标水平集函数。
在其中一个实施例中，步长和迭代次数的设定方法包括：根据感兴趣区域的直方图分布，估计病灶的回声类型；基于回声类型确定对应的步长和迭代次数。
在其中一个实施例中，回声类型包括无回声和低回声，则基于回声类型确定对应的步长和迭代次数，包括：若回声类型为无回声时，则对应的迭代次数为80～260，时间步长为0.1～1；若回声类型为低回声时，则对应的迭代次数 为280～320，时间步长为0.1～1。
在其中一个实施例中，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓，包括：对目标零水平集二值图像进行反色处理，以得到待筛选的多个前景区域；对待筛选的多个前景区域进行孔洞填充，以得到填充后的新的目标零水平集二值图像；根据邻域连通准则在新的目标零水平集二值图像中提取各连通分量，并计算各连通分量所在区域的面积；将各连通分量所在区域的面积中面积最大的连通分量所在区域确定为病灶区域，则病灶区域的边界像素为对应的目标轮廓。
在其中一个实施例中，确定病灶区域之后，还包括：基于形态学运算对病灶区域进行处理，确定处理后的病灶区域的边界像素为对应的目标轮廓。
另一方面，本申请实施例还提供了一种超声图像病灶的分割装置，所述装置包括：
感兴趣区域识别模块，用于识别超声图像中的病灶，以得到对应的感兴趣区域；
初始零水平集函数构造模块，用于根据感兴趣区域，构造初始零水平集函数，其中，初始零水平集函数表示病灶区域的初始轮廓；
能量泛函定义模块，用于基于初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
演化模块，用于通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
病灶分割结果确定模块，用于基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
又一方面，本申请实施例还提供了一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现如上所述方法的步骤。
再一方面，本申请实施例还提供了一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现如上所述方法的步骤。
上述超声图像病灶的分割方法、装置、计算机设备和存储介质，通过对超声图像中的病灶进行识别，以定位对应的感兴趣区域，进而根据感兴趣区域，并基于局部二值拟合演化算法构造初始零水平集函数及能量泛函，通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数，基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓，从而便于后续诊断工作的进行，不仅能够有效地减少医生的工作量，且能提高诊断效率。
附图说明
图1为一个实施例中超声图像病灶的分割方法的应用环境图；
图2为一个实施例中超声图像病灶的分割方法的流程示意图；
图3为原乳腺超声图像示意图；
图4为对图3进行目标检测得到的感兴趣区域的示意图；
图5为一个实施例中构造初始零水平集函数步骤的流程示意图；
图6为初始零水平集函数对应的二值化图像的示意图；
图7为像素点x及其邻域像素y的示意图；
图8为一个实施例中求解目标水平集函数步骤的流程示意图；
图9为目标水平集函数对应的目标零水平集二值图像的示意图；
图10A为低回声灰度直方图的示意图；
图10B为无回声灰度直方图的示意图；
图11为一个实施例中对目标零水平集二值图像进行后处理步骤的流程示意图；
图12为对图9进行处理后得到的新的目标零水平集二值图像的示意图；
图13为对图12进行分析后确定的病灶区域的示意图；
图14为对图13进行形态学运算后得到的示意图；
图15为在原图上显示病灶区域的轮廓示意图；
图16为一个实施例中超声图像病灶的分割装置的结构框图；
图17为一个实施例中计算机设备的内部结构图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请提供的超声图像病灶的分割方法，可以应用于如图1所示的应用环境中。其中，终端102与服务器104通过网络进行通信，在本实施例中，终端102可以是具有超声图像采集功能的设备，也可以是对采集的超声图像进行存储的设备，服务器104可以用独立的服务器或者是多个服务器组成的服务器集群来实现。具体的，终端102用于采集或存储超声图像，并将采集或存储的超声图像通过网络发送至服务器104，服务器104则对超声图像中的病灶进行识别，以定位对应的感兴趣区域，进而根据感兴趣区域，并基于局部二值拟合(Local Binary Fitting，简称LBF)演化算法构造初始零水平集函数及能量泛函，通过 梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数，基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓，从而便于后续诊断工作的进行，不仅能够有效地减少医生的工作量，且能提高诊断效率。
在一个实施例中，如图2所示，提供了一种超声图像病灶的分割方法，以该方法应用于图1中的服务器为例进行说明，包括以下步骤：
步骤202，识别超声图像中的病灶，以得到对应的感兴趣区域。
其中，超声图像为临床上乳腺结节、乳腺肿瘤、甲状腺结节、甲状腺肿瘤、肝脏囊肿、肾脏囊肿、脾囊肿等病灶的超声图像，病灶则是指乳腺结节、乳腺肿瘤、甲状腺结节、甲状腺肿瘤、肝脏囊肿、肾脏囊肿、脾囊肿等病变部分。感兴趣区域(region of interest，简称ROI)是从超声图像中选择的一个需要处理的图像区域，这个区域是进行图像分析所关注的重点。具体的，可以通过目标检测对超声图像中的病灶进行自动识别、定位，从而在超声图像中得到对应的感兴趣区域。以下以病灶为乳腺肿瘤为例进行说明，则对应的超声图像可以为乳腺超声图像(如图3所示)，从而在乳腺超声图像中得到对应的感兴趣区域(如图4所示)。
步骤204，根据感兴趣区域，构造初始零水平集函数。
由于一幅数字图像可以理解为一个二元函数，在本实施例中，基于感兴趣区域的各顶点坐标，得到病灶区域的初始轮廓，进而由感兴趣区域以及病灶区域的初始轮廓确定对应的初始零水平集函数，在本实施例中，初始零水平集函数用于表示病灶区域的初始轮廓，其也是演化算法的初始对象。其中，水平集方法的主要思想是将曲线作为零水平集嵌入到更高一维的曲面上，通过曲面的演化方程得到函数的演化方程。
步骤206，基于初始零水平集函数利用局部二值拟合演化算法定义能量泛函。
其中，泛函是一种特殊的函数，即函数的函数。泛函的自变量既有普通的变量，如x、y，也有函数。而本实施例中的能量泛函(energy functional)的自变量是上述步骤构造的初始零水平集函数，其也可以称为基于局部二值拟合的能量方程。
步骤208，通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数。
在本实施例中，基于感兴趣区域，利用局部二值拟合(以下简称LBF)演化算法求解能量泛函的最小值，使得初始零水平集朝目标边界演化，从而得到演化终止时刻的目标水平集函数。其中，零水平集就是将一个封闭的二维曲线表达于三维的空间里。
步骤210，基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
其中，目标水平集函数表示病灶区域的目标轮廓。后处理则包括对图像的形态学运算以及筛选等处理。由于一幅数字图像可以理解为一个二元函数，则相应的，一个二元函数也可以表示为一幅数字图像。在本实施例中，基于目标水平集函数可以得到对应的目标零水平集二值图像，进而对该目标零水平集二值图像进行形态学运算以及筛选等处理，从而筛选出精确的病灶区域的目标轮廓。
上述超声图像病灶的分割方法，通过对超声图像中的病灶进行识别，以定位对应的感兴趣区域，进而根据感兴趣区域，并基于LBF演化算法构造初始零水平集函数及能量泛函，通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数，基于目标水平集函数得到对应的目标零水平集二 值图像，对目标零水平集二值图像进行后处理，以得到精确的病灶区域的目标轮廓，从而便于后续诊断工作的进行，不仅能够有效地减少医生的工作量，且能提高诊断效率。
在一个实施例中，如图5所示，根据感兴趣区域，构造初始零水平集函数，具体可以包括如下步骤：
步骤502，将感兴趣区域的各顶点坐标确定为参考点，根据设定的平移量对参考点进行平移，以得到相对于各参考点的新的顶点坐标。
其中，设定的平移量可以是任意的常数，该常数应小于感兴趣区域的最小边长。在本实施例中，以感兴趣区域的各顶点坐标为参考点，按设定的平移量将各参考点的X、Y坐标分别向感兴趣区域的内部平移，从而得到相对于各参考点的新的顶点坐标，如图4中矩形区域40的各顶点。
步骤504，根据新的顶点坐标得到病灶区域的初始轮廓。
具体的，根据新的顶点坐标将各坐标点依次连接，从而得到对应的闭合曲线(如图4中的矩形区域40)，则该闭合曲线圈定的区域即为乳腺肿瘤病灶的初始区域，该闭合曲线则为病灶区域的初始轮廓。
步骤506，基于感兴趣区域以及病灶区域的初始轮廓得到初始零水平集函数。
具体的，根据感兴趣区域(ROI)以及病灶区域的初始轮廓创建一幅与ROI尺寸相同的初始图像(如图6所示)，其中，图6所示的黑色区域表示的是初始轮廓的内部，即为乳腺肿瘤病灶的初始区域，设定其像素值均为-2；白色区域表示的是初始轮廓的外部，设定其像素值均为2。
由于上述初始图像中水平集初始区域的像素灰度值为-2，而其他区域的像素灰度值为2，因此，该初始图像为二值化图像，该二值化图像即为初始零水平集图像，也是演化算法的初始对象。又由于数字图像可以通过二元函数表示， 因此，对于图6所示的二值化图像可通过如下函数(即初始零水平集函数)表示：
Figure PCTCN2020133028-appb-000001
式中r为图像中任意一个像素点的行坐标，c为列坐标，R 0表示病灶的初始区域。
进一步的，基于上述初始零水平集函数定义能量泛函，具体的，假设x是原图像中的任意一个像素点，y则是与像素点x相邻的任意一个像素点(称为x的邻域像素)，如图7所示，其中，x和y均为二维向量，可表示为x(c,r),y(c,r)。则定义能量泛函为：
F(φ,f 1,f 2)＝E LBF(φ,f 1,f 2)+μΡ(φ)+υL(φ)     (2)
其中，等式右侧的第一项是能量泛函的主体项，第二项中的P是惩罚项，第三项中的L是水平集函数的零水平曲线的长度，μ，υ则是正常数。
在一个实施例中，如图8所示，通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数，包括如下步骤：
步骤802，通过梯度下降法使得能量泛函最小化，得到水平集活动轮廓演化方程。
具体的，上式(2)中：
Figure PCTCN2020133028-appb-000002
Figure PCTCN2020133028-appb-000003
Figure PCTCN2020133028-appb-000004
其中，式(3)中的H是Heaviside函数，本文中用正则化的Heaviside函数 H ε，近似Heaviside函数，K σ(x)是标准差为σ的高斯核函数(该函数基于对感兴趣区域进行高斯处理后获得)。I(y)表示如图6所示的二值化图像中任意像素点x的邻域像素y的像素灰度值，λ 1、λ 2是正的常数，为对应积分项的权重，在本实施例中，λ 1＝1(即恒为1)，λ 2则可以根据感兴趣区域的回声类型确定。式(4)中
Figure PCTCN2020133028-appb-000005
是对零水平集函数φ(c，r)求梯度。式(5)中的Dirac函数δ是Heaviside函数的一阶导数。正则化的Dirac函数表示为δ ε。则有：
Figure PCTCN2020133028-appb-000006
Figure PCTCN2020133028-appb-000007
Figure PCTCN2020133028-appb-000008
将方程式(2)右侧的第一项和第三项正则化，则可以近似表达为：
F ε(φ,f 1,f 2)＝E ε LBF(φ,f 1,f 2)+μΡ(φ)+υL ε(φ)       (9)
式(9)中f 1(x)和f 2(x)恒大于零，其中：
Figure PCTCN2020133028-appb-000009
Figure PCTCN2020133028-appb-000010
通过梯度下降法计算能量泛函的最小值，具体的，保持f 1和f 2固定不变，使用标准的梯度下降法将关于φ的能量泛函F ε(φ,f 1,f 2)最小化，从而得到水平集活动轮廓演化方程：
Figure PCTCN2020133028-appb-000011
式(12)中，
e 1(x)＝∫ ΩK σ(y-x)|I(x)-f 1(y)| 2dy        (13)
e 2(x)＝∫ ΩK σ(y-x)|I(x)-f 2(y)| 2dy       (14)
步骤804，基于水平集活动轮廓演化方程采用设定的步长和迭代次数迭代计算，以得到演化终止时刻的目标水平集函数。
具体的，将式(12)中的偏微分方程近似转化为离散的有限差分形式：
Figure PCTCN2020133028-appb-000012
式中的
Figure PCTCN2020133028-appb-000013
就是式(12)中的等式的右侧表达式，采用设定的步长Δt和迭代次数k对式(15)进行迭代计算，以得到演化终止时刻的目标水平集函数，即得到当能量泛函F ε(φ,f 1,f 2)取得最小值时对应的水平集函数φ的零水平集轮廓即为最终结果，如图9所示。
在一个实施例中，上述步长和迭代次数的设定包括：根据感兴趣区域即图4的直方图分布，估计病灶的回声类型，进而根据回声类型确定对应的步长和迭代次数。通常，回声类型包括低回声和无回声，如图10A和图10B所示，分别表示低回声和无回声的灰度直方图。其横轴表示0到255共256个区间(即灰度可能的取值)，纵轴为每个灰度值出现的频数。很显然，无回声的直方图中出现频数最多的灰度值的频数与其它灰度值出现的平均频数相差很大。因此可用直方图的这种分布特征来区分病灶的回声类型。具体可以采用如下公式来定量表示：
ratio＝maxFrequency/mean_num，其中，maxFrequency是直方图中出现频数最多的灰度值的频数，mean_num是其它灰度值出现的平均频数。比值ratio越大，则越可能是无回声病灶。具体的，当比值大于7时，则可以确定为无回声病灶，当比值小于7时，则可以确定为低回声病灶。
在本实施例中，若通过上述方法确定回声类型为无回声时，则对应的迭代次数为80～260，时间步长为0.1～1.0，且对应式(3)中的λ 2取值为2.0～3.3， 对应式(1)中的V取值为0.003*255*255～0.008*255*255；若回声类型为低回声时，则对应的迭代次数为280～320，时间步长为0.1～1.0，且对应式(3)中的λ 2取值为1.5～2.2，对应式(1)中的V取值为10～8。基于回声类型代入对应的参数进行计算，从而得到演化终止时刻的目标水平集函数，该函数对应如图9所示的目标零水平集二值图像。
在一个实施例中，如图11所示，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓，具体可以包括如下步骤：
步骤1102，对目标零水平集二值图像进行反色处理，以得到待筛选的多个前景区域。
具体的，对如图9所示的目标零水平集二值图像进行反色处理，以得到反色处理后的多个待筛选的前景区域，其中，反色处理后的前景区域的像素灰度值为255(即白色)，背景像素灰度值为0(即黑色)。
步骤1104，对待筛选的多个前景区域进行孔洞填充，以得到填充后的新的目标零水平集二值图像。
基于上述反色处理后的多个待筛选的前景区域，对多个待筛选的前景区域进行孔洞填充，从而得到填充后的新的目标零水平集二值图像，如图12所示。
步骤1106，根据邻域连通准则在新的目标零水平集二值图像中提取各连通分量，并计算各连通分量所在区域的面积。
具体的，根据邻域连通准则采用8邻域在新的目标零水平集二值图像中提取各邻域的连通分量，此时图12中有4个连通分量，很显然，图12中的各白色区域与其它白色区域没有连通，即4个彼此独立的白色区域。进而计算各连通分量所在区域的面积(即该连通分量区域所占的像素数)。
步骤1108，将各连通分量所在区域的面积中面积最大的连通分量所在区域 确定为病灶区域。
在本实施例中，根据上述计算的各连通分量所在区域的面积，将各连通分量所在区域的面积中面积最大的连通分量所在区域确定为病灶区域，如图13所示，则病灶区域的边界像素为对应的目标轮廓，即图13中前景与背景的边界像素为病灶区域的目标轮廓。
在一个实施例中，为了使得病灶区域的目标轮廓更为精确、平滑，则在确定病灶区域之后，即得到图13后，还可以基于形态学运算对病灶区域进行处理，即对图13进行处理，从而得到处理后的图14，则处理后的图14中前景与背景的边界像素为病灶区域的目标轮廓，以达到病灶区域的目标轮廓更为精确、平滑的目的。进一步的，基于该目标轮廓可以在原感兴趣区域(即图4)上显示乳腺肿瘤病灶区域的轮廓(如图15所示)。
应该理解的是，虽然图1-15的流程图中的各个步骤按照箭头的指示依次显示，但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明，这些步骤的执行并没有严格的顺序限制，这些步骤可以以其它的顺序执行。而且，图1-15中的至少一部分步骤可以包括多个子步骤或者多个阶段，这些子步骤或者阶段并不必然是在同一时刻执行完成，而是可以在不同的时刻执行，这些子步骤或者阶段的执行顺序也不必然是依次进行，而是可以与其它步骤或者其它步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
在一个实施例中，如图16所示，提供了一种超声图像病灶的分割装置，包括：感兴趣区域识别模块1601、初始零水平集函数构造模块1602、能量泛函定义模块1603、演化模块1604和病灶分割结果确定模块1605，其中：
感兴趣区域识别模块1601，用于识别超声图像中的病灶，以得到对应的感兴趣区域；
初始零水平集函数构造模块1602，用于根据感兴趣区域，构造初始零水平集函数，其中，初始零水平集函数表示病灶区域的初始轮廓；
能量泛函定义模块1603，用于基于初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
演化模块1604，用于通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
病灶分割结果确定模块1605，用于基于目标平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
在一个实施例中，初始零水平集函数构造模块1602具体用于：将感兴趣区域的各顶点坐标确定为参考点，根据设定的平移量对参考点进行平移，以得到相对于各参考点的新的顶点坐标；根据新的顶点坐标得到病灶区域的初始轮廓；基于感兴趣区域以及病灶区域的初始轮廓得到初始零水平集函数。
在一个实施例中，演化模块1604具体用于：通过梯度下降法使得能量泛函最小化，得到水平集活动轮廓演化方程；基于水平集活动轮廓演化方程采用设定的步长和迭代次数迭代计算，以得到演化终止时刻的目标水平集函数。
在一个实施例中，回声类型包括无回声和低回声，则若回声类型为无回声时，则对应的迭代次数为80～260，时间步长为0.1～1；若回声类型为低回声时，则对应的迭代次数为280～320，时间步长为0.1～1。
在一个实施例中，病灶分割结果确定模块1605具体用于：对目标零水平集二值图像进行反色处理，以得到待筛选的多个前景区域；对待筛选的多个前景区域进行孔洞填充，以得到填充后的新的目标零水平集二值图像；根据邻域连通准则在新的目标零水平集二值图像中提取各连通分量，并计算各连通分量所 在区域的面积；将各连通分量所在区域的面积中面积最大的连通分量所在区域确定为病灶区域，则病灶区域的边界像素为对应的目标轮廓。
在一个实施例中，确定病灶区域之后，病灶分割结果确定模块1605还用于：基于形态学运算对病灶区域进行处理，确定处理后的病灶区域的边界像素为对应的目标轮廓。
关于超声图像病灶的分割装置的具体限定可以参见上文中对于超声图像病灶的分割方法的限定，在此不再赘述。上述超声图像病灶的分割装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中，也可以以软件形式存储于计算机设备中的存储器中，以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中，提供了一种计算机设备，该计算机设备可以是服务器，其内部结构图可以如图17所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中，该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储超声图像数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种超声图像病灶的分割方法。
本领域技术人员可以理解，图17中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定，具体的计算机设备可以包括比图中所示更多或更少的部件，或者组合某些部件，或者具有不同的部件布置。
在一个实施例中，提供了一种计算机设备，包括存储器和处理器，存储器 中存储有计算机程序，该处理器执行计算机程序时实现以下步骤：
识别超声图像中病灶，以得到对应的感兴趣区域；
根据感兴趣区域，构造初始零水平集函数，其中，初始零水平集函数表示病灶区域的初始轮廓；
基于初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：将感兴趣区域的各顶点坐标确定为参考点，根据设定的平移量对参考点进行平移，以得到相对于各参考点的新的顶点坐标；根据新的顶点坐标得到病灶区域的初始轮廓；基于感兴趣区域以及病灶区域的初始轮廓得到初始零水平集函数。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：通过梯度下降法使得能量泛函最小化，得到水平集活动轮廓演化方程；基于水平集活动轮廓演化方程采用设定的步长和迭代次数迭代计算，以得到演化终止时刻的目标水平集函数。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：根据感兴趣区域的直方图分布，估计病灶的回声类型；基于回声类型确定对应的步长和迭代次数。
在一个实施例中，回声类型包括无回声和低回声，则处理器执行计算机程序时还实现以下步骤：若回声类型为无回声时，则对应的迭代次数为80～260，时间步长为0.1～1；若回声类型为低回声时，则对应的迭代次数为280～320， 时间步长为0.1～1。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：对目标零水平集二值图像进行反色处理，以得到待筛选的多个前景区域；对待筛选的多个前景区域进行孔洞填充，以得到填充后的新的目标零水平集二值图像；根据邻域连通准则在新的目标零水平集二值图像中提取各连通分量，并计算各连通分量所在区域的面积；将各连通分量所在区域的面积中面积最大的连通分量所在区域确定为病灶区域，则病灶区域的边界像素为对应的目标轮廓。
在一个实施例中，处理器执行计算机程序时还实现以下步骤：确定病灶区域之后，基于形态学运算对病灶区域进行处理，确定处理后的病灶区域的边界像素为对应的目标轮廓。
在一个实施例中，提供了一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现以下步骤：
识别超声图像中的病灶，以得到对应的感兴趣区域；
根据感兴趣区域，构造初始零水平集函数，其中，初始零水平集函数表示病灶区域的初始轮廓；
基于初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
通过梯度下降法求解能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
基于目标水平集函数得到对应的目标零水平集二值图像，对目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：将感兴趣区域的各顶点坐标确定为参考点，根据设定的平移量对参考点进行平移，以得到相对于各参考点的新的顶点坐标；根据新的顶点坐标得到病灶区域的初始轮 廓；基于感兴趣区域以及病灶区域的初始轮廓得到初始零水平集函数。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：通过梯度下降法使得能量泛函最小化，得到水平集活动轮廓演化方程；基于水平集活动轮廓演化方程采用设定的步长和迭代次数迭代计算，以得到演化终止时刻的目标水平集函数。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：根据感兴趣区域的直方图分布，估计病灶的回声类型；基于回声类型确定对应的步长和迭代次数。
在一个实施例中，回声类型包括无回声和低回声，则计算机程序被处理器执行时还实现以下步骤：若回声类型为无回声时，则对应的迭代次数为80～260，时间步长为0.1～1；若回声类型为低回声时，则对应的迭代次数为280～320，时间步长为0.1～1。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：对目标零水平集二值图像进行反色处理，以得到待筛选的多个前景区域；对待筛选的多个前景区域进行孔洞填充，以得到填充后的新的目标零水平集二值图像；根据邻域连通准则在新的目标零水平集二值图像中提取各连通分量，并计算各连通分量所在区域的面积；将各连通分量所在区域的面积中面积最大的连通分量所在区域确定为病灶区域，则病灶区域的边界像素为对应的目标轮廓。
在一个实施例中，计算机程序被处理器执行时还实现以下步骤：确定病灶区域之后，基于形态学运算对病灶区域进行处理，确定处理后的病灶区域的边界像素为对应的目标轮廓。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于 一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合，为使描述简洁，未对上述实施例中的各个技术特征所有可能的组合都进行描述，然而，只要这些技术特征的组合不存在矛盾，都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本申请构思的前提下，还可以做出若干变形和改进，这些都属于本申请的保护范围。因此，本申请专利的保护范围应以所附权利要求为准。

Claims (10)

  1. 一种超声图像病灶的分割方法，其特征在于，所述方法包括：
    识别超声图像中的病灶，以得到对应的感兴趣区域；
    根据所述感兴趣区域，构造初始零水平集函数，所述初始零水平集函数表示病灶区域的初始轮廓；
    基于所述初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
    通过梯度下降法求解所述能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
    基于所述目标水平集函数得到对应的目标零水平集二值图像，对所述目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述感兴趣区域，构造初始零水平集函数，包括：
    将所述感兴趣区域的各顶点坐标确定为参考点，根据设定的平移量对所述参考点进行平移，以得到相对于各参考点的新的顶点坐标；
    根据所述新的顶点坐标得到所述病灶区域的初始轮廓；
    基于所述感兴趣区域以及所述病灶区域的初始轮廓得到所述初始零水平集函数。
  3. 根据权利要求1所述的方法，其特征在于，所述通过梯度下降法求解所述能量泛函的最小值，以得到演化终止时刻的目标水平集函数，包括：
    通过梯度下降法使得所述能量泛函最小化，得到水平集活动轮廓演化方程；
    基于所述水平集活动轮廓演化方程采用设定的步长和迭代次数迭代计算，以得到演化终止时刻的目标水平集函数。
  4. 根据权利要求3所述的方法，其特征在于，所述步长和迭代次数的设定方法包括：
    根据所述感兴趣区域的直方图分布，估计所述病灶的回声类型；
    基于所述回声类型确定对应的步长和迭代次数。
  5. 根据权利要求4所述的方法，其特征在于，所述回声类型包括无回声和低回声，所述基于所述回声类型确定对应的步长和迭代次数，包括：
    若所述回声类型为无回声时，则对应的迭代次数为80～260，时间步长为0.1～1；
    若所述回声类型为低回声时，则对应的迭代次数为280～320，时间步长为0.1～1。
  6. 根据权利要求1至5任一项所述的方法，其特征在于，所述对所述目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓，包括：
    对所述目标零水平集二值图像进行反色处理，以得到待筛选的多个前景区域；
    对所述待筛选的多个前景区域进行孔洞填充，以得到填充后的新的目标零水平集二值图像；
    根据邻域连通准则在所述新的目标零水平集二值图像中提取各连通分量，并计算所述各连通分量所在区域的面积；
    将所述各连通分量所在区域的面积中面积最大的连通分量所在区域确定为所述病灶区域，则所述病灶区域的边界像素为对应的目标轮廓。
  7. 根据权利要求6所述的方法，其特征在于，所述确定所述病灶区域之后，还包括：
    基于形态学运算对所述病灶区域进行处理，确定处理后的所述病灶区域的边界像素为对应的目标轮廓。
  8. 一种超声图像病灶的分割装置，其特征在于，所述装置包括：
    感兴趣区域识别模块，用于识别超声图像中的病灶，以得到对应的感兴趣区域；
    初始零水平集函数构造模块，用于根据所述感兴趣区域，构造初始零水平集函数，所述初始零水平集函数表示病灶区域的初始轮廓；
    能量泛函定义模块，用于基于所述初始零水平集函数利用局部二值拟合演化算法定义能量泛函；
    演化模块，用于通过梯度下降法求解所述能量泛函的最小值，以得到演化终止时刻的目标水平集函数；
    病灶分割结果确定模块，用于基于所述目标水平集函数得到对应的目标零水平集二值图像，对所述目标零水平集二值图像进行后处理，以得到病灶区域的目标轮廓。
  9. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，其特征在于，所述处理器执行所述计算机程序时实现权利要求1至7中任一项所述方法的步骤。
  10. 一种计算机可读存储介质，其上存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至7中任一项所述方法的步骤。
PCT/CN2020/133028 2019-12-25 2020-12-01 超声图像病灶的分割方法、装置和计算机设备 WO2021129324A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911358146.2 2019-12-25
CN201911358146.2A CN113034508B (zh) 2019-12-25 2019-12-25 超声图像病灶的分割方法、装置和计算机设备

Publications (1)

Publication Number Publication Date
WO2021129324A1 true WO2021129324A1 (zh) 2021-07-01

Family

ID=76458524

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/133028 WO2021129324A1 (zh) 2019-12-25 2020-12-01 超声图像病灶的分割方法、装置和计算机设备

Country Status (2)

Country Link
CN (1) CN113034508B (zh)
WO (1) WO2021129324A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114612483A (zh) * 2022-03-04 2022-06-10 北京理工大学 硬管内窥镜视野区域分割方法及装置

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113657553B (zh) * 2021-09-01 2023-12-26 什维新智医疗科技(上海)有限公司 一种结节回声类型判断装置
CN117274216B (zh) * 2023-10-09 2024-04-16 聆数医疗科技(苏州)有限公司 基于水平集分割的超声下颈动脉斑块检测方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102903103A (zh) * 2012-09-11 2013-01-30 西安电子科技大学 基于迁移活动轮廓模型的胃部ct序列图像分割方法
CN104732213A (zh) * 2015-03-23 2015-06-24 中山大学 一种基于乳腺磁共振图像的计算机辅助肿块检测方法
US20170039725A1 (en) * 2015-08-05 2017-02-09 Algotec Systems Ltd. Method and system for spatial segmentation of anatomical structures
CN106803250A (zh) * 2016-12-28 2017-06-06 辽宁师范大学 结合全局和局部区域信息的医学超声图像分割方法
US20170199654A1 (en) * 2016-01-08 2017-07-13 National Yang-Ming University System and method of mapping medical image of single heart chamber

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB0818561D0 (en) * 2008-10-09 2008-11-19 Isis Innovation Visual tracking of objects in images, and segmentation of images
CN107481252A (zh) * 2017-08-24 2017-12-15 上海术理智能科技有限公司 医学图像的分割方法、装置、介质和电子设备

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102903103A (zh) * 2012-09-11 2013-01-30 西安电子科技大学 基于迁移活动轮廓模型的胃部ct序列图像分割方法
CN104732213A (zh) * 2015-03-23 2015-06-24 中山大学 一种基于乳腺磁共振图像的计算机辅助肿块检测方法
US20170039725A1 (en) * 2015-08-05 2017-02-09 Algotec Systems Ltd. Method and system for spatial segmentation of anatomical structures
US20170199654A1 (en) * 2016-01-08 2017-07-13 National Yang-Ming University System and method of mapping medical image of single heart chamber
CN106803250A (zh) * 2016-12-28 2017-06-06 辽宁师范大学 结合全局和局部区域信息的医学超声图像分割方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114612483A (zh) * 2022-03-04 2022-06-10 北京理工大学 硬管内窥镜视野区域分割方法及装置

Also Published As

Publication number Publication date
CN113034508A (zh) 2021-06-25
CN113034508B (zh) 2024-02-13

Similar Documents

Publication Publication Date Title
WO2021129323A1 (zh) 超声图像病灶描述方法、装置、计算机设备和存储介质
WO2021129325A1 (zh) 超声图像病灶的分割方法、装置和计算机设备
CN108010021B (zh) 一种医学图像处理系统和方法
Huang et al. Breast ultrasound image segmentation: a survey
WO2021129324A1 (zh) 超声图像病灶的分割方法、装置和计算机设备
CN105957063B (zh) 基于多尺度加权相似性测度的ct图像肝脏分割方法及系统
Foruzan et al. Improved segmentation of low-contrast lesions using sigmoid edge model
Abbas et al. Skin tumor area extraction using an improved dynamic programming approach
Sreeja et al. An improved feature based image fusion technique for enhancement of liver lesions
JP2023502814A (ja) 画像検出モデルのトレーニング方法及びその関連装置、機器、記憶媒体
JP2016007270A (ja) 医用画像処理装置
Abbas et al. Combined spline and B-spline for an improved automatic skin lesion segmentation in dermoscopic images using optimal color channel
Anter et al. Automatic liver parenchyma segmentation system from abdominal CT scans using hybrid techniques
Zhu et al. Fast feature-preserving speckle reduction for ultrasound images via phase congruency
Cavalcanti et al. A coarse-to-fine approach for segmenting melanocytic skin lesions in standard camera images
Kaur et al. Computer-aided diagnosis of renal lesions in CT images: a comprehensive survey and future prospects
Jalab et al. Fractional Renyi entropy image enhancement for deep segmentation of kidney MRI
Jain et al. IFCM based segmentation method for liver ultrasound images
Singh et al. Mesh-free technique for enhancement of the lung CT image
CN112634280B (zh) 基于能量泛函的mri图像脑肿瘤分割方法
CN111784664B (zh) 肿瘤淋巴结分布图谱生成方法
Anter et al. Automatic segmentation and classification of liver abnormalities using fractal dimension
Huang et al. Neural network analysis applied to tumor segmentation on 3D breast ultrasound images
Joda et al. Digital mammogram enhancement based on automatic histogram clipping
Mahmoud et al. Novel feature extraction methodology based on histopathalogical images and subsequent classification by Support Vector Machine

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20905141

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20905141

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20905141

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 26.01.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20905141

Country of ref document: EP

Kind code of ref document: A1