CN109636810B

CN109636810B - Pulmonary nodule segmentation method and system of CT image

Info

Publication number: CN109636810B
Application number: CN201811469539.6A
Authority: CN
Inventors: 王兴维; 邰从越; 刘龙; 王慧; 史黎鑫; 尹延伟; 刘慧芳
Original assignee: Senyint International Digital Medical System Dalian Co ltd
Current assignee: Senyint International Digital Medical System Dalian Co ltd
Priority date: 2018-11-28
Filing date: 2018-11-28
Publication date: 2020-11-03
Anticipated expiration: 2038-11-28
Also published as: CN109636810A

Abstract

A lung nodule segmentation method, a system, a region growing method, a segmentation end judging method and a method for cutting an adhesion region of a CT image belong to the field of medical image processing, and in order to solve the problem that different types of nodules in a lung nodule image in the prior art cannot be automatically and accurately segmented according to actual edges of the nodules, the key points are that S3, a connected region of the nodules is obtained, and S5, whether the segmentation is ended or not is judged by using an adjacent interlayer relation; s6, cutting the adhesion area, wherein the effect is that the adhesion between the nodule and the blood vessel can be quickly and effectively segmented.

Description

Pulmonary nodule segmentation method and system of CT image

Technical Field

The invention belongs to the field of medical image processing, and relates to a lung nodule automatic segmentation method based on a CT image.

Background

Lung cancer is the cancer with the highest mortality rate. The lung cancer generally appears in the form of lung nodules in early stage, and the diagnosis and treatment of the lung nodules in early stage can improve the 5-year survival rate of patients after the operation. Computed Tomography (CT) is a good imaging method for detecting lung nodules, but the image quantity generated by CT scanning is huge, which can reach dozens to hundreds of layers, so that radiologists have heavy reading burden and are easy to cause fatigue, and further the reading efficiency and quality are reduced, and misdiagnosis and missed diagnosis are caused at a certain probability. A Computer Aided Detection (CAD) system can automatically help doctors to detect nodules, assist the doctors in finding focuses and improve the accuracy of diagnosis. The lung nodule segmentation from the lung CT image is an important application of an image processing technology in medical images and has important significance in computer-aided detection. The common pulmonary nodule types are mainly divided into isolated type, vascular adhesion type and the like. Solitary pulmonary nodules are generally not connected to any tissue in the lung; the blood vessel adhesion type pulmonary nodules are usually connected with blood vessels, the malignant probability of the nodules is the highest, the gray value of the nodules is usually close to that of the blood vessels, and accurate segmentation cannot be achieved only by considering gray information. Therefore, it is significant and difficult to accurately segment the vascular adhesion-type pulmonary nodules, and the existing nodule segmentation methods mainly include morphology-based methods, dynamic programming-based methods, iterative threshold-based methods, neural network-based methods, and the like. The method mainly has the problems of low segmentation precision, poor robustness and the like, and can not accurately separate nodules from blood vessels, so that the subsequent pulmonary nodule detection part can be influenced.

Disclosure of Invention

In order to solve the problem that different types of nodules in a lung nodule image in the prior art can not be automatically and accurately segmented according to actual edges of the nodules, and realize accurate segmentation of blood vessel adhesion type lung nodules from a lung CT image, the invention provides the following technical scheme: a lung nodule automatic segmentation method based on CT images comprises the following steps:

s1, inputting volume data V of a region of interest containing suspected nodules cut according to rules and a central point of the suspected nodules;

s2, according to the central point coordinate Pc, two-dimensional data S of a Z-direction corresponding layer is obtained, a threshold value T is obtained within the range of the two-dimensional data S, and the threshold value T is used for acting on the volume data V to generate binary volume data M;

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

s5, judging whether the segmentation is finished or not by utilizing the relation between adjacent layers;

and S6, cutting the adhesion area.

Further, step S2 includes the following method:

(1) the center point coordinate Pc is generally expressed as (x, y, z), and the data of the z-th layer is taken to generate two-dimensional data S;

(2) taking two-dimensional data S as input, and solving a threshold value T by using the Otsu method;

(3) after obtaining the threshold value T, traversing the volume data V, setting the point of which the gray value is greater than the threshold value T as 1, setting the point of which the gray value is less than the threshold value T as 0,

new binarized volume data M is generated.

Further, step S3 includes the following method, which uses the idea of region growing, and is performed in two steps:

(1) taking the input central point Pc as a seed point, carrying out region growth with the length of 2 adjacent regions being 1 in the vertical direction of the binary volume data M, and growing seed points SPs of each layer;

(2) and (3) performing region growth with the length of 4 adjacent regions being 1 on each layer of the binary volume data M by using the seed points generated in the step (1), and finally growing new binary volume data G with only one connected region.

Further, step S4 includes the following method: the small-area adhesion is cut off by using a morphological opening operation on the binarized volume data G to generate binarized volume data O with small connections broken.

Further, step S5 includes the following method: comparing the areas of the communicated regions of the adjacent layers in the binary volume data O, and finishing the segmentation of the nodule if the area difference is within a set reasonable range; if the area difference exceeds the set reasonable range, the corresponding layer with larger area has blood vessels connected with the suspected nodule area, and is set as a layer to be processed, and then the step S6 is carried out.

Further, step S6 includes the following method: extracting two-dimensional data of a layer to be processed from the binary volume data O, extracting boundary information of adjacent layers, taking median points of upper and lower adjacent layers as cutting boundaries according to the attribute that the boundary of the nodule is in gentle transition, and cutting off the nodule and the blood vessel to realize segmentation between the nodule and the blood vessel.

A system for automatic segmentation of lung nodules based on CT images, having stored thereon a plurality of instructions adapted to be loaded and executed by a processor:

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

and S6, cutting the adhesion area.

A region growing method of CT image includes:

(2) using the generated seed points, region growth with 4 adjacent regions of length 1 is performed on each layer of the binarized volume data M, and finally, new binarized volume data G having only one connected region is grown.

A CT image segmentation end judging method is characterized in that in binary volume data O, the areas of adjacent communicated regions are compared, and if the area difference is within a set reasonable range, nodule segmentation is completed; if the area difference exceeds the set reasonable range, the corresponding layer with larger area is provided with blood vessels connected with the suspected nodule area and is set as a layer to be treated.

A method for cutting adhesion areas of CT images comprises the steps of extracting two-dimensional data of a layer to be processed from binary volume data O, extracting boundary information of adjacent layers, taking median points of upper and lower adjacent layers as cutting boundaries according to the attribute that a nodule boundary is in gentle transition, cutting off nodules and blood vessels, and achieving segmentation between the nodules and the blood vessels.

Has the advantages that: the invention can segment the nodules at different positions through one process, and can rapidly and effectively segment the nodules especially aiming at the condition that the nodules are adhered to blood vessels. For the nodules at the isolated positions, the nodules are obvious, the gray difference with the surrounding is large, and the segmentation is easy; for a nodule with adhesion to a blood vessel, the gray level of the blood vessel is similar to that of the nodule, and the nodule region cannot be accurately segmented only through a gray level threshold. After the initial threshold segmentation, a novel region growing mode is used, so that the condition of blood vessel adhesion is more effectively reduced compared with the traditional mode. The region growing in the vertical direction is to consider the incidence relation among layers of the three-dimensional data, the region growing in the horizontal direction is to process the actual communication condition of single-layer two-dimensional data, and a nodule has one and only one communication region in any dimension. Therefore, the invention combines the vertical direction and the horizontal direction, and can ensure that only one communication area exists in any layer. For the data layer with adhesion, a method is adopted to cut the nodule from the blood vessel. The edge of the nodule is a gentle transition process, and the cliff type growth cannot occur, so that the method fully utilizes the process, takes out the upper and lower adjacent layers of the layer to be segmented, counts effective boundary information, calculates the intermediate value of the upper and lower layer boundaries as a segmentation line, and accords with the attribute of edge transition.

Drawings

FIG. 1 is a graph comparing results;

FIG. 2 is a diagram of the results of an ON operation;

FIG. 3 is a graph of segmentation results;

fig. 4 is a process flow diagram of the method of the present invention.

Detailed Description

Referring to fig. 1, a method for automatically segmenting lung nodules based on CT images includes the following steps:

s1, inputting region-of-interest volume data V containing suspected nodules and cut according to a certain rule and a central point of the suspected nodules.

The step is the input data of the invention, considering the calculation speed and the actual engineering requirements, the original input data of the invention is the three-dimensional data V of the suspected nodule part in the lung CT image and the center point coordinate Pc of the corresponding suspected nodule, the two input data can be obtained by other automatic processing processes in the product of the company, and can also be generated by clicking the suspected point and the radius or the surrounding frame by the user or obtaining the corresponding area by other methods, and the method of the invention has no influence.

And S2, according to the central point coordinate Pc, two-dimensional data S of a layer corresponding to the Z direction is obtained, a threshold value T is obtained by applying the Otsu method within the range of the data S, and the threshold value T is applied to the volume data V to generate binary volume data M. The reason why the threshold is calculated by using the two-dimensional data S instead of the volume data V is to reduce the amount of calculation and save the calculation time. In addition, the related Z direction is the direction of the Z axis of the three-dimensional coordinate system, the Z axis is divided into a plurality of layers, and one layer is the Z-th layer;

wherein, the following methods are involved:

(1) the center point coordinate Pc is generally expressed as (x, y, z), and the data of the z-th layer is taken to generate two-dimensional data S, which can also be called as a two-dimensional array S;

(2) the two-dimensional data S is used as input, and Otsu' S method is used to obtain the threshold value T. When using the Otsu method, it should be considered that the input data may have the background region removed, i.e. the non-lung region is set as the background value, such as-2000, or other special regions are set as special values but should not be counted. For these cases, when using Otsu, these special values should be excluded and not included in the threshold calculation;

(3) after the threshold value T is obtained, traversing the volume data V, setting the point with the gray value larger than T as 1, and setting the point with the gray value smaller than T as 0, and generating new binary volume data M.

S3, acquiring connected regions of nodules

Using the region growing idea, the method comprises two steps:

(1) taking the input central point Pc as a seed point, carrying out region growth with the length of 2 adjacent regions being 1 in the vertical direction of the data M, and growing seed points SPs of each layer;

(2) using the seed points generated in (1), region growth with 4 adjacent regions of length 1 is performed on each layer of the data M, and finally new binarized volume data G with only one connected region is grown.

Conventional three-dimensional 6-neighborhood growth can effectively grow isolated nodules, but nodules that adhere to blood vessels cannot separate the nodules from the vessel region. Due to the growth in 6 three-dimensional directions, there may be a case where the a layers are connected and the b layers are not connected, so that two or more connected regions of the two-dimensional data on a certain layer may be left (as shown in the middle column of fig. 1). The method can ensure that only one connected area is reserved for the two-dimensional data of each layer, and other non-adhered parts can be simply and quickly deleted.

S4, morphological opening operation and disconnection of interference connection

Some nodules may have inevitable adhesion with blood vessels, and small-area adhesion can be cut off by using morphological opening operation on the data G, so that the binary volume data O with small connection broken is generated.

S5, judging whether the division is finished or not by utilizing the relation between adjacent layers

In the data O, comparing the areas of the communicated regions of the adjacent layers, and if the area difference is within a reasonable range, completing the segmentation of the nodule; if the area difference is larger, the layer with larger area is correspondingly provided with blood vessels connected with the suspected nodule area, the layer is set as a layer to be treated, and then the step 6 is carried out.

S6, cutting the adhesion area

And extracting the two-dimensional data of the layer to be processed in the step S5 from the data O, extracting boundary information of adjacent layers, taking median points of upper and lower adjacent layers as cutting boundaries according to the property that the nodule boundaries are in smooth transition, and cutting off the nodule and the blood vessel to realize the segmentation between the nodule and the blood vessel. Since this step processes only a small number of layers of data, the amount of computation is reduced compared to processing all volume data.

The region growing method in the step 3 is a simple and effective new application method

The method in step 5 can distinguish the nodule type, and different treatment modes are used for different types of nodules, for example, the nodules with vascular adhesion can be treated in step 6, the calculated amount and the calculated occupied space can be effectively reduced in the process, and the method is suitable for engineering requirements

The segmentation method in the step 6 is a rapid segmentation method meeting engineering requirements

The detection of pulmonary nodules can be generally divided into the following processes, 1. initial detection of nodules and determination of suspected areas; 2. accurately segmenting nodules; 3. and judging whether the nodule is true or false. The invention solves the problem of accurate segmentation of the nodule in step 2.

The following is a description of further technical solutions to explain the above solutions in more detail.

Step S1, which is data input from outside.

Step S2, two-dimensional data S is obtained from the three-dimensional volume data V by using a conventional method, the two-dimensional data S is recorded as S (x, y), the segmentation threshold values of the foreground (namely the suspected nodule area) and the background (lung parenchyma) are recorded as T, the proportion of the number of pixels belonging to the foreground in the whole image is recorded as omega 1, and the average gray scale of the pixel is mu 1; the ratio of the number of background pixels to the whole image is ω 0, and the average gray level is μ 0. The total mean gray level of the image is denoted as μ and the between-class variance (i.e., the variance between the foreground and background) is denoted as g.

Assuming that the size of the image is M × N and the nodules are higher than the lung parenchyma gray level, the number of pixels in the image with the gray level of the pixel less than the threshold T is denoted as N0, and the number of pixels with the gray level greater than the threshold T is denoted as N1, and there are:

ω0＝N0/(M×N) (1)

ω1＝N1/(M×N) (2)

N0+N1＝M×N (3)

ω0+ω1＝1 (4)

μ＝ω0*μ0+ω1*μ1 (5)

g＝ω0*(μ0-μ)^2+ω1*(μ1-μ)^2 (6)

substituting formula (5) for formula (6) yields the equivalent formula:

g＝ω0*ω1*(μ0-μ1)^2 (7)

equation (7) is the inter-class variance. And obtaining the threshold T which enables the inter-class variance g to be maximum by adopting a traversal method.

Generally, when traversing each gray value of an image, a gray histogram of the image is counted first, and then each gray value in the histogram is traversed to perform calculation. Since the gray-level value of the CT image is negative, a preprocessing is performed to shift the gray-level value to a region without negative value. The method adopted can translate according to a fixed value, such as adding 2000 (the minimum value of the gray value is 2000), the method can cause a large number of situations with a lower value of 0, and wastes space, the translation method adopted in the invention is to count the maximum value and the minimum value (Val _ max and Val _ min) of data, then translate according to the minimum value, such as adding 500 to all gray values, and initialize the length of a histogram to Val _ max-Val _ min +1, so that the memory space can be fully utilized.

In the process of histogram statistics, there are some special cases, such as the case where the lung parenchyma may have been segmented in advance in the incoming image data, and the background region other than the lung is set to a special value, such as-2000, or there are other similar cases where the special value is set but should not be recorded in the subsequent calculation process, and for this case, the position statistics of the special value is to be recorded as 0 in the histogram statistics.

After the threshold value T is obtained, traversing each pixel point (x, y, z) of the three-dimensional volume data V, and obtaining binary data M according to the following rules:

in step S3, acquiring connected regions of nodules is different from the conventional region growing method, and is divided into two steps of vertical region growing and horizontal region growing:

(1) vertical direction region growth

And taking the input central point Pc as a seed point, and performing region growth with the length of 2 adjacent regions being 1 in the vertical direction (namely the z direction) of the data M, wherein the growth operator is [ (0,0,1), (0,0, -1) ]. And the data M can grow when the value of the corresponding point is 1, and stops growing when the value of the corresponding point is 0, so that a point meeting the conditions can be grown in each layer according to the growing conditions, and the grown points are used as the standby seed points SPs. Since the shape of the nodule is not a perfect regular sphere, the input center point Pc may also be slightly shifted, and considering the deviation, the initial seed point is expanded by a proper proportion to grow the true edge of the nodule, for example, in the two-dimensional data S, the center point Pc is used as the center, 2 pixels (each pixel should satisfy M (x, y, z) is 1) are expanded outwards to serve as a seed point group, and then the region growth in the vertical direction is performed.

(2) Horizontal direction region growth

Using the seed points SPs grown in (1), region growth with 4 adjacent regions of length 1 is performed on each layer of the data M, the growth operators are [ (0, -1,0), (-1,0,0), (0,1,0), (1,0,0) ], the stop condition is as (1), and finally, new binarized volume data G with only one connected region is grown.

FIG. 1 is a graph showing the results of comparing the method with the conventional method. In the figure, only 3 layers of data (30, 31 and 32 layers) are taken as illustrations, the first column is an original figure with the lung parenchyma segmented, the second column is a result figure obtained by using a traditional region growing method after binarizing data, and the third column is a result obtained by using the method in the invention, interference data of non-connected regions are removed, and the data are in accordance with the expectation. For the data at layer 30, the nodule is not completely separated from the vessel and is subsequently processed accordingly.

In step S4, the interference connection is disconnected using a conventional morphological open operation. The operation result is schematically shown in fig. 2, the first column is an original diagram of the segmented lung parenchyma, the second column is a result diagram after the binarization data, and the third column is a result diagram after the opening operation is used, the small connection is broken, and the result diagram is in accordance with the expectation. However, after some data is opened, the data is not only provided with one connected unit, and in actual operation, the result data is subjected to conventional region growth with the unit of 1 in the three-dimensional 6-adjacent domain, and finally new binary volume data O is generated.

In step S5, it is determined whether the division process is finished.

And traversing data of each layer, recording the area of the connected region, and comparing the area difference of adjacent layers. The specific situation is different because the layer thicknesses of different data are different, and if the layer thickness is larger, the area difference between adjacent layers is also larger. For unifying the judgment rule, after calculating the area difference ratio, the conversion relation between the pixel coordinate and the physical coordinate is introduced, the thickness of the adjacent layer is considered, and finally, a reasonable range of the area difference ratio under the thickness is given. Specifically, assuming that the layer thickness of the data is slicewickness, the area of the current layer connected region is a0, and the area of the adjacent layer connected region is a1, the area difference ratio is calculated as follows:

note that the pixel pitch in the X, Y direction cancels out each other when dif is obtained, so no special calculation is required. The converted ratio is as follows:

dif_convert＝dif/SliceThickness

if dif_convertIf the number is less than 1, the nodule segmentation is finished; if dif_convertAnd (4) more than or equal to 1, indicating that the corresponding layer with larger area has blood vessels which are connected with the suspected nodule area greatly, defining the layer as a layer to be treated, and carrying out the treatment of the step 6.

And step S6, extracting the data C of the layer to be processed, and cutting the adhesion area.

And taking the data U of the upper layer and the data D of the lower layer of the data C layer to be processed, wherein the communication area of the data C layer is between the communication area of the data U layer and the communication area of the data D layer in principle.

And calculating the coordinates of the outer enclosure box of each layer of data, and comparing the coordinate ranges to obtain the approximate position of the adhesion area. And calculating the boundary point of the corresponding position communication area of the upper layer data U and the lower layer data D of the current data layer C in the adhesion area, and calculating the middle position point Pe of the two layers of data. After traversing the boundary points of all the adhesion areas, generating a position point set Pes, wherein the Pes is a segmentation boundary between the data C and the suspected nodule and the blood vessel. The segmentation result is shown in fig. 3, the first is an original image of the segmented lung parenchyma, the second is a binarization result image after the region growing in the present invention, and the third is a result image obtained by using the above-mentioned segmentation method.

And if the adjacent multiple layers are the data layers to be processed, continuously taking the non-adhered data layers upwards or downwards, and acquiring the segmentation boundary by using the thought.

Specifically, if the upper layer or the lower layer of the layer to be processed has no connected region, i.e. no-doubt similar region, and the layer to be processed is the first layer or the last layer of the suspected region, the position of the adhesion region can be obtained in the same manner, and the divided boundary points can be calculated according to the percentage of the adjacent layers. The value of the percentage can be found by the ratio between the adjacent second layer and the adjacent first layer.

The invention can segment the nodules at different positions through one process, and can rapidly and effectively segment the nodules especially aiming at the condition that the nodules are adhered to blood vessels. For the nodules at the isolated positions, the nodules are obvious, the gray difference with the surrounding is large, and the segmentation is easy; for a nodule with adhesion to a blood vessel, the gray level of the blood vessel is similar to that of the nodule, and the nodule region cannot be accurately segmented only through a gray level threshold. After the initial threshold segmentation, a novel region growing mode is used, so that the condition of blood vessel adhesion is more effectively reduced compared with the traditional mode. The region growing in the vertical direction is to consider the incidence relation among layers of the three-dimensional data, the region growing in the horizontal direction is to process the actual communication condition of single-layer two-dimensional data, and a nodule has one and only one communication region in any dimension. Therefore, the invention combines the vertical direction and the horizontal direction, and can ensure that only one communication area exists in any layer.

For the data layer with adhesion, a method is adopted to cut the nodule from the blood vessel. The edge of the nodule is a gentle transition process, and the cliff type growth cannot occur, so that the method fully utilizes the process, takes out the upper and lower adjacent layers of the layer to be segmented, counts effective boundary information, calculates the intermediate value of the upper and lower layer boundaries as a segmentation line, and accords with the attribute of edge transition.

The above description is only for the purpose of creating a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can substitute or change the technical solution and the inventive concept of the present invention within the technical scope of the present invention.

Claims

1. A lung nodule automatic segmentation method based on CT image is characterized by comprising the following steps:

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

s6, cutting the adhesion area;

step S5 includes the following method: comparing the areas of the communicated regions of the adjacent layers in the binary volume data O, and finishing the segmentation of the nodule if the area difference is within a set reasonable range; if the area difference exceeds the set reasonable range, the corresponding layer with larger area has blood vessels connected with the suspected nodule area, and is set as a layer to be processed, and then the step S6 is carried out.

2. The method for automatically segmenting lung nodules based on CT image as claimed in claim 1, wherein step S2 includes the following steps:

(3) after the threshold value T is obtained, traversing the volume data V, setting the point with the gray value larger than the threshold value T as 1, and setting the point with the gray value smaller than the threshold value T as 0, and generating new binary volume data M.

3. The method for automatically segmenting lung nodules based on CT image as claimed in claim 1, wherein step S3 includes the following method, using the idea of region growing, in two steps:

4. The method for automatically segmenting lung nodules based on CT image as claimed in claim 1, wherein step S4 includes the following steps: the small-area adhesion is cut off by using a morphological opening operation on the binarized volume data G to generate binarized volume data O with small connections broken.

5. The method for automatically segmenting lung nodules based on CT image as claimed in claim 1, wherein step S6 includes the following steps: extracting two-dimensional data of a layer to be processed from the binary volume data O, extracting boundary information of adjacent layers, taking median points of upper and lower adjacent layers as cutting boundaries according to the attribute that the boundary of the nodule is in gentle transition, and cutting off the nodule and the blood vessel to realize segmentation between the nodule and the blood vessel.

6. A lung nodule automatic segmentation method based on CT image is characterized by comprising the following steps:

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

s6, cutting the adhesion area;

step S6 includes the following method: extracting two-dimensional data of a layer to be processed from the binary volume data O, extracting boundary information of adjacent layers, taking median points of upper and lower adjacent layers as cutting boundaries according to the attribute that the boundary of the nodule is in gentle transition, and cutting off the nodule and the blood vessel to realize segmentation between the nodule and the blood vessel.

7. The method for automatically segmenting lung nodules based on CT image as claimed in claim 6, wherein step S2 includes the following steps:

8. The method for automatically segmenting lung nodules based on CT image as claimed in claim 6, wherein step S3 includes the following method, using the idea of region growing, in two steps:

9. The method for automatically segmenting lung nodules based on CT image as claimed in claim 6, wherein step S4 includes the following steps: the small-area adhesion is cut off by using a morphological opening operation on the binarized volume data G to generate binarized volume data O with small connections broken.

10. A system for automatic segmentation of lung nodules based on CT images, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor to:

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

s6, cutting the adhesion area;

step S5 is implemented based on the following manner: comparing the areas of the communicated regions of the adjacent layers in the binary volume data O, and finishing the segmentation of the nodule if the area difference is within a set reasonable range; if the area difference exceeds the set reasonable range, the corresponding layer with larger area has blood vessels connected with the suspected nodule area, and is set as a layer to be processed, and then the step S6 is carried out.

11. A system for automatic segmentation of lung nodules based on CT images, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor to:

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

s6, cutting the adhesion area;

step S6 is implemented based on the following manner: extracting two-dimensional data of a layer to be processed from the binary volume data O, extracting boundary information of adjacent layers, taking median points of upper and lower adjacent layers as cutting boundaries according to the attribute that the boundary of the nodule is in gentle transition, and cutting off the nodule and the blood vessel to realize segmentation between the nodule and the blood vessel.

12. A system for automatic segmentation of lung nodules based on CT images, wherein a plurality of instructions are stored, the instructions being adapted to be loaded and executed by a processor to:

s3, acquiring a connected region of the nodule;

s4, morphologically opening operation to disconnect interference connection;

s6, cutting the adhesion area;

step S5 is implemented based on the following manner: comparing the areas of the communicated regions of the adjacent layers in the binary volume data O, and finishing the segmentation of the nodule if the area difference is within a set reasonable range; if the area difference exceeds the set reasonable range, connecting the corresponding layer with larger area with blood vessels and the suspected nodule area, setting the layer as a layer to be processed, and then performing step S6;