Initial sketch based stomach CT (Computed Tomography) image lymph node detection system

A technology for CT images and lymph nodes, applied in the field of image processing, can solve the problems of high false alarm rate of lymph nodes, long use time, and inaccurate location of lymph node boundaries, so as to achieve accurate extraction, improve processing speed, and reduce false alarm rate.

Active Publication Date: 2016-03-16
XIDIAN UNIV
4 Cites 2 Cited by

AI-Extracted Technical Summary

Problems solved by technology

This method is based on the grayscale features of the pixels in the CT image, so it takes a long time, and because the shape of the lymph node is simila...
View more

Abstract

The present invention discloses an initial sketch based stomach CT (Computed Tomography) image lymph node detection system. The detection system comprises eight functional modules: a gray-scale value statistic module collects statistics of gray-scale values of pixel points in a CT image; a sketch line pre-processing module performs pre-processing on sketch lines; a sketch line based adaptive window determining module is an adaptive window for extracting the sketch lines; a suspected lymph node image block sequence extracting module extracts an image block sequence based on the sketch lines and the adaptive window; a line-row sectional drawing based lymph node detection module tracks and determines a suspected lymph node, the image block sequence of which is greater than 11*11; a low rank decomposition module and a centroid tracking module track and determine a suspected lymph node, the image block sequence of which is not greater than 11*11; and a sketch line marking module marks the sketch lines enclosed by the tracked suspected lymph nodes as processed. According to the initial sketch based stomach CT image lymph node detection system provided by the present invention, sketch information of the CT image is used, so that the lymph node detection speed is improved and the false alarm rate is reduced, and the system cab be applicable to medical image processing.

Application Domain

Image enhancementImage analysis

Technology Topic

Computed tomographyDecomposition +8

Image

  • Initial sketch based stomach CT (Computed Tomography) image lymph node detection system
  • Initial sketch based stomach CT (Computed Tomography) image lymph node detection system
  • Initial sketch based stomach CT (Computed Tomography) image lymph node detection system

Examples

  • Experimental program(1)

Example Embodiment

[0045] The technical solutions and effects of the present invention are described in detail below in conjunction with the accompanying drawings:
[0046] refer to figure 1 The detection system of the present invention includes: a gray value statistics module 1, a sketch line preprocessing module 2, a sketch line-based adaptive window determination module 3, a suspected lymph node image block sequence extraction module 4, and a lymph node detection based on row and column cut map Module 5 , low-rank decomposition module 6 , centroid tracking module 7 , sketch line marking module 8 . in:
[0047] The gray value statistics module 1 is used to divide each CT image in the original CT sequence diagram into four categories through the otsu classification algorithm, and count the gray value range where each category is located. The largest order is the gray value range of the background, the gray value range of the fat, the gray value range of the suspected lymph node, and the gray value range of the highlighted area;
[0048] The sketch line preprocessing module 2 includes: sketch line connection sub-module 21, sketch line length feature extraction sub-module 22, sketch line-based CT image grayscale feature extraction sub-module 23, sketch line polygonal gathering shape feature extraction sub-module twenty four.
[0049] The sketch line connection sub-module 21 is used to find out whether there is an end point of other sketch lines in the 3×3 neighborhood centered on the sketch point where one end point of the current sketch line is located. The sketch line is connected to the current sketch line; if there are multiple endpoints, find the angle between the sketch line segment where each endpoint is located and the sketch line segment where the endpoint of the current sketch line is located, and take the sketch line segment with the largest angle value. The sketch line is connected; the sketch line connection is to change the found end point to the end point of the current sketch line; perform the above operation on the new sketch line formed by the connection, until all the end points of all sketch lines are in the 3×3 area. If there is no other end point of the sketch line, the connection operation is completed;
[0050] The sketch line length feature extraction sub-module 22 is used to extract sketch lines with a length in the range of [6, 40], wherein the sketch line length refers to the sum of the number of sketch points on all sketch line segments included in each sketch line;
[0051] The sketch line-based CT image grayscale feature extraction sub-module 23 is used to make a vertical line for the longest sketch line segment in the input sketch line, and its length is 11 sketch points, that is, the longest sketch line segment in the sketch line is drawn. Take 5 sketch points on each side of the long sketch line segment, map these 11 sketch points to the corresponding 11 pixel points in the original CT image, and find the maximum gray value of these 11 pixel points in the original CT image gray_max and Minimum gray value gray_min, extract the sketch lines with gray_max within [90, 200] and gray_min within (0, 75], where [90, 200] is the gray value range of the suspected lymph node in the original CT image; (0, 75] is the original The gray value range of fat in CT images;
[0052] The sketch line polygon shape feature extraction sub-module 24 is used to extract the shape features of a sketch line containing two or more sketch line segments; note that the number of sketch line segments included in the sketch line is n, n≥2, if n=2, find the angle value of the unique vertex, if the angle value is in the range of [10,145], mark the sketch line as a sketch line with a polygonal gathering shape; if n≥3, find the corresponding n-1 The angle value of the vertex, if the n-1 angle values ​​are all in the range of [10,145] degrees, then find out whether the i-1 and i+1 lines of the two sketch line segments adjacent to the i-th sketch line segment are in the ith sketch line segment. The i sketch line segments are on the same side of the straight line, 2≤i≤n-1, if they are, it is considered that these three sketch line segments have a tendency to wrap around. If the sketch line segments on the sketch line have a tendency to wrap around, mark the Sketch Lines Sketch lines with polygonal gather shapes.
[0053] The sketch line-based adaptive window determination module 3 includes: a rectangular window extraction sub-module 31 for polygonal gathering shape sketch lines, a maximum area window extraction sub-module 32 based on polygonal gathering shape sketch line matching, and a square window extraction for a single sketch line segment. Submodule 33 , maximum area window extraction submodule 34 based on single sketch line segment matching, and sketch line adaptive window extraction submodule 35 .
[0054] The rectangular window extraction sub-module 31 of the polygonal gathering shape sketch line is used to obtain the maximum value x of the abscissas of the endpoints of all sketch line segments contained in the sketch line line marked as a polygonal gathering shape. max , the minimum value of the abscissa x min , the maximum value of the ordinate y max , the minimum value of the ordinate y min , and extract it with (x min , y min ), (x min , y max ), (x max , y min ) and (x max , y max ) The rectangular area with these four points as vertices is used as the rectangular window RWin of the polygonal gathering shape sketch line line, and the frame where the polygonal gathering shape sketch line rectangular window RWin is located is taken out in the initial sketch sequence diagram as the current frame, forward , the sketch map of each 5 frames backwards, in the sketch map of each frame, extract the sketch lines in the rectangular area corresponding to the rectangular window RWin or part of the sketch lines in the rectangular area in the sketch map, and use these sketch lines Match with the polygonal gathering shape sketch line line, and extract the rectangular window of the sketch lines matched in the first and last 5 frames according to the matching sketch lines; perform the above operations on all marked polygonal gathering shape sketch lines ;
[0055] The maximum area window extraction sub-module 32 based on the matching of polygonal gathering shape sketch lines is used to extract the matching degree of these sketch lines and polygonal gathering shape sketch lines for the sketch lines matched in each of the five frames of sketches before and after, and based on the matching Extract the rectangular window with the largest area corresponding to the polygonal gathering shape sketch line line;
[0056] The square window extraction sub-module 33 of the single sketch line segment is used to find a square window with the midpoint of the sketch line segment as the center, and the length of the sketch line segment is twice as long as the side length for the sketch line sline containing only one sketch line segment SWin, and take the frame where the sketch line sline is located as the current frame in the initial sketch sequence diagram, and take the sketch diagram of 5 frames forward and backward, in the sketch diagram of each frame, extract the sketch diagram with the square in the sketch diagram For the sketch lines in the square area or part of the square area corresponding to the window SWin, match these sketch lines with the sketch line sline, and extract the matched sketch lines in the previous and next 5 frames according to the matched sketch lines the square window of ; do the above for all sketch lines that contain only one sketch line segment and whose length is in the range [6,10];
[0057] The maximum area window extraction sub-module 34 based on the matching of a single sketch line segment is used to extract the matching degree of these sketch lines and the sketch line sline for the sketch lines matched in each of the five frames of sketches before and after, and extract the matching degree based on the matching degree. The square window with the largest area corresponding to the sketch line sline of a sketch line segment;
[0058] The sketch line adaptive window extraction sub-module 35 is used to convert the rectangular window with the largest area corresponding to the polygonal gathering shape sketch line line and the square window with the largest area corresponding to the sketch line sline containing only one sketch line segment upward, downward, and upward. Extend 5 sketch points to the left and right to obtain the adaptive window of the polygonal gathering shape sketch line and the adaptive window of the sketch line sline containing only one sketch line segment.
[0059] The suspected lymph node image block sequence extraction module 4 includes: a local image block extraction sub-module 41, a sub-module 42 for determining the area surrounded by the sketch line, a sub-module 43 for determining a suspected lymph node, and an adaptive window extracting sub-module for the suspected lymph node on the CT image. Module 44, a sub-module 45 for extracting a sequence of suspected lymph node image blocks.
[0060] The local image block extraction sub-module 41 is used to extract the local image in the area where the adaptive window is located from the CT image of the frame where the adaptive window of the current sketch line is located, and set the gray value in the local image in the gray value where the suspected lymph node is located. The value of the pixel within the range of the degree value is set to 1, and the others are set to 0;
[0061] The sub-module 42 for determining the area surrounded by the sketch line is used to extract the center sketch point of the longest sketch line segment contained in the current sketch line, and in the partial image block, find the pixel corresponding to the center sketch point and the distance from the pixel The connected area where the pixel with the nearest grayscale value of 1 is located. If there are other connected areas with a value of 1 in the local image block, the value of the pixels contained in these areas is set to 0;
[0062] The determination sub-module 43 of the suspected lymph node is used to determine whether the connected region extracted by the determination sub-module (42) surrounded by the sketch line is a suspected lymph node; that is, the distance between the boundary pixels of the region and the centroid pixels of the region is calculated. The mean value c and variance e of the ellipse are recorded as class ellipticity t=c/e. If t≥1, the area is determined as a suspected lymph node, and the regional information of the suspected lymph node is retained; otherwise, it is determined as a non-suspected lymph node and will not be retained. ;
[0063] The self-adaptive window extraction submodule 44 of the suspected lymph node on the CT image is used to extract the minimum circumscribed rectangle of the region where the suspected lymph node is located, and the smallest circumscribed rectangle is used as the self-adaptive window of the suspected lymph node on the CT image;
[0064] The suspected lymph node image block sequence extraction sub-module 45 is used to take the area where the adaptive window of the suspected lymph node is located as a fixed area, take the frame where the suspected lymph node is located on the original CT sequence image as the current frame, and take one frame forward and one backward each. The local image blocks of the fixed area of ​​​​are sorted according to the order of the frame, to obtain the image block sequence D of the suspected lymph node, where l=20.
[0065]The lymph node detection module 5 based on the row-column cut map includes: a row-cut map construction sub-module 51, a row-cut map to determine the first and last frames sub-module 52, a column-cut map to construct a sub-module 53, a column-cut map to determine the first and last frames sub-module 54, Lymph node determination sub-module 55 .
[0066] The row-cut map construction sub-module 51 is used for the suspected lymph node image block sequence D to sequentially take out the first sub-image to the i-th row of the (21+1)-th sub-image, and construct a row-cut map R according to the following rules i :
[0067] Take the i-th row of the first sub-image as the first row of the new image, take the i-th row of the second sub-image as the second row of the new image, and so on, take the i-th row of the j-th sub-image as the new image's row The jth row, until the ith row of all images in the sequence diagram D is taken, and the row cut diagram R corresponding to the ith row of the sequence diagram D is obtained. i , where i=1~m, m is the number of rows of each image in sequence diagram D;
[0068] The line cut map determines the head and tail frame sub-module 52, which is used for the line cut map R (z-1) Find the abscissa l1 of the uppermost row and the abscissa l4 of the lowermost row in the area where the point (l+1,ceny) is located; cut the graph R to the row z Find the abscissa l2 of the top row and the abscissa l5 of the bottom row in the area where the point (l+1, ceny) is located; cut the graph R to the row (z+1) Find the abscissa l3 of the top row and the abscissa l6 of the bottom row of the area where the point (l+1, ceny) is located; calculate the first frame position s and tail frame position e of the suspected lymph node in the suspected lymph node image block sequence D:
[0069] s=min(min(l1,l2),l3)
[0070] e=max(max(l4,l5),l6)
[0071] Where z=m/2+1, ceny is the ordinate of the currently tracked suspected lymph node centroid;
[0072] The column cut map construction sub-module 53 is used to sequentially take out the first sub-image to the kth column of the (21+1)th sub-image in the suspected lymph node image block sequence D, and construct a column cut map according to the following rules C k :
[0073] Take the k-th column of the first sub-image as the first row of the new image, take the k-th column of the second sub-image as the second column of the new image, and so on, take the k-th column of the j-th sub-image as the new image's The jth column, until the kth column of all images in the sequence diagram D is taken, and the column cut diagram C corresponding to the kth column of the sequence diagram D is obtained. k , where k=1~n, n is the number of columns of each image in the sequence diagram D;
[0074] The column cut map determines the first and last frame sub-module 54, which is used for the column cut map C (z-1) Find the ordinate h1 of the leftmost column and the ordinate h4 of the rightmost column in the region where the point (cenx,l+1) is located; z Find the ordinate h2 of the leftmost column and the ordinate h5 of the rightmost column in the area where the point (cenx, l+1) is located; (z+1) , find the ordinate h3 of the leftmost column and the ordinate h6 of the rightmost column of the region where the point (cenx,l+1) is located, and update the first frame position s and tail frame position e of the suspected lymph node in sequence D:
[0075] s=min(s,s t )
[0076] ,
[0077] e=max(e,e t )
[0078] where s t =min(min(h1,h2),h3), e t =max(max(h4,h5),h6), z=n/2+1, cenx is the abscissa of the currently tracked suspected lymph node centroid;
[0079] The lymph node determination sub-module 55 is used to determine the suspected lymph nodes whose image blocks are larger than 11×11 in the suspected lymph node image block sequence D according to the area c of the suspected lymph node currently tracked and the area s of the suspected lymph node in the first frame 1 and tail frame suspected lymph node area e 1 mark
[0080] Note whether it is a lymph node:
[0081] if Then mark it as a lymph node, and save the frame number, region information and label information of the suspected lymph node in the suspected lymph node sequence diagram, otherwise, mark it as a non-lymph node.
[0082] The low-rank decomposition module 6 includes: an observation matrix construction sub-module 61 , a decolor decomposition sub-module 62 , and an initial head and tail frame determination sub-module 63 .
[0083] The observation matrix construction sub-module 61 copies cl copies of the image block where the suspected lymph node currently tracked in the image block sequence D of the suspected lymph node is located, to obtain a low-rank decomposed image block sequence F. The secondary image blocks are pulled into a column and placed in the order of the image blocks to obtain the observation matrix data M, where cl is 40;
[0084] The decolor decomposition submodule 62 performs low-rank decomposition on the observation matrix M through the decolor low-rank model to obtain the background sequence B;
[0085] The initial head and tail frame determination sub-module 63 calculates the variance v=[v of each column in the background sequence B left ,v cur ,v right ], where v left Represents the variance corresponding to the leftmost l column in the B sequence, v right Indicates the corresponding to the rightmost l column in the B sequence, v cur Represents the variance corresponding to columns l+1 to l+cl in the middle of the B sequence, and calculates v left The frame number whose variance is less than 1, and the largest frame number is marked as the initial first frame, and v is calculated right Frame numbers with a variance less than 1 in the middle, and mark the smallest frame number as the initial tail frame.
[0086] The centroid tracking module 7 includes: a forward centroid tracking sub-module 71 , a backward centroid tracking sub-module 72 , and a lymph node and blood vessel sub-module 73 .
[0087] The forward centroid tracking submodule 71 is used to track the target area corresponding to the current suspected lymph node forward on the original CT sequence diagram, starting from the current frame, to obtain the first frame of the currently tracked suspected lymph node. Its tracking principle is:
[0088] On the original CT sequence diagram, starting from the current frame, taking the size of the adaptive window of the suspected lymph node of the current frame as the fixed size of the window, and taking the centroid of the tracked target as the centroid of the window, the target matching in the window is performed forward. If it is successful, the target area that is successfully matched is the area of ​​the tracking target, and the target centroid is updated, and the matching continues forward to obtain the forward sequence of the tracking target, and the leftmost frame in the forward sequence is used as the current tracked suspected lymph node. The first frame; if the match fails, the forward tracking is terminated;
[0089] The backward centroid tracking sub-module 72 is used to start from the current frame on the original CT sequence diagram, and track backward the target area corresponding to the current suspected lymph node to obtain the tail frame of the currently tracked suspected lymph node. Its tracking principle is:
[0090] On the original CT sequence diagram, starting from the current frame, taking the size of the adaptive window of the suspected lymph node of the current frame as the fixed size of the window, taking the centroid of the tracking target as the centroid of the window, and performing the target matching in the window backwards, if it matches If it is successful, the target area that is successfully matched is the area of ​​the tracking target, and the target centroid is updated, and the matching is continued backward to obtain the backward sequence of the tracking target, and the rightmost frame in the backward sequence is used as the current tracked suspected lymph node. End the frame; if the match fails, terminate the backward tracking;
[0091] The lymph node and blood vessel area sub-module 73, for the image blocks in the image block sequence D of the currently tracked suspected lymph node is not larger than 11×11 suspected lymph nodes, according to the sequence length l of the complete sequence of suspected lymph nodes and the area s of the suspected lymph node in the first frame 3 , the area e of the suspected lymph node in the tail frame 3 , the distance d between the centroid of the suspected lymph node in the first frame and the centroid of the suspected lymph node in the last frame, the average value of the area change of the suspected lymph node between adjacent frames in the sequence a, the area s of the suspected lymph node in the initial first frame 2 , the area e of the suspected lymph node in the initial tail frame 2 , to distinguish between lymph nodes and blood vessels:
[0092] If l<20,l>=6,s 3 2 ,e 3 2 , d<10, a>5, it is marked as a lymph node, and the frame number, region information and labeling information of the suspected lymph node in the suspected lymph node sequence map are saved; otherwise, it is not marked.
[0093] The sketch line marking module 8 is used to mark sketch lines surrounded by all suspected lymph nodes tracked between the first and last frames as processed from the pre-processed set of sketch lines.
[0094] refer to figure 2 , the present invention based on the initial sketch map of the stomach CT image lymph node detection method, comprising the following steps:
[0095] Step 1: Count the gray value.
[0096] Each CT image in the original CT sequence image is divided into four categories by the otsu classification algorithm, the gray value range of each category is counted separately, and the gray value range of the background is in the order of gray value from small to large. gray1, gray value range gray2 where the fat is located, gray value range gray3 where the suspected lymph node is located, and gray value range gray4 where the highlighted area is located;
[0097] Step 2, extract the initial sketch map for each CT image in the original CT sequence map, such as image 3 shown, where image 3 (a) is the CT image to be detected, 3(b) is from image 3 (a) Extracted initial sketch map.
[0098] Extracting the initial sketch map References: C.E.Guo, S.C.Zhu, Y.N.Wu.PrimalSketch:IntegratingTextureandStructure[J].ComputerVisionandImageUnderstanding,2007,106(1):5-19.
[0099] Step 3: Preprocess the sketch lines in all the extracted initial sketch images to obtain a set of preprocessed sketch lines PreLines.
[0100] 3.1) For each sketch line line included in the initial sketch map currently being processed, find out whether there are other sketch line endpoints in the 3×3 neighborhood centered on the sketch point where one endpoint of the sketch line line is located, and if so And there is only one endpoint, connect the sketch line where the endpoint is located with the current sketch line line; if there are multiple endpoints, find the clip between the sketch line segment where each endpoint is located and the sketch line segment where the endpoint is located in the current sketch line line. Angle, take the sketch line corresponding to the sketch line segment with the largest included angle value for connection, where the sketch line connection is to change the found end point to the end point of the current sketch line; perform the above operations on the new sketch line formed by the connection, until all the If there are no endpoints of other sketch lines in the 3×3 area where all the endpoints of the sketch line are located, the connection operation is completed, and the initial sketch line set InitLines is obtained;
[0101] 3.2) Retain the sketch lines whose length is in the range of [6,40] in the initial sketch line set InitLines, and obtain the length sketch line set LenLines, where the length of the sketch line refers to the number of sketch points on all sketch line segments included in each sketch line sum; as Figure 4 shown, where Figure 4 (a) is the rendering of mapping the sketch lines in the set of length sketch lines LenLines to the CT image, Figure 4 (b) is the rendering of the sketch lines contained in the set of length sketch lines LenLines;
[0102] 3.3) To each sketch line in the length sketch line set LenLines, the longest sketch line segment in the sketch line is made a vertical line, and its length is 11 sketch points, that is, 5 are taken on both sides of the longest sketch line segment. Sketch points, map these 11 sketch points to the corresponding 11 pixel points in the original CT image, and find the maximum gray value gray_max and minimum gray value gray_min of these 11 pixel points in the original CT image, and extract gray_max in the For the sketch lines within [90,200] and gray_min within (0,75], the grayscale sketch line set GrayLines is obtained, where [90,200] is the gray value range gray3 of the suspected lymph node in the original CT image; (0,75] is the original The gray value range gray2 of fat in CT images; such as Image 6 shown, where Image 6 (a) is the rendering of mapping the sketch lines in the grayscale sketch line set GrayLines to the CT image, Image 6 (b) is the rendering of the sketch lines included in the grayscale sketch line collection GrayLines;
[0103] 3.4) Take the sketch line that contains n sketch line segments in the gray-scale sketch line set GrayLines, n>=2, if n=2, then find the angle value of its unique vertex, if the angle value is in the range of [10,145] , then mark the sketch line as a sketch line with a polygonal gathering shape, and keep the sketch line; if n≥3, find the angle value of the corresponding n-1 vertices, if the n-1 angle values ​​are all in [10,145] Within the range of degrees, then find out whether the i-1 and i+1-th sketch line segments adjacent to the i-th sketch line segment are on the same side of the straight line where the i-th sketch line segment is located, 2≤i≤n-1, if it is , then it is considered that these three sketch line segments have a tendency to wrap around. If the sketch line segments on the sketch line have a tendency to wrap around, then mark the sketch line as a sketch line with a polygonal gathering shape, and keep the sketch line; keep all the sketch lines. Lines form a collection of polygonal gather shape sketch lines ShapeLines; such as Figure 7 7(a) is the rendering of mapping the sketch lines in the polygonal sketch line collection ShapeLines to the CT image, Figure 7 (b) is the rendering of the sketch lines contained in the polygonal sketch line collection ShapeLines;
[0104] Among them, whether the i-1 and i+1-th sketch line segments adjacent to the i-th sketch line segment are on the same side of the straight line where the i-th sketch line segment is located is judged by the following method:
[0105] Note that the four endpoints of the three sketch line segments i-1, i and i+1 are p1, p2, p3, p4, where the endpoint p1 is the unique endpoint of the i-1 sketch line segment, the endpoint p2 is the common endpoint of the i-1 and i-th sketch line segments, the endpoint p3 is the common endpoint of the i-th and i+1-th sketch line segments, and the endpoint p4 is the unique endpoint of the i+1-th sketch line segment, If the endpoint p1 and the endpoint p4 are on one side of the i-th sketch line, the i-1 and i+1-th sketch line segments are considered to be on the same side of the i-th sketch line segment; otherwise, they are not on the same side;
[0106] 3.5) Take the grayscale sketch line set GrayLines that only contains one sketch line segment and the length is within the range of [6, 10], and store these sketch lines and the sketch lines contained in the polygonal gathering shape sketch line set ShapeLines into the preprocessing In the sketch line collection PreLines; the above preprocessing operation is performed for each initial sketch image, and the sketch lines obtained after preprocessing and their corresponding frame number information are stored in the preprocessing sketch line collection PreLines.
[0107] Step 4: Extract the adaptive window of the current sketch line based on the initial sketch sequence diagram.
[0108] 4.1) Take the first unprocessed sketch line line in the preprocessing sketch line set PreLines as the current sketch line, and the frame cur where it is located as the current frame, determine whether the current sketch line line contains only one sketch line segment, and if so, execute the steps 4.2), if not, go to step 4.3);
[0109] 4.2) Take the midpoint of the current sketch line as the center, make a square window SWin with twice the length of the sketch line as the side length, and take out the initial sketches of the first 5 frames and the last 5 frames of the current frame in the initial sketch sequence diagram Figure, in the initial sketch map of each frame, extract the sketch lines in the area corresponding to the square window SWin or part of the area, obtain the local sketch line set LocalLines, and judge whether the local sketch line set LocalLines exists and the current sketch A sketch line whose included angle is less than 30 degrees and whose center point distance is not more than 5 sketch points, if it exists, the sketch line with the closest distance is taken as the sketch line matching the current sketch line line in the initial sketch image of the frame, with Use the same method as the current sketch line to make a square window; if it does not exist, it is considered that there is no sketch line matching the current sketch line in the initial sketch image of the frame, and no other operations are performed; take it out from all the obtained square windows The square window with the largest area, expand the window up, down, left, and right by 5 sketch points to obtain the adaptive window PrimWin of the current sketch line; remember the frame where the adaptive window PrimWin is located as win_zhen, and set the The current sketch line is marked as processed; eg Figure 9 (a) shows the process of extracting the largest square window of a sketch line that only contains one sketch line segment, in which ☆ marks the frame cur where the current sketch line line is located, and ★ marks the frame where the largest square window is located. The implementation in the image is the matched sketch line, and the box is the square window found based on the matched sketch line on the frame; Figure 9 The box in (b) is the adaptive window PrimWin of the sketch line obtained by expanding the extracted largest square window.
[0110] 4.3) Find the maximum value x of the abscissa of the endpoints of all sketch line segments contained in the current sketch line max , the minimum value of the abscissa x min , the maximum value of the ordinate y max , the minimum value of the ordinate y min , will be represented by (x min , y min ), (x min , y max ), (x max , y min ) and (x max , y max ) The rectangular area with these four points as vertices is used as the rectangular window RWin of the current sketch line, and the initial sketches of the first 5 frames and the last 5 frames of the current frame are taken out in the initial sketch sequence diagram, and in each frame of the initial sketch diagram , extract the sketch lines in or part of the area corresponding to the rectangular window RWin, obtain the local sketch line set LocalLines, and determine the distance from the local sketch line set LocalLines for each sketch line segment cline in the current sketch line line Whether the angle between the sketch line closest to the sketch line cline and the sketch line cline is less than 30 degrees, if so, it is considered that the sketch line matches the sketch line cline; for all the sketch line segments that match each other, find the distance between their center points, If the average value of all the obtained distances is not greater than 5 pixels, it is considered that there is a sketch line matching the current sketch line line on the initial sketch map of the frame, and a rectangular window is made in the same way as the current sketch line line; If the distance is greater than 5, it is considered that there is no sketch line matching the current sketch line on the initial sketch map of the frame, and it will not be processed; take out the rectangular window with the largest area among all the obtained rectangular windows, and move the window up and down. , expand 5 sketch points to the left and right to obtain the adaptive window PrimWin of the current sketch line line, record the frame where the adaptive window PrimWin is located as win_zhen, and mark the current sketch line line as processed; such as Figure 8 shown, where Figure 8 (a) is the display of the extraction process of the largest rectangular window of the polygonal gathering shape sketch line, where ☆ marks the frame cur where the current sketch line line is located, and ★ marks the frame where the largest rectangular window is located. is the matched sketch line, the box is the rectangular window found based on the matched sketch line on this frame, Figure 8 The box in (b) is the adaptive window PrimWin of the polygonal gathering shape sketch line obtained by expanding the extracted largest rectangle.
[0111] Step 5: Extract the suspected lymph node region based on the position of the current sketch line and its adaptive window.
[0112] 5.1) Input the CT image of the frame win_zhen where the adaptive window PrimWin of the current sketch line line is located, extract the local image of the area where the adaptive window PrimWin is located on the image, and set the gray value in the local image at the gray level where the suspected lymph node is located. The gray value of the pixels in the value range gray3 is set to 1, and the others are set to 0;
[0113] 5.2) Take the center sketch point of the longest sketch line segment contained in the current sketch line, find the pixel point corresponding to the center sketch point in the local image block, and find the gray value closest to the pixel point is 1 The connected area of ​​the pixel with the nearest value of 1 is the area surrounded by the current sketch line on the CT image of the frame where the adaptive window is located. If there are other areas in the local image block Connected regions with a value of 1, set the value of the pixels contained in these regions to 0;
[0114] 5.3) Calculate the average value c and variance e of the distance between the boundary pixel point of the connected area area and the centroid pixel point of the area, and record the class ellipticity t=c/e. If t≥1, the area is determined to be suspected Lymph node, keep the area information of the suspected lymph node, and continue to step 6; otherwise, it is determined as a non-suspected lymph node, and determine whether there are unprocessed sketch lines in the preprocessing sketch line set PreLines, if so, go back to step 4; if not , the detection is over, and the area marked as lymph node is displayed on the original CT image. like Figure 10 shown, where Figure 10 The larger box in (a) is the effect of mapping the adaptive window of the sketch line to the CT image, where the solid line is the position of the current sketch line; Figure 10 (b) The white dots in the box represent the extracted suspected lymph nodes.
[0115] Step 6, extract a sequence of suspected lymph node image blocks.
[0116] Taking the extracted suspected lymph node as the current suspected lymph node, extracting the smallest circumscribed rectangle of the region where it is located, taking this region as the fixed region, taking the frame where the suspected lymph node is located on the original CT sequence image as the current frame, and taking l forward and backward. The local image blocks in the fixed area of ​​the frame are sorted according to the order of the frame, to obtain the image block sequence D of the suspected lymph node, where l=20, such as Figure 10 The smaller box in (b) is the adaptive window of the extracted suspected lymph nodes, Figure 11 Image block sequence of suspected lymph nodes marked in (a).
[0117] Step 7: Determine the size of the image block in the image block sequence D of the current suspected lymph node. If it is greater than 11×11, perform step 8; otherwise, jump to step 12.
[0118] Step 8: Calculate the head and tail frame areas of the currently tracked suspected lymph node.
[0119] 8.1) Input the current suspected lymph node image block sequence D obtained in step 7, take out the first sub-image to the i-th row of the (21+1)-th sub-image in sequence, and construct a row cut map R i , that is, the i-th row of the first sub-image is taken as the first row of the new image, the i-th row of the second sub-image is taken as the second row of the new image, and so on, the i-th row of the j-th sub-image is taken as the new image The jth row of , until the ith row of all images in the sequence diagram D is taken, and the row cut diagram R corresponding to the ith row of the sequence diagram D is obtained. i , where i=1~m, m is the number of rows of each image in sequence diagram D;
[0120] 8.2) Calculate the upper and lower boundary information of the line cut graph in the area where the point (l+1,ceny) is located:
[0121] Row cut graph in R (z-1) Find the abscissa l1 of the upper boundary and the abscissa l4 of the lower boundary of the area where the point (l+1,ceny) is located;
[0122] Row cut graph in R z Find the abscissa l2 of the upper boundary and the abscissa l5 of the lower boundary of the area where the point (l+1, ceny) is located;
[0123] Row cut graph in R (z+1) Find the abscissa l3 of the upper boundary and the abscissa l6 of the lower boundary of the area where the point (l+1, ceny) is located;
[0124] Where z=m/2+1, ceny is the ordinate of the currently tracked suspected lymph node centroid;
[0125] like Figure 11 The white frame in (b) is the area where the point (l+1,ceny) is located;
[0126] 8.3) According to the result of step 8.2), calculate the first frame position s and the last frame position e of the suspected lymph node in the original row and column cut sequence diagram D:
[0127] s=min(min(l1,l2),l3)
[0128];
[0129] e=max(max(l4,l5),l6)
[0130] 8.4) In the current suspected lymph node image block sequence D, take the first sub-image to the k-th column of the (21+1) sub-image in sequence, and construct a row cut map C k , that is, the k-th column of the first sub-image is taken as the first row of the new image, the k-th column of the second sub-image is taken as the second column of the new image, and so on, the k-th column of the j-th sub-image is taken as the new image The jth column of , until the kth column of all images in the sequence diagram D is taken, and the column cut diagram C corresponding to the kth column of the sequence diagram D is obtained. k , where k=1~n, n is the number of columns of each image in the sequence diagram D;
[0131] 8.5) Calculate the left and right boundary information of the column cut graph in the area where the point (cenx, l+1) is located:
[0132] In Lechet Diagram C (z-1) Find the ordinate h1 of the left boundary of the area where the point (cenx, l+1) is located and the ordinate h4 of the right boundary;
[0133] In Lechet Diagram C z Find the ordinate h2 of the left boundary of the area where the point (cenx, l+1) is located and the ordinate h5 of the right boundary;
[0134] In Lechet Diagram C (z+1) Find the ordinate h3 of the left boundary of the area where the point (cenx, l+1) is located and the ordinate h4 of the right boundary;
[0135] Where z=n/2+1, cenx is the abscissa of the currently tracked suspected lymph node centroid;
[0136] like Figure 11 The white box in (c) is the area where the point (cenx, l+1) is located;
[0137] 8.6) According to the result of step 8.5), calculate the first frame position s and the last frame position e of the suspected lymph node in the current suspected lymph node image block sequence D:
[0138] s=min(s,x)
[0139] ,
[0140] e=max(e,y)
[0141] Where x=min(min(h1, h2), h3), y=max(max(h4, h5), h6), z=n/2+1, cenx is the abscissa of the currently tracked suspected lymph node centroid;
[0142] 8.7) According to the first and last frames of the suspected lymph node determined in step 8.6), calculate the area s of the suspected lymph node in the first frame 1 , the area e of the suspected lymph node in the tail frame 1 and the area c of the suspected lymph node currently tracked.
[0143] In step 9, the suspected lymph nodes are marked and preserved.
[0144] According to the area s of the suspected lymph node in the first frame obtained in step 8 1 , the area e of the suspected lymph node in the tail frame 1 , the currently tracked area c of the suspected lymph node to mark the lymph node: if then labelled as lymph node; otherwise, labelled as non-lymph node; such as Figure 12 The white rectangle in the middle is the tracked lymph node, the position marked by the five-pointed star is the suspected lymph node in the current frame, the position marked by the left diamond is the first frame, and the position marked by the right diamond is the tail frame.
[0145] Step 10, obtain the complete sequence of the suspected lymph node.
[0146] From the image block sequence D of the currently tracked suspected lymph node, take out the image block sequence between the first frame and the last frame, take the frame where the currently tracked suspected lymph node is located as the current frame, and use the fixed-window centroid tracking method to move forward and backward respectively. Then, the suspected lymph node region corresponding to the current suspected lymph node on each frame is extracted, and the complete sequence of the suspected lymph node is obtained.
[0147] Step 11, mark the sketch line.
[0148] For each suspected lymph node in the complete sequence of suspected lymph nodes, determine whether the sketch line corresponding to it in the initial sketch map is in PreLines, if so, mark it as processed; If the processed sketch line exists, return to step 4 for execution; otherwise, the detection ends, and the region marked as lymph node is displayed on the original CT image.
[0149] Step 12: Calculate the area of ​​the initial head and tail frames of the suspected lymph node whose image block sequence is not larger than 11×11.
[0150] 12.1) Copy cl copies of the image block where the currently tracked suspected lymph node is located in the image block sequence D of the currently tracked suspected lymph node to obtain a low-rank decomposed image block sequence F;
[0151] 12.2) Pull each image in the low-rank decomposition image block sequence F into a column to form an observation matrix M;
[0152] 12.3) Perform low-rank decomposition on the observation matrix M through the decolor low-rank model to obtain the background sequence B;
[0153] 12.4) Calculate the variance v=[v of each column in the background sequence B left ,v cur ,v right ], where v left Represents the variance corresponding to the leftmost l column in the B sequence, v right Represents the variance corresponding to the rightmost l column in the B sequence, v cur Represents the variance corresponding to columns l+1 to l+cl in the B sequence;
[0154] take out v left All column numbers whose variance is less than 1, and the largest column number plus cur-l is recorded as the initial first frame, and the area s of the suspected lymph node in the initial first frame is calculated. 2 , that is, the number of 1s in the largest column number column in the background sequence B;
[0155] take out v right All column numbers with variance less than 1, and the smallest column number plus cur-l-cl is recorded as the initial tail frame, and the area e of the suspected lymph node in the initial tail frame is calculated. 2 , that is, the number of 1s in the minimum column number column in the background sequence B; calculate the area c of the suspected lymph node currently tracked, that is, the number of 1s in the l+1th column in the background sequence B.
[0156] Step 13: Obtain a complete sequence of suspected lymph nodes whose image block sequence is not larger than 11×11.
[0157] 13.1) Input the original CT sequence image, set the value of the pixel whose gray value is within the range of the suspected lymph node gray value in each CT image to 1, and set the others to 0 to obtain the binary image corresponding to the original CT sequence image sequence;
[0158] 13.2) Take the frame where the current suspected lymph node is located as the current frame, on the binary sequence image obtained in step 13.1), start with the current frame, and take the size of the current frame suspected lymph node adaptive window as the fixed size of the window to track The centroid of the target is the centroid of the window, and the target matching in the window is performed forward and backward respectively:
[0159] 13.21) Carry out the target matching in the window forward, and judge whether the pixel value of the point where the target centroid is located is 1. If it is 1, the matching is successful, update the target centroid to the center of the area where the target centroid is located, and continue to match forward; otherwise , the matching fails, stop forward matching;
[0160] 13.22) Carry out the target matching in the window backwards, determine whether the pixel value of the point where the target centroid is located is 1, if it is 1, the matching is successful, update the target execution to the center of the area where the target centroid is located, and continue to match backwards; otherwise, match If it fails, stop the backward matching and get the backward sequence of the tracking target;
[0161] 13.3) The complete sequence of the suspected lymph node is composed of the forward tracking sequence and the backward tracking sequence.
[0162] Step 14, distinguish between lymph nodes and blood vessels.
[0163] Enter the sequence length l of the complete sequence of the suspected lymph node obtained in step 13, and the area s of the suspected lymph node in the first frame 3 , the area e of the suspected lymph node in the tail frame 3 , the distance d between the centroid of the suspected lymph node in the first frame and the centroid of the suspected lymph node in the last frame, the average value a of the area change of the suspected lymph node between adjacent frames in the sequence, and the area s of the suspected lymph node in the initial first frame obtained in step 12 2 , the area e of the suspected lymph node in the initial tail frame 2 , to distinguish between lymph nodes and blood vessels:
[0164] If l>=20,s 3=s 2 ,e 3=e 2 , d>10, a
[0165] The above description is only a specific example of the present invention, and does not constitute any limitation to the present invention. Obviously, for those skilled in the art, after understanding the content and principle of the present invention, they may not deviate from the principle and structure of the present invention. Various modifications and changes in form and details are made under the circumstances of the present invention, but these modifications and changes based on the idea of ​​the present invention are still within the protection scope of the claims of the present invention.

PUM

PropertyMeasurementUnit
Angle<= 30.0deg

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.

Similar technology patents

Data processing method and device

Owner:ADVANCED NEW TECH CO LTD

Driver smoking behavior real-time monitoring method based on vehicle video monitoring

InactiveCN105976570Aprocessing speedSoftware and hardware resource requirements are not high
Owner:SHANDONG NORMAL UNIV

JPEG_LS rule coding hardware achieving method based on scanning sequence changing

InactiveCN104717497Aprocessing speed
Owner:北京润光泰力科技发展有限公司

Multipicture synthesizing method and system

InactiveCN104168427Aprocessing speedgood effect
Owner:SUZHOU CODYY NETWORK SCI & TECH

Preparation method and application for graphene multi-metal composite material

InactiveCN110015746AExcellent organic matter degradation abilityprocessing speed
Owner:SHANGHAI UNIV

Classification and recommendation of technical efficacy words

  • reduce processing time
  • processing speed

Method for preprocessing in photographing process of mobile terminal and mobile terminal

ActiveCN103581528AReduce preprocessing timereduce processing time
Owner:BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD

Radiotherapeutic apparatus

InactiveUS6879659B2reduce processing timeshorter treatment time
Owner:ELEKTA AB

Truth random number producing method and device

InactiveCN102176199Agood randomnessprocessing speed
Owner:XI'AN INST OF OPTICS & FINE MECHANICS - CHINESE ACAD OF SCI

Vision auxiliary device and method for automatic landing of unmanned aerial vehicle

Owner:CHANGCHUN INST OF OPTICS FINE MECHANICS & PHYSICS CHINESE ACAD OF SCI
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products