KR101274283B1 - Liver automatic segmentation method and system - Google Patents

Abstract

PURPOSE: A liver automatic segmentation method and a system thereof are provided to use a cumulative probability map to correct a liver area which is largely modified, thereby being utilized for a plan of a bio-segmental transplant surgery. CONSTITUTION: A liver volume definition unit(310) defines an optimal liver volume value. A liver volume segmentation unit(320) segments an initial liver volume based on a brightness value of the defined optimal liver volume value. A peripheral organ removal unit(330) uses the multiple cross sectional dissection information on the segmented liver volume to remove peripheral organs. A correction unit(340) corrects a boundary line by using a modification model of a liver, which is removed with the peripheral organs, and a cumulative probability map. The correction unit uses the modification model to smoothly correct the rear surface of the liver. [Reference numerals] (310) Interval volume definition unit; (320) Interval volume segmentation unit; (330) Peripheral organ removal unit; (340) Correction unit

Description

Liver Automatic Segmentation Method and System

The present invention relates to a method and a system for automatically dividing the liver, and more particularly, to a method and a system for automatically dividing the liver using multi-section anatomical information and a deformation model in a contrast-enhanced abdominal computed tomography image.

In order to detect local lesions of the liver, such as metastatic liver cancer or hepatocellular carcinoma, and to observe the surgical planning of the liver disease and the effects of radiotherapy and radiofrequency coagulation, hepatic segmentation is primarily performed on computed tomography (CT). It is a process that should be done. However, in the CT image, the liver is similar to the brightness value of surrounding organs such as the heart, stomach, and spleen, and since the shape and size of the liver vary depending on a person, there is a limit in automatically segmenting it.

The present invention was devised to solve the above-mentioned problem, and the anatomical information of the cross-sectional view is used to precisely divide the boundary portion of the peripheral organ where the liver and the brightness value are similar, and the variation is large using the deformation model and the probability accumulation map. An object of the present invention is to provide a method and a system for automatically dividing a liver by correcting the liver region.

Liver automatic segmentation method according to an embodiment of the present invention for achieving the above object comprises the steps of defining an optimal liver volume comprising a liver site; Dividing the initial inter-volume based on the brightness value with respect to the optimal inter-volume defined by the defining step; Removing peripheral organs using the cross-sectional anatomical information on the liver volume divided by the dividing step; And correcting the liver boundary line using the deformation model and the probability accumulation map for the liver volume from which the peripheral organs are removed by the peripheral organ removing step.

In the defining step, in order to extract the optimal volume of the liver including the liver, the upper boundary is defined by the lung and heart segmentation technique based on the brightness value in the coronal cross-section, and the lower boundary is determined by the pelvic bone and ribs recognition technique. In the axial sectional view, the side boundary can be defined through the rib recognition technique.

In the dividing step, an anisotropic divergence filtering may be used to highlight an area having brightness information of the liver, and the initial liver volume may be divided using an adaptively calculated threshold value.

Here, it is preferable to remove using a circular morpheme in order to prevent the hole inside the liver.

In the peripheral organ removal step, in order to remove peripheral organs similar in brightness to the liver, small tissues and kidneys are removed from the axial sectional view through a morphological opening operation and a connection area labeling technique, and the connection area labeling in the coronal cross sectional view. Techniques can remove the spleen and stomach.

In the correcting step, the rear surface of the liver may be smoothly corrected using the deformed model, and the left lobe region lost in the peripheral organ removal may be detected.

In this case, in order to solve a problem in which a partial region of the left lobe is recognized as a peripheral organ and is lost in the case of large variance data, a liver having a large shape variation using the probability cumulative map calculated through the segmentation information in the coronal cross section. The border can be corrected.

Liver automatic splitting system according to an embodiment of the present invention for achieving the above object, liver volume defining unit for automatically calculating the optimal liver volume; An inter volume dividing unit dividing an initial inter volume based on a brightness value with respect to an optimal inter volume defined by the inter volume defining unit; Peripheral organ removal unit for removing the peripheral organs using the multi-sectional view anatomical information for the liver volume divided by the liver volume divider; And a correction unit for correcting the liver boundary line using the deformation model and the probability accumulation map for the liver volume from which the peripheral organs are removed by the peripheral organ removing unit.

The liver volume defining unit may define an upper boundary of the optimal liver volume including the liver part through a brightness value-based lung and heart segmentation technique in the coronal cross-sectional view to extract the optimal liver volume including the liver part, and the pelvis. The bone and rib recognition technique defines the lower boundary of the optimal liver volume including the liver, and the rib section recognition technique defines the lateral boundary of the optimal liver volume including the liver.

The liver volume dividing unit may emphasize the region having the brightness information of the liver using anisotropic divergence filtering and divide the initial liver volume using an adaptively calculated threshold value.

In this case, the liver volume divider is preferably removed using a 7 × 7 morpheme to prevent holes in the liver.

The peripheral organ removal unit removes peripheral tissues and kidneys through a morphological opening operation and a connection area labeling technique in an axial sectional view to remove the peripheral organs similar in brightness to the liver, and a connection area labeling technique in a coronal cross sectional view. The spleen and stomach can be removed.

The correction unit may smoothly correct the posterior surface of the liver using the deformed model, and detect the left lobe region lost in the peripheral organ removal step.

In addition, in order to solve a problem in which a partial region of the left lobe is recognized as a peripheral organ and is lost in the case of data having a large variation, the correction unit may use a shape variation using the probability accumulation map calculated through the segmentation information in the coronal cross-sectional view. It is desirable to correct the boundary line between the large livers.

The automatic liver segmentation method according to an embodiment of the present invention can be used for diagnosing diseases such as metastatic liver cancer, liver cancer treatment using radiation therapy, and biofragmentation planning, and not only liver data but also brightness values of objects to be segmented It can be applied when it is desired to accurately divide objects having similar and various shapes.

1 is a diagram showing the limits of liver segmentation in contrast-enhanced abdominal CT images, (a) shows an example of a liver CT image in the axial sectional view, (b) shows an example of a histogram of brightness values of the liver and peripheral organs, (c) is a diagram showing an example of the diversity of liver shape and size.
2 is a flowchart illustrating a method for automatically dividing liver according to an embodiment of the present invention.
3 is a diagram illustrating an automatic liver volume estimation process.
4 is a diagram showing the results of each step of the initial intersegmentation scheme.
5 is a view showing the step-by-step results of the peripheral organ removal technique.
6 is a view showing the step-by-step results of the deformation model.
7 is a diagram showing the results of each step of the liver boundary correction method using a probability accumulation map.
8 is a diagram illustrating a liver segmentation result using an automatic liver segmentation method according to the present invention.
9 is a view showing a step-by-step performance of the liver splitting method according to the present invention.
FIG. 10 is a diagram illustrating a result of measuring an average distance difference between divided liver-shaped surfaces. FIG.
11 is a diagram illustrating an example of a result of measuring a redundancy ratio between a manual division volume and an automatic division volume.
12 is a diagram illustrating another example of a result of measuring a redundancy ratio between manual division and automatic division volume.
FIG. 13 is a diagram schematically showing a liver autodividing system according to an embodiment of the present invention. FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, a detailed description of known techniques well known to those skilled in the art may be omitted.

In describing the constituent elements of the present invention, the same reference numerals may be given to constituent elements having the same name, and the same reference numerals may be given thereto even though they are different from each other. However, even in such a case, it does not mean that the corresponding components have different functions according to the embodiments, or does not mean that they have the same functions in different embodiments, and the functions of the respective components may be implemented. Judgment should be made based on the description of each component in the example.

In the following description of the embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

In addition, in describing the component of this invention, terms, such as 1st, 2nd, A, B, (a), (b), can be used. These terms are intended to distinguish the constituent elements from other constituent elements, and the terms do not limit the nature, order or order of the constituent elements. When a component is described as being "connected", "coupled", or "connected" to another component, the component may be directly connected or connected to the other component, Quot; may be "connected," "coupled," or "connected. &Quot;

Conventional techniques for automatically segmenting the liver from CT images can be divided into brightness and gradient based segmentation, classification based segmentation, and shape information based segmentation.

First, as a brightness and slope-based segmentation technique, Massoptier, etc., is an area in which the average of the brightness values of each region divided into equally sized regions within the liver brightness value and the standard deviation is less than a certain standard deviation is obtained. After defining it as a candidate region, we proposed a method of segmenting the boundary line by applying a gradient vector filed (GVF) model. In addition, Paola et al. Proposed a liver region segmentation technique by detecting initial liver region using brightness information and performing morphological operations to remove peripheral organs with similar liver and brightness values. In addition, Amir et al. Proposed a technique for segmenting the initial liver region by extracting the liver region using anatomical information and brightness information, and removing surrounding organs using split thresholding. In addition, Liu et al. Remove noise edges that are extracted together using template mask information when performing Canny edge detection to extract the boundary line. At this time, the extracted edge information is calculated as a connected outline through morphological operations and divided by converging the GVF model. In addition, Rusko et al., Based on the segmentation of the liver region using the brightness value information, align the segmented liver region in the venous (portal venuos phase) image and the arterial phase image through affine transformation. We propose a technique for dividing overlapping regions into final interregions. At this time, peripheral organs such as the spleen, pancreas and vena cava, which appear similar to the brightness of the liver in the venous image, appear brighter than the liver in the arterial image. Such a segmentation technique based on the brightness value or the slope information has a problem that it is difficult to accurately segment the boundary information at the region having the brightness value similar to the surrounding tissue.

As a classification-based segmentation technique, Furukawa et al. Classify each voxel for the liver, heart, and spleen using a maximum a posteriori probability prediction technique, and use the level-set technique to determine the boundaries of the liver. A method of dividing by correction is proposed. In addition, Van et al. Classified the liver region and other regions using the k-NN classification technique based on the brightness value information, and then corrected the boundary line by applying morphological operations in the region classified as liver. Susomboon et al. Also calculate and classify the distribution of brightness values for the liver and surrounding organs based on the expectation maximization algorithm, and based on the slice with the largest texture information in each sliced region in each slice. Segmentation is proposed by applying the growth method. In addition, Frelman et al. Classify each voxel into four regions using prior information on the brightness range information of the liver and surrounding organs to detect candidate liver regions, and then remove the internal holes of the liver using morphological operations. The geodesic active contour model was used to divide the boundaries of the divided liver regions as initial outlines. Since the classification method based on the classification method is classified based on the brightness value information, it is difficult to accurately divide the adjacent organs having similar liver and brightness values in the adjacent region and takes a long time to optimize.

As a shape information-based segmentation technique, Kainmuller et al. Proposed a segmentation technique by defining a statistical shape model calculated from a training set and fitting it to the shape that optimizes the heuristic technique in the image to be segmented. Heimann et al. Also segmented each shape by moving the vertices of the shape model and fitting the shape model to a location that optimizes the k-NN classification technique in order to fit the statistical shape model. In addition, Jiamin et al. Segmented initial shapes using an active shape model based on shape information, and then applied live-wire techniques based on brightness values to improve local accuracy. A method of dividing the height is proposed. In addition, Zhang et al. Estimated the initial position of the average model through the GHT (generated hough transform) method to minimize the segmentation results when the statistical shape model based on shape information is affected by the initial position of the average model. A segmentation technique is proposed through the graph-cut technique. In addition, Wimmer et al. Divided the level-set method by additionally designing an expression representing the region and shape information of the liver in the general level-set method. In addition, Linguraru et al. Defined the average atlas calculated from the training set and transformed it by applying affine transformation and b-spline non-rigid matching technique to the data to be split, and then geodesiced the boundary of the modified atlas as the initial outline. Finally, we propose a final segmentation technique using the dynamic outline technique. This shape information-based segmentation technique can be segmented relatively accurately in areas with similar brightness and surroundings, but the size and shape of the liver appear differently depending on the person, and in particular, because the left lobe variation is large, it is divided by the shape information of the training set. It has a limit that is greatly affected by accuracy.

In the present invention, the liver segmentation technique is accurately divided at the boundary region of the surrounding organ where the liver and the brightness value are similar by using the anatomical information of the multi-sectional view, and the segmentation is corrected by correcting the liver region with large variation using the deformation model and the probability accumulation map. Suggest.

Rib and pelvic bone recognition using brightness information and cardiac removal technique minimizes liver partitioning time by calculating the optimal liver volume, and peripheral organ removal using multi-section anatomical information Remove peripheral organs such as kidneys, stomach and spleen. At this time, the embodiment of the present invention proposes a deformation model using the directional information to smoothly correct the rear surface of the liver in which the peripheral organs are located, and solves a problem in which the left lobe of the liver is recognized as the peripheral organ and is lost. In addition, by suggesting a method for calibrating the liver boundary using a probability cumulative map, the shape of the liver is varied evenly.

FIG. 2 is a flowchart illustrating a method for automatically dividing a liver according to an exemplary embodiment of the present invention, and illustrating an automatic segmentation technique using anatomical information and a deformation model of a multi-section view.

Referring to FIG. 2, first, in order to extract an optimal liver volume including a liver region, the upper boundary is defined by the lung and heart segmentation technique based on the brightness value in the coronal cross section, and the lower portion is determined through the pelvic bone and rib recognition technique. A boundary is defined and a side boundary is defined through the rib recognition technique in the axial cross section (S210).

Second, anisotropic divergence filtering is used to emphasize the region having the brightness information of the liver, and the initial volume is partitioned using an adaptively calculated threshold value (S220). At this time, in order to prevent the hole in the liver to remove using a 7x7 circular morpheme.

Third, in order to remove peripheral organs with similar liver and brightness values, small tissues and kidneys are removed from the axial cross section through morphological opening and linkage labeling, and spleen and stomach from the coronal cross section. Remove (S230).

Finally, the rear surface of the liver is gently corrected using the deformation model, and the left lobe region lost in the peripheral organ removal step is detected (S240). In this case, in order to solve a problem in which a partial region of the left lobe is recognized as a peripheral organ in the case of large variation, the boundary line between the large shape variation is corrected by using a probability accumulation map calculated through the segmentation information in the coronal cross section.

First, the process of automatically calculating the optimal liver volume in step S210 will be described.

The abdominal CT image of the liver is taken from the bottom of the lungs and heart to the pelvis. Therefore, by automatically defining the liver volume that optimally includes the liver area, unnecessary peripheral information is removed and the calculation time of the automatic liver segmentation is minimized.

The optimal liver volume consists of defining the upper boundary by separating the lung and the heart located at the upper part of the liver, defining the lower boundary using the position information of the ribs and pelvis, and defining the lateral boundary using the position information of the ribs. In the upper boundary defining step, the lung is divided by a threshold technique using the brightness value (-450HU: Hounsfield Unit) of the lung in a vertical coronal cross-sectional view of the axial cross-section as shown in (a) of FIG. 3, and (b) of FIG. 3. As shown below, the point where the average brightness profile value calculated using the 1 × 3 mask in the cardiac candidate boundary region is the minimum is defined as the boundary point between the liver and the heart. As shown in (c) of FIG. 3, the candidate boundary region of the heart is defined using location information of the leftmost lowermost point of the divided right lung and the rightmost lowermost point of the left lung, and the calculated boundary points are defined as b-splines. By defining the upper boundary of the volume between the optimal liver. The lower boundary definition step is the lowest point of the rib and the top of the pelvis through the rib and pelvic bone recognition technique using the bone brightness value (200HU) information and location information in the middle slice of the coronal cross-sectional view as shown in FIG. Define as 3/4 ratio point of the point. The side boundary defining step is defined as the inside of the outline connecting the center points of the ribs detected in the middle slice of the axial sectional view as b-splines as shown in FIG.

The optimal liver volume estimation using anatomical information is to separate the cardiac region, which is similar to the brightness of liver in the upper boundary definition step, using coronal cross-sectional image information to prevent leakage to the heart and to define the lower boundary. By estimating the pelvic bones at the upper part of the pelvis, we can reduce unnecessary segmentation and prevent bleeding out of the ribs.

Next, a process of initial volume division based on the brightness value of step S220 will be described.

The brightness value distribution of the liver region in the calculated optimal liver volume can be distinguished from the region indicated by high brightness values, such as some peripheral regions and bones, as shown in FIG. An initial volume partitioning technique for extracting candidate regions is performed.

The initial hepatic region segmentation technique using brightness information consists of anisotropic divergence filtering, adaptive threshold technique, and internal hole elimination. Anisotropic divergence filtering, which is performed to preserve boundary information and remove internal noise, smoothes the area of the slope value smaller than the slope defined in consideration of the directional information of the brightness value and preserves the area of the large slope value. It is calculated through Equation 1 and shows a result as shown in FIG.

In this case, fc represents a brightness value calculated at time t, and c (| f |) is a coefficient representing spreading information of an image and adjusts a smoothing degree according to a user-defined K parameter.

In the adaptive threshold step, the mean brightness value μ and the standard deviation s of the liver are calculated by histogram analysis in the middle slice of the axial cross section to estimate the range of brightness values of the liver. In general, the global threshold technique used to solve the problem that may include the costal cartilage region, as shown in Figure 4 (c), it is solved by removing the border region of the liver and the costal cartilage appearing with a lower brightness value than the surroundings . To this end, a threshold value is adaptively calculated according to an area as shown in FIG. At this time, the threshold value range in the R1 region is [μ-σ, μ + σ], and the lower threshold in the R2 region uses a brightness value having a ratio of 10% of the R1 region threshold value to remove the boundary region of the liver and peripheral organs. The upper threshold value is used the same as the upper threshold value of R1.

The hole elimination process defines a circular mask having a size to prevent holes caused by the removal of blood vessels when the adaptive threshold technique is performed, and the brightness value corresponding to the boundary point of the mask is a hole with a brightness value corresponding to the center point of the circular mask. If all of these have a range of brightness values in the liver, the holes are determined by replacing the average brightness value of the liver with the hole inside the liver.

Anisotropic divergence filtering preserves liver border information while preserving liver boundary information, and the anatomical information-based adaptive threshold technique enables accurate segmentation of the liver region even on the ribbone where the liver and the brightness are similar.

Next, the peripheral organ removal step using the multi-section anatomy information of step S230 will be described.

In order to solve the limitations of the detection of peripheral tissues, which are represented by brightness values similar to the initial segmentation time using the brightness value information, and peripheral organs such as the kidney, stomach and spleen, we propose a technique for removing peripheral organs using anatomical information in multiple sections. do. In the axial cross section, the liver and surrounding organs have similar brightness information and anatomical information appearing adjacent to each other. To remove peripheral organs efficiently.

Peripheral organ removal using anatomical information consists of the removal of peripheral tissues and kidneys in the axial cross-section and the removal of the stomach and spleen in the coronal cross-section. The process of removing small tissues and kidneys around the liver area removes the kidneys by opening the operator in the axial cross section and extracting the largest liver area through the junction labeling technique.

Since the spleen and stomach appear adjacent to the liver in the axial cross section, they appear separate from the liver in the coronal section, so that the splice and stomach area can be efficiently removed by applying the joint labeling technique. 5 (a) and 5 (b) show an image in which the liver and peripheral organs are adjacent to each other and a vertical cross-sectional view thereof. 5C to 5G show the removal of peripheral organs in the axial cross section and the tubular cross section.

Peripheral tissues and organs similar to liver brightness values are efficiently removed using anatomical information in the axial and coronal sections.

Finally, the final volume splitting process using the deformation model and the probability accumulation map of step S240 will be described.

Since the shape and size of the liver varies from person to person, there is a limitation in that some areas are lost in the left lobe with large variation when removing the surrounding organs based on anatomical information. In order to solve this problem, we propose a technique to detect the left lobe region lost through the deformation model using the directional information of the slope, and to accurately segment even the large left lobe region through the boundary correction method using the probability map. In this case, the boundary correction step using the split difference accumulation map is additionally performed when the left lobe variation is determined to be large data.

The final inter-volume segmentation technique consists of deformation model fitting, left lobe detection, and boundary correction. In the step of fitting the deformed model, in order to estimate the position of the left lobe lost and smoothly correct the posterior surface of the liver, the uppermost vertex and the lowest vertex of the hepatic region divided in the coronal cross section as shown in FIG. An initial deformation model is defined, and each profile calculated at 15 pixel intervals calculates the feature point Fi corresponding to the liver surface using Equation 2. The computed feature points are redefined by the b-spline technique to redefine the deformation model, and are repeated to fit the interfacial surface.

In this case, the feature points calculated in the i th profile are calculated using the directional information of the labeling value L (k) of the k th index of the profile and the labeling value L (k + 1) of the k + 1 th index.

 The location of the lost left lobe is defined as the area in which the curvature of the fitted deformed model appears more than 10 pixels unchanged, and the area corresponding to the range of brightness values in the sagittal direction in the normal vector direction is detected.

In the boundary correction step, a probability accumulation map is calculated for data having a large shape variation and the boundary line is corrected. The large variance data is relatively difficult to define the boundary between the liver and the spleen because the relatively wide width of the liver and appear very close to the spleen as shown in (a) of FIG. As shown in FIG. 7B, the probability cumulative map detects an area in which the difference between the divided liver areas in the coronal cross-sectional view is 150 pixels or more and calculates the cumulative probability value. At this time, the maximum probability region calculated in the coronal cross-sectional view is shown as an area including the boundary region of the liver and spleen in the axial cross-sectional view as shown in FIG. 7C, and the boundary line of the liver is corrected by linear interpolation based on both endpoints of the maximum probability region. do. 7 (d) shows the step-by-step results of the liver boundary correction method using the probability accumulation map.

In the boundary correction step, the lateral information-based deformation model not only smoothly corrects the posterior surface of the liver, but also detects the position of the missing left lobe. It also divides correctly in the adjacent areas.

The data used in the experiments of the present invention was applied to 10 living liver donor data as an abdominal contrast liver image taken by SIEMENS CT. The image size is 512x512, the pixel size is 0.54 ~ 0.7mm, the slice spacing is 2.0mm, and the whole slice is 136 ~ 229. In order to evaluate the accuracy of the automatic liver segmentation method according to the embodiment of the present invention, the average symmetrical surface distance and the overlapping volume error between the visual evaluation and the manual segmentation of the clinician and the automatic segmentation method between the liver were measured.

8 is a result of segmentation of data with large liver shape variation, and each row shows two-dimensional segmentation results and three-dimensional surface rendering results at the upper, middle, and lower ends of the liver. The division result is shown. The two-dimensional segmentation results show that the liver and the bright, such as the heart, kidney, stomach, and spleen, are correctly divided at the borderline of the surrounding organs, which have similar values.

9 is a result of performing the proposed method step by step in the data of the large left lobe, each column represents the top, middle, bottom surface. Using the anatomical information of the axial and coronal sections together in the initial liver segmentation results, the surrounding organs could be removed efficiently. Could be solved through. In addition, it can be seen that the mutation was correctly divided even in the large left lobe.

The average symmetrical surface distance during the evaluation of accuracy can be measured by the results of the manual division of two clinical experts and the surface distance of the liver shape, which is the result of automatic division using the automatic division method according to the embodiment of the present invention, as shown in Equation 3. have.

At this time, S (A) represents a vertex on the surface automatically divided, S (B) represents a vertex on the surface that has been manually divided, and N (·) represents the number of vertices constituting the divided surface. D is Euclidean distance.

10 is a graph showing a box plot of the difference in average distances of the divided liver surfaces for the ten experimental data used in the present experiment.

The experimental results show that the mean distances and standard deviations between outlines for manual division 1, manual division 2, and automatic division are 0.28 ± 0.22 mm for manual division 1: automatic division, manual division 2: 0.31 ± 0.19 mm for automatic division, and manual division 1: Manual division 2 was measured at 0.27 ± 0.16 mm. The difference between manual division was measured to be smaller than 0.3mm, and the difference between automatic division and manual division was also measured to be smaller than the pixel size within 0.3mm.

The duplicate volume error during the accuracy evaluation measures the error of the overlap ratio between the volume divided by the two clinical experts and the result of applying the proposed method as shown in Equation 4.

In this case, A is the number of voxels of the automatic division result, and B is the number of voxels of the manual division result.

FIG. 11 is a graph showing box plots of measurement results of overlap ratio errors with respect to divided prostate volumes.

The average and standard deviation of the measured duplicated volume errors were 0.86 ± 0.45% in autodivision: manual division 1, autodivision: 0.86 ± 0.45% in manual division 2, and manual division 1: 0.73 ± 0.33% in manual division 2. . As a result of measuring duplicated volume error, it shows error rate within 1%.

The experiment was performed on a PC with an Intel Core i7 920 2.6GHZ CPU and 4.0GB memory. FIG. 12 shows the execution time by dividing the optimal liver volume definition step of the proposed method, the initial liver segmentation step using the brightness value information, the peripheral organ removal step using the multi-section information, and the boundary line correction step using the deformation model and the probability accumulation map. Is the result of measuring. According to an embodiment of the present invention, the average execution time of the automatic liver segmentation method is 46 seconds, 2.26 seconds in the optimal boundary region calculation step, 22.74 seconds in the initial liver segmentation step, 7.20 seconds in the peripheral organ removal step and final liver segmentation step. It took 14.23 seconds. The optimal boundary area estimation step required the shortest execution time of around 2 seconds on average, and the initial interdivision phase required the most execution time due to the performance of anisotropic divergence filtering.

In an embodiment of the present invention, an anatomical information of multiple sections and an automatic segmentation method based on deformation model were developed in an abdominal contrast-enhanced CT image. The optimal liver volume definition technique was used to remove the heart with a similar brightness to the liver, and the initial segmentation technique effectively segmented only the candidate liver region in the intercostal area that appeared adjacent to the liver region. At this time, the problem that the peripheral organs, which are similar to the brightness of the liver, is detected as the initial liver region could be solved through the peripheral organ removal technique using anatomical information of the multi-section. In addition, the proposed boundary boundary correction step to solve the limitations of the loss of some areas of the peripheral organ removal technique was smoothly corrected the posterior surface of the liver through the deformation model, The location was detected and tightly split. In particular, the left lobe of the liver, which appears in various anatomical shapes, could be accurately segmented by correcting the borderline of the liver using a split difference accumulation map. As a result of evaluating the accuracy of the hepatic autodividing method according to an embodiment of the present invention, the average symmetrical surface distance difference was 0.31 ± 0.19mm, which was smaller than one pixel size, and the overlapping volume error was 0.86 ± 0.45, which is less than 1%. It became. In addition, the execution time of the automatic segmentation by applying the liver automatic segmentation method according to an embodiment of the present invention showed a fast convergence within an average of 46 seconds. The method for automatically dividing the liver according to an embodiment of the present invention can be used for diagnosing diseases such as metastatic liver cancer, treating liver cancer using radiation therapy, and planning for biofragmentation. It can be applied when it is desired to accurately divide objects having similar and various shapes.

FIG. 13 is a diagram schematically showing a liver autodividing system according to an embodiment of the present invention. FIG.

Referring to FIG. 13, the liver auto splitting system may include a liver volume definer 310, a liver volume divider 320, a peripheral organ definer 330, and a corrector 340.

Here, the liver volume defining unit 310 performs the function of step S210 of the liver autodividing method according to the embodiment of the present invention as described above, the liver volume divider 320 performs the function of step S220, The peripheral organ defining unit 330 may perform the function of step S230, and the correction unit 340 may perform the function of step 240. Therefore, detailed descriptions of operations and functions of the liver volume defining unit 310, the liver volume dividing unit 320, the peripheral organ defining unit 330, and the correcting unit 340 will be omitted.

 In the above description, all elements constituting the embodiments of the present invention are described as being combined or operating in combination, but the present invention is not necessarily limited to the embodiments. In other words, within the scope of the present invention, all of the components may be selectively operated in combination with one or more. In addition, although all of the components may be implemented in one independent hardware, each or all of the components may be selectively combined to perform some or all functions combined in one or a plurality of hardware. It may be implemented as a computer program having a. In addition, such a computer program is stored in a computer readable media such as a USB memory, a CD disk, a flash memory, and the like, and is read and executed by a computer, thereby implementing an embodiment of the present invention. As the storage medium of the computer program, a magnetic recording medium, an optical recording medium, a carrier wave medium, or the like may be included.

Furthermore, all terms including technical or scientific terms have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined in the Detailed Description. Terms used generally, such as terms defined in a dictionary, should be interpreted to coincide with the contextual meaning of the related art, and shall not be interpreted in an ideal or excessively formal sense unless explicitly defined in the present invention.

 The above description is merely illustrative of the technical idea of the present invention, and those skilled in the art to which the present invention pertains may various modifications and changes without departing from the essential characteristics of the present invention. In addition, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. Therefore, the protection scope of the present invention should be interpreted by the claims, and all technical ideas within the equivalent scope will be construed as being included in the scope of the present invention.

310: inter volume definition unit 320: inter volume division unit
330: peripheral organ removal unit 340: correction unit

Claims (14)

Defining an optimal liver volume;
Dividing the initial inter-volume based on the brightness value with respect to the optimal inter-volume defined by the defining step;
Removing peripheral organs using the cross-sectional anatomical information on the liver volume divided by the dividing step; And
And correcting a liver boundary using a deformation model and a probability accumulation map for the liver volume from which the peripheral organs are removed by the peripheral organ removing step. The method of claim 1, wherein the defining step,
In order to extract the optimal volume of liver including the liver area, the upper boundary is defined by the brightness-based lung and heart segmentation technique in the coronal section, the lower boundary is defined by the pelvic and rib recognition technique, Liver segmentation method characterized in that by defining the lateral boundary through the rib recognition technique. The method of claim 1, wherein the dividing step,
And anisotropic divergence filtering to emphasize the region having the brightness information of the liver and divide the initial liver volume using an adaptively calculated threshold value. The method of claim 3, wherein
Automatic liver segmentation method characterized in that to remove by using a 7x7 morpheme in order to prevent the hole in the liver. According to claim 1, The peripheral organ removal step,
In order to remove peripheral organs similar in brightness to the liver, small tissues and kidneys are removed by morphological opening operation and connection area labeling technique in the axial cross section, and spleen and stomach in the coronal cross section through the connection region labeling technique. Liver autodividing method characterized in that the removal. The method of claim 1, wherein the correcting step,
Smoothing the rear surface of the liver using the deformed model, and the liver segmentation method characterized in that for detecting the left lobe region lost in the peripheral organ removal step. The method according to claim 6,
In order to solve a problem in which a partial region of the left lobe is recognized as a peripheral organ and lost in case of large variance data, the boundary line of the liver having a large shape variance is corrected using the probability accumulation map calculated through the segmentation information in the coronal cross section. Liver splitting method characterized in that. Liver volume defining unit for automatically calculating the optimal liver volume to define;
An inter volume dividing unit dividing an initial inter volume based on a brightness value with respect to an optimal inter volume defined by the inter volume defining unit;
Peripheral organ removal unit for removing the peripheral organs using the multi-sectional view anatomical information for the liver volume divided by the liver volume divider; And
And a correction unit for correcting a liver boundary using a deformation model and a probability accumulation map with respect to the liver volume from which the peripheral organs are removed by the peripheral organ removing unit. The method of claim 8, wherein the liver volume defining unit,
In order to extract the optimal liver volume including the liver, the upper boundary of the optimal liver volume is defined by the lung and heart segmentation technique based on the brightness value in the coronal section and the optimal liver volume by the pelvic bone and ribs recognition technique. Auto-segmentation system, characterized in that it defines the lower boundary, the side boundary of the optimal liver volume through the rib recognition method in the axial cross-sectional view. The method of claim 8, wherein the inter-volume division,
And anisotropic divergence filtering to emphasize the region having the brightness information of the liver and divide the initial liver volume using an adaptively calculated threshold value. The method of claim 10,
Liver automatic splitting system, characterized in that for removing the hole in the liver using a circular morpheme. The method of claim 8, wherein the peripheral organ removal unit,
In order to remove the peripheral organs with similar brightness values to the liver, peripheral tissues and kidneys are removed from the axial cross-section through a morphological opening operation and a joint labeling technique, and a spleen and a stomach are removed from the coronal profile through a joint labeling technique. Liver automatic splitting system, characterized in that. The method of claim 8, wherein the correction unit,
Using the deformed model to gently correct the back surface of the liver, the liver segmentation system characterized in that for detecting the left lobe region lost in the peripheral organ removal step. The method of claim 13,
In order to solve a problem in which a partial region of the left lobe is recognized as a peripheral organ and lost in case of large variance data, the boundary line of the liver having a large shape variance is corrected using the probability accumulation map calculated through the segmentation information in the coronal cross section. Liver automatic splitting system, characterized in that.

Images