KR101153627B1 - Image analysys system of cellular tissue and method thereof - Google Patents

Abstract

An image analysis system of cell tissue is disclosed. The disclosed image analysis system forms a contour topography of beam intensities attenuated by cellular tissue from a level of spacing between contours to a level of dense spacing between contours in sequence, and the inclusion relationship between the topology and the topology And an extracting unit for extracting a region of interest by calculating a nesting depth from the inclusion relationship between the derived imaging unit and the topology.

Topology, Image, Tumor, X-ray

Description

IMAGE ANALYSYS SYSTEM OF CELLULAR TISSUE AND METHOD THEREOF

The present invention relates to an image analysis system and method for disease diagnosis, and more particularly, to an image analysis system and method for disease diagnosis using a topology.

Medical image analysis devices for diagnosing diseases include an X-ray device, a magnetic resonance imaging (MRI), and an ultrasound device.

X-ray apparatus refers to a device that analyzes an image in which electromagnetic waves penetrate human tissue and then is printed on a film. Recently, an X-ray apparatus has evolved and developed into a digital x-ray apparatus using TFT.

MRI is a diagnostic device that uses a magnetic field to create a picture of the structure inside the body, using the pulse of electromagnetic waves to create an image based on the principle that the hydrogen atoms in the body react with the magnetic field and electromagnetic waves. MRI images can combine multiple thin slices of organs, bodies and parts of cells into three-dimensional images.

 An ultrasound device is an apparatus for analyzing target cell tissues by imaging ultrasound waves reflected from cell tissues using ultrasound waves having an audible frequency of 20 kHz.

The medical image analyzing apparatus described above may form an image using various wavelengths to diagnose various diseases and analyze the image to distinguish between cell tissues showing abnormal symptoms. However, the image of the conventional image analysis device does not have a high resolution, so there is an inconvenience that the medical personnel must manually check and distinguish the abnormal cell tissue. In other words, the image presented by the image analysis device needs to show as many suspected tumor candidates as possible to medical personnel, such as malignant tumors and benign tumors, and even when it is necessary to distinguish specific substances, it is necessary to present the relevant candidates to the maximum possible.

The present invention has been made to improve the prior art as described above, and an object of the present invention is to provide an image analysis system and method that can display a cell tissue with a high resolution using a topology.

Still another object of the present invention is to provide an image analysis system and apparatus capable of extracting a region of interest with high reliability by using topology and matching.

In order to achieve the above object and to solve the problems of the prior art, according to one embodiment of the present invention, the contour topography of the beam intensity attenuated by the cell tissue is densely spaced between contours sequentially from the coarse level between the contour lines An imaging unit which forms up to one level and derives a containment relationship between the topologies from the change of the topologies; And an extraction unit for extracting a region of interest by calculating a nesting depth from the inclusion relationship between the topologies.

According to another embodiment of the present invention, the method comprises: forming a contour topology of beam intensity attenuated by cellular tissue from a level where spacing between contours is coarse to a level where the spacing between contours is denser; Deriving a containment relationship between the topologies from the change of the topologies; And extracting a region of interest by calculating a nesting depth from the inclusion relationship between the topologies.

According to one embodiment of the present invention, the cell tissue can be displayed with high resolution.

In addition, according to an embodiment of the present invention, the ROI may be extracted with high reliability.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a schematic view showing the structure of an image analysis system of cell tissue according to an embodiment of the present invention. The image analysis system 100 of the cell tissue according to the exemplary embodiment of the present invention intends to extract a region of interest of the target cell tissue. Here, the region of interest may be a target cell tissue to be searched for a medical diagnosis, such as a tissue that is denser (or different) than a surrounding cell tissue such as a tumor, or a material different from other substances such as bone.

Referring to FIG. 1, the image analysis system 100 of a cell tissue according to an exemplary embodiment of the present invention may sequentially remove the contour topologies of beam intensities attenuated by the cell tissue. An imaging unit 101 which forms a dense level and derives an inclusion relationship between the topologies from the change of the topology, and an extraction unit for extracting the region of interest by calculating the nesting depth from the inclusion relationship between the topologies ( 102).

The beam irradiated to the cell tissue may use various light sources such as X-rays, MRI electromagnetic waves, and ultrasonic waves, and the image analysis system of the cell tissue of the present invention may be applied to any light source.

The imaging unit 101 of the image analysis system 100 of the cell tissue according to an embodiment of the present invention changes the topology according to the intensity of the beam from the image formed by the beam once penetrating, reflecting, or refracting the cell tissue. Analyze

The image I may be represented by Equation 1 when Ω represents a domain of the image I (x).

)에 있어 표면 입면도(elevation)는 다음의 수학식과 같이 나타낼 수 있다. In this case, the strength (Φ⊂

The surface elevation can be expressed by the following equation.

)가 이미지 내 주어진 강도보다 더 높은 강도를 가진 콤팩트한 연결 영역인 경우 다음과 같이 정의된다.R (t) ⊂Ω (

) Is a compact connection area with a higher intensity than a given intensity in the image.

The contour line Γ (t) of the image at a given level t is defined as the boundary of the region R (t).

Since region R is a compact connection region, its boundary line contour Γ is represented by a simple closed curve. Simple closed curves do not intersect at any point and appear similar to a single circle and are identical to the Jordan curve. Topographical representation is obtained from a set of contours at a number of distinct partial values throughout the intensity range of the image. This display may be referred to as a contour map. The contour map M _N (I) with _N quantizations for image I is given as follows.

는 최소 강도

와 최대 강도

사이에 주어진 이미지 (I)의 강도 범위인 강도

에서의 등고선이다. M은 정량화 레벨의 수로서 스케일 인수(scale factor)이다. N의 선택은 검출되는 특징의 특성(specificity)과 관계가 있다. 큰 N은 조밀한(fine) 스케일 특징을 특징화하는 등고선 맵을 형성하 고, 작은 N은 성긴(coarse) 스케일 특징을 가지는 등고선 맵을 형성한다. 등고선 맵의 다중 스케일 특성은 성긴 스케일부터 조밀 스케일까지의 계층적 방식으로 분석되는 이미지 분석 특징에 대해 분해능을 강화시킬 수 있다.here,

Is the minimum strength

And maximum strength

Intensity, which is the intensity range of the image (I) given in between

Is the contour line at. M is the scale factor as the number of quantification levels. The choice of N is related to the specificity of the feature to be detected. Large N forms a contour map that characterizes fine scale features, while small N forms a contour map with coarse scale features. The multi-scale nature of the contour map can enhance resolution for image analysis features that are analyzed in a hierarchical manner from coarse to dense scale.

FIG. 2A shows the contour line according to the surface model of the image and the intensity value quantified, and FIG. 2B shows the contour line superimposed on the image quantified to the set level. In order to display contours on a three-dimensional image surface, an image having a series of horizontal planes can be sliced and displayed.

으로 강도 레벨이 설정되어 이미지가 생성된다. 설정된 강도값에 대한 등고선 추적 과정에서, 그 연속성은 8개의 이웃 픽셀을 참조할 수 있다. 등고선에 대한 연속 곡선을 얻기 위하여, 이산(discrete) 이미지 모델을 연속 이미지로 변형하는 리샘플링이 필요하고 이산 보간법(bilinear interpolation)이 사용될 수 있다.2A and 2B, the image surface is modeled in a semi-ellipse shape. The curve pattern of the contour lines is a threshold value.

The intensity level is set so that an image is generated. In the contour tracking process for the set intensity value, the continuity may refer to eight neighboring pixels. In order to obtain continuous curves for the contours, resampling to transform the discrete image model into continuous images is necessary and bilinear interpolation can be used.

Contour maps are effective for extracting topology and geometric information from images. The pattern of contours is distinguished according to the geometry and topology properties of the corresponding image content. One of the most important contour patterns is the nesting pattern. Contours of salient objects generally form a dense nesting shape. Often, contour shapes of objects appear similar to object contours. Dense nesting patterns of contours often form similar to the contours of objects distinguished by high contrast.

Along with the nesting pattern, the gradient of the contour can be selected as a useful feature for the feature of the projecting object. Gradually decreasing intensity as the intensity crosses the contour of an object forms a prominent contour pattern in a dense scale contour map. A change in single intensity across the object contour causes a single change in the contour pattern.

Dense nesting contour patterns can be found in the region of abrupt changes in intensity gradients. A sparse contour pattern can be found in the progressive change area of the intensity gradient. Closely located contours are produced in sharp slope regions, and widely spaced contours are produced in smooth slope regions. Thus, the dense nesting pattern of the contours represents the distinct boundary of the object with a sudden change in strength and terrain from the object to its local background. In the image analysis system according to an embodiment of the present invention, the entire contour is analyzed by topology and geometry in order to select the contour from which the ROI is extracted.

An advantage of the image analysis system according to an embodiment of the present invention using a topology is that the topology display gives shape, size, and position information about an object in a context, from which a region of interest including a tumor in an image, such as an x-ray, is classified and It is an important feature to detect. One useful feature of the contour map is that it is invariant even when the intensity changes. Images of cellular tissue can be reliably analyzed under changing environmental conditions such as patients and x-ray machines. Image intensity values of cellular tissue can vary significantly even in the same patient. Indeed, the image analysis system and method according to an embodiment of the present invention using topologies provides robustness to change with affine invariance (affine). do.

The inclusion relationship between the topologies may be derived from the change of the contour topography of the beam intensity generated by the imaging unit 101 of the image analysis system 100 according to the exemplary embodiment.

)의 차수(degree)는

로 나타낸다. 함유 트리에서 그 차수에 따라 노드는 브랜칭 노드 또는 비브랜칭 노드로 분류될 수 있다. 하나 이상의 자(子) 노드를 가진 노드는 브랜칭 노드라고 하고 단지 하나의 자 노드만을 가지거나 자 노드가 없는 노드는 비브랜칭 노드라고 한다. 브랜칭 노드의 윤곽선을 브랜칭 윤곽선이라고 하고 바로 아래의 자 노드의 윤곽선을 베이스 윤곽선이라고 한다. 터미널 노드는 0 차수의 윤곽선을 가지게 된다. contour(

) Degree is

Respectively. Depending on their order in the containing tree, nodes may be classified as either branched nodes or non-branched nodes. Nodes with more than one child node are called branching nodes, and nodes with only one child node or no child nodes are called non-branching nodes. The contour of the branching node is called the branching contour and the contour of the child node immediately below is called the base contour. Terminal nodes will have zero-order contours.

FIG. 3A is a contour map showing the inclusion relationship, and FIG. 3B shows the containing tree corresponding to FIG. 3A.

Referring to FIG. 3A, the contour lines C1, C2, and C3 form an inclusion relationship, and the contour lines S1 and S2 form an inclusion relationship, and the contour line B includes them. The inclusion relationship between the contours is based on the Jordan curve theory. The Jordan curve on the Euclidean plane, consisting of exactly two connecting components, internal and external, has a common contour. In two distinct contours, the relationship of the two contours has a nesting pattern and is not connected to each other.

FIG. 3B forms the containing tree so as to correspond to FIG. 3A. The nodes of the containing tree represent outlines, and the parent node includes child nodes. The largest outline I of the containing tree represents the outline of the whole image. Contour B becomes a branching contour, and C3 and S2 represent the base contours with contours of varying topologies, respectively. And C1 and S1 become terminal contours.

In the imaging unit 101 of the image analysis system 100 according to the exemplary embodiment of the present invention, such contour topologies are formed from the level of the spacing between the contour lines to the level of the spacing between the contour lines in order.

FIG. 4A shows a contour topography formed at a sparse level, for example, a first level, and FIG. 4B shows a containing tree corresponding to FIG. 4A.

4A and 4B, the contours C4, C3, C2, and C1 are included in order from the largest contour C5 in turn, and another contour C6 is included in which the topology changes within the C5 contour. It can be seen from the corresponding containing tree that C5 becomes the base contour.

FIG. 5A shows a contour topography formed at a dense level, e.g., a second level, and FIG. 5B shows a containing tree corresponding to FIG. 5A.

5A and 5B, C4 ', C4, C3', C3, C2 ', C2, C1' and C1 contours are included inside the largest contour C5 and the topology changes within the C5 contour. Another topology is contour C6, with C6 'in C6. Contours C4, C3, C2, C1 and C6 will be included therein when forming contour topologies at the second level with contours already created at the first level. It can be seen from the containing tree that C5 becomes the base contour.

In the imaging unit 101 of the image analysis system 100 according to an exemplary embodiment of the present invention, the contour of the first node where the topology changes is set as the base closed curve of the ROI.

The extraction unit 101 of the image analysis system 100 according to an exemplary embodiment of the present invention calculates the nesting depth from the inclusion relationship of the topology and extracts the region of interest. The inclusion relationship of the topology can be presented as a containing tree as described above, and the extraction unit 101 of the image analysis system 100 according to an embodiment of the present invention includes a closed curve included in one topology as a nesting depth. Calculate the number of.

)인 윤곽선(

) 은 다음의 조건을 만족한다.The path Pi, j from the contour Γi to the contour Γj may be defined as the number of edges from the contour Γi to Γj. That is, the path (

)

) Satisfies the following conditions:

는 공간적 포함관계를 나타낸다. 경로(Pi,j)의 길이는 L(Pi,j)로 나타내고 이는 네스팅 깊이를 나타낸다. here,

Represents a spatial inclusion relationship. The length of the path Pi, j is represented by L (Pi, j) which represents the nesting depth.

)에 대해 최소 네스팅 깊이값

은 다음과 같이 정의된다.The base contour is regarded as the boundary of the protruding region, ie the region of interest, in the topology property. The extraction unit 102 of the image analysis system 100 according to an exemplary embodiment of the present invention calculates the depth of the nesting structure to measure the protrusion of the selected base contour. This is due to the assumption that the protruding areas generally form a dense nesting pattern. The base contour with a large nesting depth may mark the boundary of the protruding region. Nesting depth in the containing tree is defined by the number of contours in the base contour. That is, the length from the base contour to the terminal contour is given by the nesting depth. However, several rule trees may be included in the set contour. In this case, the minimum nesting depth value among the nesting depth values from the outline to the terminal outline is defined as the degree of protrusion. Base outline (

Nesting depth value for

Is defined as

)로부터 베이스 윤곽선(

)까지의 경로를 나타낸다.Where Pt, b is the terminal outline (

From the base contour (

Route to).

Therefore, in FIG. 4B, the minimum nesting depth of the child topology is 2 and the nesting depth of the parent topology is 2. In FIG. 5B, the minimum nesting depth of the magnetic topology becomes 3 and this value becomes the nesting depth of the parent topology.

As described above, the region of interest to be extracted needs to be matched in multiple images. In other words, in order to increase the reliability of the region of interest, it is necessary to take a plurality of images and match the region of interest for the same cell tissue as closely as possible in the multiple images.

The extraction unit 101 of the image analysis system 100 according to an embodiment of the present disclosure matches a plurality of regions of interest to the same cell tissue to match regions of interest having mutual similarities higher than a preset value. In this case, the matching may be performed in consideration of information such as the position, shape or size of the contour, which is a node, to improve the reliability of the ROI.

FIG. 6A is a photograph of a breast by X-ray, FIG. 6B is a drawing of a contour map at a coarse level, FIG. 6C is a drawing of a contour map at a medium level, and FIG. 6D is a drawing of a contour map at a dense level. to be. The more contours drawn from the coarse level to the dense level, the more precisely the area of interest can be derived. However, in some cases, the density of the level can be adjusted, and once the region of interest is derived to some extent, the level can be stopped.

7 is a flow chart showing an image analysis method of cell tissue according to an embodiment of the present invention.

Referring to FIG. 7, first, contour topologies of beam intensities attenuated by cell tissues are formed from coarse level lines to densely spaced intervals between contour lines (S1). When the coarse level is set as the first level, the contour topologies having nesting depths below the threshold value among the contour topologies formed in the first level are filtered first, and then, the interval between the contour lines is denser than the first level. The second level may be set to form a contour topology. At this time, the level can be set sequentially according to the density of the interval between the contour lines, the nesting depth can be calculated from the contour topology, and the process can be repeated until a predetermined region of interest can be derived.

Then, the inclusion relationship between the topologies is derived from the change of the topologies (S2). At this time, the branching node and the non-branching node are separated, and the first child node of the branching node becomes the base outline. One or more child contours may be included in the base contour.

Next, the nesting depth is calculated from the inclusion relationship between the topologies (S3). The nesting depth may be set to the number of child contours included in the base contour. In addition, when two or more child topologies exist in the base contour of the parent topology, the nesting depth of the child topology that has the smallest nesting depth is set as the nesting depth value of the topology.

Finally, the tissue of the topology having a predetermined nesting depth value is extracted into the region of interest (S4).

The image analysis method of cell tissues according to an embodiment of the present invention may be implemented in the form of program instructions that may be executed by various computer means, and may be recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above-described embodiments, which can be variously modified and modified by those skilled in the art to which the present invention pertains. Modifications are possible. Accordingly, the spirit of the present invention should be understood only in accordance with the following claims, and all equivalents or equivalent variations thereof are included in the scope of the present invention.

1 is a schematic view showing the structure of an image analysis system of cell tissue according to an embodiment of the present invention;

2A shows contours according to the surface model of an image and intensity values quantified,

2b is a view showing contours superimposed on an image quantified to a set level;

3A is a view showing a contour map showing a containment relationship;

3b shows a containing tree corresponding to FIG. 3a,

4a shows a contour topography formed at a sparse level, for example at a first level, between contours;

4b shows a containing tree corresponding to FIG. 4a, FIG.

5a shows a contour topography formed at a dense level, e.g., a second level, between contours;

5b shows a containing tree corresponding to FIG. 5a,

Figure 6a is a photograph taken of the breast by X-rays,

FIG. 6B illustrates a contour map at a sparse level corresponding to FIG. 6A; FIG.

FIG. 6C illustrates a contour map at an intermediate level corresponding to FIG. 6A.

FIG. 6D illustrates a contour map at a dense level corresponding to FIG. 6A

7 is a flow chart showing an image analysis method of cell tissue according to an embodiment of the present invention.

Claims (21)

An imaging unit for forming a plurality of contour topologies of beam intensity attenuated by cellular tissues, ranging from coarse contour lines to densely spaced contour lines, and deriving a containment relationship between the topologies from the change of the topology. ; And And an extractor configured to extract a region of interest hierarchically by calculating a nesting depth from the inclusion relationship between the topologies. The imaging unit forms the contour topologies at a first level with spacing between contours to filter the topologies having a nesting depth of less than or equal to a threshold value, and the interval between contours is equal to the first level until the region of interest is extracted. Contour contour topology is formed by sequentially setting denser levels of at least a second level or more, The imaging unit determines a level above the second level according to precision for hierarchical analysis of the ROI, and multi-scale images of the captured image, wherein the multi-scale image is an image in which each of the plurality of contour topologies is displayed. Characterized in that the display, Image analysis system of cell tissue. The method of claim 1, The imaging unit, Analyze the change in topology according to the intensity of the beam from the image formed by the beam that penetrates, reflects or refracts the cellular tissue, Form a contour map M _N (I) with _N quantizations for the image,

, here,

Is the minimum strength

And maximum strength

Intensity, which is the intensity range of the image (I) given in between

Characterized in that the contour line, Image analysis system of cell tissue. The method of claim 1, And the region of interest is a protruding cell tissue having a different density or a cell tissue formed of a material different from a surrounding cell tissue. The method of claim 1, And the imaging unit forms a sequential inclusion tree indicating mutual inclusion by using outlines included in a topology as nodes, thereby deriving a relationship of inclusion of the topologies. 5. The method of claim 4, The imaging unit of claim 1, wherein the imaging unit sets the contour of the first node where the topology changes as a base contour representing the region of interest. The method of claim 1, And the extractor calculates the number of contour lines included in one topology as the nesting depth. The method of claim 6, When the extraction unit has a plurality of child topologies in one parent topology, the image analysis of the cell tissue calculating the nesting depth of the parent topology as the nesting depth having the smallest number of outlines among the plurality of child topologies system. delete The method of claim 1, And the extracting unit matches a plurality of regions of interest for the same cell tissue to match regions of interest having mutual similarities higher than a predetermined value. 5. The method of claim 4, The extracting unit matches an area of interest having a higher mutual similarity than a preset value by matching the position, shape, or size of an outline of each node when matching a plurality of regions of interest for the same cell tissue. Forming a plurality of contour topologies of beam intensities attenuated by the cell tissue from a level where spacing between contours is coarse to a level where spacing between contours is dense in sequence; Deriving a containment relationship between the topologies from the change of the topologies; Extracting a region of interest by calculating a nesting depth from the inclusion relationship between the topologies; And Displaying multi-scale images of the captured image, wherein the multi-scale image is each of the plurality of contour topologies being displayed; In the step of forming the level, the contour topography is formed at the first level of spacing between contours to filter the topologies having a nesting depth of less than or equal to a threshold, and the interval between contours until the region of interest is extracted. Contour contour topology is formed by sequentially setting at least a second level or more denser than the first level, In the step of forming the level, characterized in that for determining the level above the second level according to the precision for the hierarchical analysis of the region of interest, Image analysis method of cell tissue. The method of claim 11, Forming the level, Analyzing a change in topology according to the intensity of the beam from an image formed by the beam that is transmitted through, reflected, or refracted from the cell tissue; And Forming a contour map M _N (I) with _N quantizations for the image,

, here,

Is the minimum strength

And maximum strength

Intensity, which is the intensity range of the image (I) given in between

Contours in, Image analysis method of cell tissue. The method of claim 11, The region of interest extracts protruding cell tissues having different densities or cell tissues formed of materials different from surrounding cell tissues. The method of claim 11, In the step of deriving the inclusion relationship, the image analysis method of the cell tissue to derive the inclusion relationship of the topology by forming a sequential inclusion tree that indicates whether the inclusion of the contour included in the topology as a node. The method of claim 14, In the step of deriving the inclusion relationship, the image analysis method of the cell tissue to set the contour of the first node that the topology changes to the base contour indicating the region of interest. The method of claim 11, Extracting the region of interest, and calculating the number of contours included in one topology as the nesting depth. The method of claim 16, An image analysis method of a cell tissue in which the nesting depth of the parent topology is calculated as the nesting depth of the parent topology when the plurality of children topologies are included in one parent topology. delete The method of claim 11, Extracting the region of interest, and matching a plurality of regions of interest for the same cell tissue to match regions of interest with mutual similarity higher than a predetermined value. The method of claim 11, In the extracting of the region of interest, when matching a plurality of regions of interest with respect to the same cell tissue, by matching the position, shape or size of the contour of each node to match the region of interest with mutual similarity higher than the predetermined value to each other Image analysis method. delete

Images