KR102318619B1 - Apparatus and method for improving area segmentation performance in medical image data - Google Patents

Abstract

Disclosed are an apparatus and method for improving region segmentation performance of medical image data, wherein the method for improving region segmentation performance of medical image data according to an embodiment of the present application divides a region of interest for each of a plurality of input 2D medical image data determining a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary for the region of interest obtained as a result of the region division, and the plurality of 2D medical care products based on the region division result Constructing a 3D model for a region of interest from image data, generating a plurality of cross-sections in a predetermined angular direction to include at least a portion of the candidate region from the 3D model, and the candidate region in each of the plurality of cross-sections The method may include determining, as a selected cross-section, a cross-section in which segmentation performance of the region of interest is improved based on a change in pixel brightness associated with , and correcting the 3D model based on the selected cross-section.

Description

Apparatus and method for improving region segmentation performance of medical image data

The present application relates to an apparatus and method for improving region segmentation performance of medical image data.

Medical image data, such as computed tomography (CT), is used for observing the patient's condition and lesion detection and change, and contains a lot of information. Using such CT, an internal cross-section of the patient's body can be obtained, which is stored in the form of a 2D image and is used as a decisive tool for diagnosis.

The medical staff analyzes medical image data such as CT in order to understand the condition of the subject, such as a patient, and recognizes and judges the problem situation through careful analysis of each slide. Various methods can be applied to facilitate this analysis. For example, when the boundary surface of an organ or lesion is extracted using region segmentation technology, considerations to be considered when diagnosing a subject can be easily observed, and patients or other medical staff Because it makes it easy to explain it to people, the technical field related to domain division is gradually developing.

In particular, CT, which is a tomography image of a subject such as a patient, is in the form of a Dicom standard, which has the advantage of being easy to change to a still image format that can be checked by a computer as an international standard.

In addition, in recent years, with the development of anticancer treatment, etc., the need for 3D stereoscopic information beyond the existing 2D image data is gradually increasing in order to increase the precision when establishing a surgical plan or evaluating anticancer drug responsiveness. In particular, if the region segmentation method is applied using a plurality of 2D medical image data, 3D modeling of human organs and tissues can be performed. In the process of reconstructing a cross-sectional image of a gap, the resolution of the image data is lowered, so in the 3D modeling result, an unclear boundary surface is shown in a specific area (for example, the corner of the liver, etc.) according to the characteristics of human organs and tissues. There are many limitations.

The technology that is the background of the present application is disclosed in Korean Patent Publication No. 10-1004342.

The present application is to solve the problems of the prior art described above, and utilizes cross-sectional reconstruction from 2D medical image data, such as CT, through analysis in consideration of the anatomical characteristics of human organs, organs, tissues, etc. An object of the present invention is to provide an apparatus and method for improving the region segmentation performance of medical image data that can improve the accuracy of boundary surface extraction during 3D modeling.

The present application is intended to solve the problems of the prior art, and when 3D modeling of a region of interest such as a human organ, organ, tissue, etc., the boundary of the region where it is difficult to extract the boundary due to the large curvature or sharp end surface in the region of interest The purpose is to extract accurately.

However, the technical problems to be achieved by the embodiments of the present application are not limited to the technical problems as described above, and other technical problems may exist.

As a technical means for achieving the above technical problem, a method for improving region segmentation performance of medical image data according to an embodiment of the present application includes performing region segmentation on a region of interest on each of a plurality of input 2D medical image data. Step, determining a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary of the region of interest obtained as a result of the region division, from the plurality of 2D medical image data based on the region division result constructing a 3D model for a region of interest, generating a plurality of cross-sections in a predetermined angular direction to include at least a portion of the candidate region from the 3D model; The method may include determining, as a selected cross-section, a cross-section having improved segmentation performance for the region of interest based on a change in pixel brightness, and correcting the 3D model based on the selected cross-section.

In the determining of the candidate region, a point that does not satisfy a pixel brightness change condition in a vertical direction with respect to the ROI among points of the boundary may be determined as the candidate region.

In addition, the pixel brightness change condition includes at least a part of a region in which the differential value of the pixel brightness change exceeds a preset upper limit value in a region close to the corresponding point by a predetermined level or more, and a preset lower limit value outside the region close to the predetermined level or more It may be satisfied if the following is maintained.

Also, the constructing of the 3D model may include deriving point cloud data for the region of interest in a 3D space based on the plurality of 2D medical image data.

In addition, the generating of the plurality of cross-sections in the predetermined angular direction may include generating the plurality of cross-sections in all angular directions with respect to the candidate region.

In addition, in the generating of the plurality of cross-sections in the predetermined angular direction, it is determined that the probability that the segmentation performance of the region of interest is improved is greater than or equal to a preset threshold probability based on the sampled 3D data for the candidate region. The plurality of cross-sections may be generated by selecting the angular direction of .

Also, the region of interest may include a predetermined intra-abdominal organ.

In addition, in the generating of the plurality of cross-sections in the predetermined angular direction, the plurality of cross-sections may be generated by selecting a predetermined angular direction in consideration of anatomical characteristics of an organ corresponding to the region of interest.

Also, the determining of the candidate region may be performed in consideration of anatomical characteristics of an organ corresponding to the region of interest.

Also, the determining of the candidate region may include determining the candidate region at a point where a curvature exceeds a preset threshold curvature among boundaries of the region of interest.

Also, the 2D medical image data may include CT image data.

On the other hand, the apparatus for improving region segmentation performance of medical image data according to an embodiment of the present application includes a region divider for performing region division on a region of interest on each of a plurality of input 2D medical image data, obtained as a result of region division a candidate region determiner that determines a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary with respect to the region of interest; A three-dimensional transform unit constructing a model and a plurality of cross-sections in a predetermined angular direction are generated from the 3D model to include at least a portion of the candidate area, and pixel brightness change associated with the candidate area in each of the plurality of cross-sections The cross-section selection unit may include a cross-section selection unit that determines, as a selected cross-section, a cross-section having improved segmentation performance for the region of interest based on the .

Also, the 3D transform unit may correct the 3D model based on the selected cross-section.

In addition, the cross-section selection unit generates a plurality of cross-sections in all angular directions with respect to the candidate region or has a preset probability of improving segmentation performance of the region of interest based on sampled 3D data for the candidate region. The plurality of cross-sections may be generated by selecting a predetermined angular direction determined to be equal to or greater than a threshold probability, or a predetermined angular direction may be selected in consideration of anatomical characteristics of the ROI to generate the plurality of cross-sections.

The above-described problem solving means are merely exemplary, and should not be construed as limiting the present application. In addition to the exemplary embodiments described above, additional embodiments may exist in the drawings and detailed description.

According to the above-described problem solving means of the present application, 3D modeling of a region of interest is performed by utilizing cross-sectional reconstruction from 2D medical image data, such as CT, through analysis in consideration of anatomical characteristics of regions of interest, such as human organs, organs, and tissues. An apparatus and method for improving region segmentation performance of medical image data capable of improving boundary surface extraction accuracy may be provided.

According to the above-described problem solving means of the present application, it is possible to more accurately extract the boundary surface of the region where it is difficult to extract the boundary surface due to the large curvature or sharp end surface in the region of interest during 3D modeling of the region of interest such as human organs, organs, and tissues. have.

According to the above-described problem solving means of the present application, as the boundary of the 3D model with respect to the region of interest, such as a human organ, organ, or tissue, becomes clear, accuracy of calculating a volume associated with the region of interest may be improved.

However, the effects obtainable herein are not limited to the above-described effects, and other effects may exist.

1 is a schematic configuration diagram of a medical image data analysis system including an apparatus for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure.
FIG. 2 is a diagram for explaining a change in pixel brightness in a vertical direction at each point of a boundary of a region of interest.
3 is a graph schematically illustrating a pixel brightness change at a point satisfying a pixel brightness change condition according to an exemplary embodiment of the present disclosure.
4 is a view for explaining an angular direction for generating a plurality of cross-sections in order to generate or correct a 3D model with improved region division performance.
5 to 7 are diagrams exemplarily illustrating a plurality of cross-sections generated by the apparatus for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure based on various angular directions.
FIG. 8 is a diagram for explaining determining, as a selection cross-section, a cross-section having improved segmentation performance for a region of interest with respect to a determined candidate region.
9 is a schematic configuration diagram of an apparatus for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure.
10 is an operation flowchart of a method for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure.

Hereinafter, embodiments of the present application will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art to which the present application pertains can easily implement them. However, the present application may be embodied in several different forms and is not limited to the embodiments described herein. And in order to clearly explain the present application in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout this specification, when a part is "connected" with another part, it is not only "directly connected" but also "electrically connected" or "indirectly connected" with another element interposed therebetween. "Including cases where

Throughout this specification, when it is said that a member is positioned "on", "on", "on", "under", "under", or "under" another member, this means that a member is positioned on the other member. It includes not only the case where they are in contact, but also the case where another member exists between two members.

Throughout this specification, when a part "includes" a component, it means that other components may be further included, rather than excluding other components, unless otherwise stated.

The present application relates to an apparatus and method for improving region segmentation performance of medical image data. In particular, the present application provides an apparatus and method for improving the performance of segmentation for a specific area with respect to medical image data, in which a plurality of CT images are aligned in a three-dimensional space and the axis segmentation is clear (easy). ), and to an apparatus and method for improving the accuracy of boundary surface extraction by generating a cross-section based on the

For reference, in the description of the embodiments of the present application, terms (left direction, right direction, up-down direction, etc.) related to directions or positions have been described based on the arrangement state of each component shown in the drawings. For example, when viewed in FIG. 1 , the left direction may be the 9 o'clock direction, the right direction may be the 3 o'clock direction, the upper direction may be the 12 o'clock direction, and the downward direction may be the 6 o'clock direction.

However, the direction setting may vary depending on a direction in which the medical image data input to the apparatus of the present application is captured, a direction in which the medical image data is viewed, and the like. For example, if necessary, the medical image data may be acquired so that the upward direction of FIG. 1 faces a horizontal direction (left and right direction), and as another example, the medical image data may be rotated so that the upward direction of FIG. 1 is slanted. can

1 is a schematic configuration diagram of a medical image data analysis system including an apparatus for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure.

Referring to FIG. 1 , the medical image data analysis system 10 of the present application may include the apparatus 100 for improving the region segmentation performance of medical image data according to an exemplary embodiment of the present application. Also, referring to FIG. 1 , the apparatus 100 for improving the region segmentation performance of medical image data according to an embodiment of the present application receives, as an input, a plurality of 2D medical image data 1 captured as a 2D image, for example. Thus, it may be implemented to provide a 3D model 2 derived from a plurality of 2D medical image data. According to an exemplary embodiment of the present disclosure, the 2D medical image data 1 may include CT image data.

Also, FIG. 1A is an example of a 3D model 2 provided by the apparatus 100 for improving the region segmentation performance of medical image data according to an exemplary embodiment of the present application, and FIG. 1B is Among the implemented 3D models (2), the boundary surface of the area where the boundary surface is difficult to clearly distinguish due to the presence of difficulty in distinguishing it from other organs due to the presence of sharp or sharp edges or adjacent to other organs 3 corrected according to the present application This is an enlarged view of the dimensional mesh creation result.

Also, referring to FIG. 1 , the apparatus 100 for improving the region segmentation performance of medical image data according to an embodiment of the present application sets a liver, which is an organ in the abdominal cavity, as a region of interest, and performs three-dimensional modeling after region segmentation of the liver region. It is possible to create a volume (3D object) according to At this time, depending on the anatomical shape and positional characteristics of the liver, there are parts where the region can be divided well and parts where it cannot be divided. It may be more pronounced in narrow or dense areas (eg, the marginal area of the liver).

Also, according to an embodiment of the present application, the apparatus 100 for improving the region division performance of medical image data according to the embodiment of the present application (hereinafter, referred to as a 'region division performance improvement apparatus 100 ') may take a medical image. A plurality of 2D medical image data 1 may be received from a device (not shown). In addition, the 3D model 2 generated by the apparatus 100 for improving the region division performance may be implemented to be transmitted and stored to a user terminal (not shown) or displayed by a display module of the user terminal (not shown). Also, according to an exemplary embodiment of the present disclosure, the apparatus 100 for improving the region division performance may be provided as a device included under a medical imaging apparatus (not shown). As another example, the apparatus 100 for improving the region division performance may be provided as a device included under the user terminal (not shown).

According to an embodiment of the present application, the medical imaging apparatus (not shown) may include a medical computed tomography (CT) scanner, an X-ray imaging device, a magnetic resonance imaging (MRI) device, etc. but is not limited. That is, the medical imaging apparatus of the present application and the medical image data to be analyzed are not limited to a specific type, and include a conventional medical imaging device for imaging the inside of a subject, such as a patient, and a imaging device to be developed in the future. It is preferable to be understood as a concept that

For example, a user terminal (not shown) includes a smartphone, a smart pad, a tablet PC, and the like and a PCS (Personal Communication System), GSM (Global System for Mobile communication), PDC (Personal Digital Cellular), PHS(Personal Handyphone System), PDA(Personal Digital Assistant), IMT(International Mobile Telecommunication)-2000, CDMA(Code Division Multiple Access)-2000, W-CDMA(W-Code Division Multiple Access), Wibro(Wireless Broadband Internet) ) may include all types of wired/wireless communication devices such as terminals.

According to an embodiment of the present disclosure, the above-described medical image capturing apparatus, the region division performance enhancing apparatus 100 and the user terminal may communicate through an interconnected network (not shown).

Here, the network (not shown) is illustratively a 3rd Generation Partnership Project (3GPP) network, a Long Term Evolution (LTE) network, a 5G network, a World Interoperability for Microwave Access (WIMAX) network, the Internet, and a LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), Bluetooth network, Wi-Fi network, satellite broadcasting network, analog broadcasting network , a Digital Multimedia Broadcasting (DMB) network, and the like, but is not limited thereto.

The region segmentation performance improvement apparatus 100 may segment the ROI on each of the plurality of input 2D medical image data 1 . Here, performing region segmentation may be understood as extracting a boundary (contour) for the region of interest.

According to an embodiment of the present disclosure, the apparatus 100 for improving the region segmentation performance utilizes a semantic segmentation technique or corresponds to the intensity or distribution of contrast of each region of the 2D medical image data 1 and the region of interest. In consideration of morphological information such as organs and organs, the primary region segmentation of the region of interest may be performed.

According to an embodiment of the present disclosure, the region of interest may include a predetermined intra-abdominal organ. For example, the region of interest may include liver, kidney, spleen, lung, stomach, and the like. However, the region of interest herein is not limited to a specific organ, and may include a wide range of various elements present in the body of a subject that can be identified from medical image data, such as blood vessels, tissues, tumor lesions, bones, and pelvis. In particular, the present application provides not only the extraction of the external interface of an organ (eg, heart, kidney, spleen, etc.) having a smooth curve (curved surface), but also an external surface of an organ (eg, liver, etc.) having a large curved or sharp end surface. There is an effect that the interface can also be easily extracted.

The apparatus 100 for improving the region segmentation performance may determine a candidate region among the boundaries of the region of interest based on a change in pixel brightness at each point of the boundary of the region of interest obtained as a result of region division. Here, the candidate region may be determined as a region whose boundary is unclear or ambiguous among the boundaries of the region of interest obtained as a result of the preceding region division. In other words, when the 3D model 2 is constructed from the plurality of 2D medical image data 1 , the candidate region may refer to a region in which the boundary with the region of interest is likely to appear inaccurately. That is, the candidate region may mean a region in which the boundary surface needs to be newly analyzed.

Hereinafter, with reference to FIGS. 2 and 3 , a process for determining a candidate region by the region division performance enhancing apparatus 100 will be described in detail.

FIG. 2 is a diagram for explaining a change in pixel brightness in a vertical direction at each point of a boundary of a region of interest. Referring to FIG. 2, (a) of FIG. 2 shows a pixel brightness change graph at a point where the pixel brightness change in the vertical direction is relatively clear, and FIG. 2(b) is a point where the pixel brightness change in the vertical direction is somewhat unclear. The graph shows the change in pixel brightness in

Referring to FIG. 2 , the apparatus 100 for improving region segmentation performance based on a change in pixel brightness at each point of a boundary of an ROI (referring to FIG. 2 , a boundary line indicated in red) obtained as a result of region division. Candidate regions can be determined. Specifically, the apparatus 100 for improving the region division performance may determine, as a candidate region, a point that does not satisfy the pixel brightness change condition in the vertical direction with respect to the region of interest among the points of the boundary.

Here, in the pixel brightness change condition according to an embodiment of the present application, at least a part of a region in which the differential value of the pixel brightness change is close to the corresponding point by a predetermined level or more, the region exceeding the preset upper limit exists, and the corresponding point is close to the corresponding point by a predetermined level or more. Outside the region, it may be satisfied when the preset lower limit value is maintained or less. In this way, the present application introduces the threshold values (upper limit value and lower limit value) as a criterion for determining a candidate region, so that a candidate region requiring reconfiguration of the boundary can be detected more closely.

3 is a graph schematically illustrating a pixel brightness change at a point satisfying a pixel brightness change condition according to an exemplary embodiment of the present disclosure.

Referring to FIG. 3, (a) of FIG. 3 is a graph showing the amount of change in pixel brightness based on a predetermined point selected among boundaries, and (b) of FIG. For example, the first derivative) is a graph showing the differential value of the pixel brightness change amount based on a selected predetermined point. Predetermined points selected in (a) and (b) of FIG. 3 are indicated by a dotted line with respect to the horizontal axis.

Referring to (b) of FIG. 3 , in the case of a point satisfying the pixel brightness change condition of the present application shown in FIG. 3, a region exceeding a preset upper threshold exists in a region close to the corresponding point by a predetermined level or more. It can be confirmed that both the condition and the condition for maintaining the preset lower threshold or lower outside the region close to the corresponding point or more by a predetermined level are satisfied. If the pixel brightness change condition is satisfied, it may be determined that the region division has been performed relatively accurately so that the boundary of the region of interest is clearly distinguished at the corresponding point. In this case, the corresponding point satisfying the pixel brightness change condition is a candidate It may not be determined (not selected) as an area.

Conversely, as in the case of (b) of FIG. 2 , when the change in pixel brightness does not appear clearly based on the corresponding point (in other words, when the pixel brightness change condition of the present application is not satisfied), the corresponding point is a candidate area. After the decision (selection) is made, a process for clarifying the boundary for the corresponding candidate area is performed.

According to an embodiment of the present application, the value of the range (or length, width, etc.) of the adjacent region at the corresponding point, which is a criterion for whether the predetermined upper limit value, the preset lower limit value and the upper limit value are satisfied and whether the lower limit value is satisfied for determining the candidate region, is of interest It may be determined based on at least one of a region type, a resolution of the medical image data, an average pixel value of the 2D medical image data 1 , a median of pixel values, and a pixel value range of the 2D medical image data 1 . In addition, as another example, the range (or length, width, etc.) value of the adjacent region at the corresponding point, which is a criterion for whether the predetermined upper limit value, the preset lower limit value, and the upper limit value are satisfied and whether the lower limit value is satisfied for determining the candidate region is determined by the user input. It can be implemented to be adjusted. In this case, the user needs to variously adjust at least one of the preset upper limit value for the candidate region determination, the preset lower limit value, and the range of the adjacent region at the corresponding point, which is a criterion for whether the upper limit value is satisfied and whether the lower limit value is satisfied through a user input. It is possible to determine at what level the candidate region is determined according to It can be made to meet the requirements of the 3D model (2).

Also, according to an exemplary embodiment of the present disclosure, the apparatus 100 for improving the region division performance may be designed to determine a candidate region at a point where the curvature exceeds a preset threshold curvature among boundaries of the region of interest. Here, a point whose curvature is equal to or less than a preset threshold curvature among the boundaries of the ROI may be understood as a region in which boundary detection is performed relatively well because the shape of the ROI adjacent to the corresponding point is relatively gentle. A point at which the curvature exceeds a preset threshold curvature among the boundaries of the boundary may be understood as an area with a high probability of inaccurate boundary detection, such as a sharp shape or a sharp edge of an ROI adjacent to the corresponding point. As such, the apparatus 100 for improving the region division performance may infer the difficulty of detecting a boundary at a corresponding point based on curvature information (eg, radius of curvature, etc.) of each point of the primarily detected boundary, and the obtained boundary It may be implemented to preferentially select a candidate region that is likely to require re-detection of the boundary surface based on the detection difficulty.

Also, according to an embodiment of the present disclosure, the apparatus 100 for improving the region segmentation performance may determine a candidate region in consideration of anatomical characteristics of an organ corresponding to the region of interest. In other words, the apparatus 100 for improving the region segmentation performance does not perform a candidate region search for all parts of the region of interest in consideration of the anatomical characteristics of the organ corresponding to the region of interest, but rather a region (eg, a region with distinct characteristics of the organ). For example, it may operate to preferentially determine a candidate region, particularly in a sharp part, an angled part, etc.).

Specifically, for example, when the region of interest is 'liver', in general, when viewed as a whole, the left region of the liver may be gentle, and the right region may include a relatively sharp and sharp edge. Accordingly, the difficulty of boundary detection in the right region may be higher than in the left region. In consideration of the anatomical characteristics of the region of interest, the region division performance improving apparatus 100 may operate to preferentially determine a candidate region in the right region rather than in the left region of the liver. For example, the apparatus 100 for improving the region segmentation performance may include a region with a high probability of having a relatively gentle shape and a sharp or sharp edge based on the general shape of the corresponding organ based on information on which organ the region of interest corresponds to. A preset upper limit value for determining a candidate region in a region having a high probability of having .

Also, the apparatus 100 for improving the region segmentation performance may construct a 3D model 2 ′ of the region of interest from a plurality of 2D medical image data 1 based on the region segmentation result. Here, the 3D model 2' of the constructed ROI may include an ambiguous or inaccurate boundary surface. In other words, the 3D model 2 ′ generated in this step may refer to the 3D model before correction (resetting the boundary) on the previously determined candidate region is performed.

The three-dimensional object for the region of interest is converted into a two-dimensional image by CT scan and stored. However, if the loss of the two-dimensional image information is not large, processing into a three-dimensional object in the three-dimensional space is possible again. In this case, the lost information can be generated by interpolation.

In this regard, according to an embodiment of the present application, the apparatus 100 for improving the region division performance is configured to provide a region of interest in a 3D space based on a plurality of 2D medical image data 1 in order to construct a 3D model 2 ′. It may be to derive point cloud data for In other words, the apparatus 100 for improving the region segmentation performance may utilize a method for synthesizing 2D images taken from various directions as a technique for building a 3D model and converting it into a point cloud (point cloud) in a three-dimensional space, However, the present invention is not limited thereto. In addition, the apparatus 100 for improving the region division performance may generate a new section 1 ′ using an arbitrary axis or an arbitrary angular direction based on the subsequently generated 3D model 2 or reconstructed 3D data. .

The apparatus 100 for improving the region segmentation performance divides all pixel values on the x-axis, y-axis, and z-axis of each part of the region of interest in the 3D space at predetermined intervals based on the plurality of 2D medical image data 1 . Can be assigned (can be filled). The 3D model 2' (in other words, the 3D model before correction) built by the region segmentation performance improving apparatus 100 may be utilized to obtain a cross section in different angular directions for improving segmentation performance for a candidate region. can

The apparatus 100 for improving the region division performance may generate a plurality of cross-sections 1 ′ in a predetermined angular direction to include at least a portion of the candidate region from the constructed 3D model 2 ′.

Hereinafter, a process of generating a plurality of cross-sections 1 ′ in a predetermined angular direction by the apparatus 100 for improving the region division performance will be described in detail with reference to FIGS. 4 to 7 .

4 is a view for explaining an angular direction for generating a plurality of cross-sections in order to generate or correct a 3D model with improved region division performance.

Referring to FIG. 4 , a cross-section 1 ′ in any and all angular directions θ may be generated based on a predetermined point. In addition, referring to FIG. 4 , the angular direction ( θ ) may be measured in different ways based on the direction in which the subject looks at the body ( θ _{1 )} or θ ₂ ) .

Referring to FIG. 4 , in FIG. 4 , for convenience of explanation, the angular direction ( θ ) is measured based on the center point of the subject's body, but the reference position for calculating the angular direction ( θ ) is the position (coordinate) of the determined candidate area. can be determined as

According to an embodiment of the present application, when a predetermined angular direction θ is determined, an axis for generating a cross section 1 ′ including at least a portion of a candidate region is determined accordingly, and region division performance is improved. The device 100 may be to generate the cross section 1 ′ based on a determined arbitrary axis.

5 to 7 are diagrams exemplarily illustrating a plurality of cross-sections generated by the apparatus for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure based on various angular directions.

5 to 7 , the apparatus 100 for improving the region division performance provides a plurality of cross-sections 1 ′ in all angular directions with respect to the determined candidate region (eg, a 360-degree omnidirectional cross-section based on the candidate region) (1')) can be created.

In addition, according to an embodiment of the present disclosure, the apparatus 100 for improving the segmentation performance of the region determines that the probability that the segmentation performance of the region of interest is improved is equal to or greater than a preset threshold probability based on the sampled 3D data of the candidate region. A plurality of cross-sections 1' may be generated by selecting a predetermined angular direction. This has the advantage that accurate results can be obtained because the cross-section 1' can be generated in consideration of data for all angular directions in the case of the method of generating the cross-section 1' for all angular directions described above. This is because the computation time (eg, the time required to generate the cross section 1 ′ in all angular directions) may be increased by considering all angular directions. In other words, the region division performance apparatus 100 uses sampled 3D data instead of considering all angular directions in order to shorten the calculation time, and uses the sampled 3D data to identify a rough amount of change (eg, change in pixel brightness), and The cross section 1' can be created by selecting the most probable angular direction (in other words, the most probable axial change) among them.

For example, the region division performance improving apparatus 100 extracts only line data in a predetermined direction with respect to a candidate region as sampled 3D data, and then uses a change in brightness of a sampled pixel value in the corresponding line data to obtain a predetermined value. It may be implemented to select an angular direction. Here, the selection of a predetermined direction for line data extraction may be performed only for some directions sampled by selecting a θ value having a preset interval.

6 and 7 , the apparatus 100 for improving the region division performance selects 90 degrees, 45 degrees, 13 degrees, and 0 degrees angular directions instead of all angular directions by using sampled three-dimensional data. Create a section (1') (Fig. 6) or select an angular direction of 22.5, 45, 67.5, 90, 112.5, 135, and 157.5 degrees to create a section (1') rather than all angular directions (FIG. 7) can be done.

In addition, according to an embodiment of the present disclosure, the apparatus 100 for improving the region division performance selects a predetermined angular direction in consideration of the anatomical characteristics of an organ corresponding to the region of interest to form a plurality of cross-sections 1 ′. can create For example, when the region of interest is 'liver', in general, when viewed as a whole, the left region of the liver may be gentle, and the right region may include a relatively sharp and sharp edge. Accordingly, the difficulty of boundary detection in the right region may be higher than in the left region. In consideration of the anatomical characteristics of the ROI, the apparatus 100 for improving the region division performance considers different inclination characteristics between different angular directions based on location information of the determined candidate region in the ROI in a predetermined angular direction. In the case of generating the cross section 1 ′ by selecting

According to an embodiment of the present application, a method of generating a cross-section in all of the above-described angular directions for one candidate region, a method of generating a cross-section by selecting a predetermined angular direction based on sampled 3D data, and a region of interest A method of selecting a predetermined angular direction in consideration of the anatomical characteristics of an organ corresponding to and a pixel value range of the 2D medical image data 1 may be implemented to be determined based on at least one of the ranges. As another example, selecting one of the three cross-section generating methods described above may be performed as a method selected by a user based on a separate user input applied to the region division performance improving apparatus 100 .

The apparatus 100 for improving the region segmentation performance may determine a cross section with improved segmentation performance for the region of interest as the selected cross section based on a change in pixel brightness associated with the candidate region in each of the generated cross-sections 1 ′. Here, the cross-section with improved segmentation performance refers to a cross-section 1 including data that can more clearly distinguish the boundary of the candidate region compared to the 2D medical image data 1 used when the previously constructed 3D model 2 is generated. ') is present among the generated cross-sections 1', if the corrected 3D model 2 is generated by correcting the 3D model 2' built on the basis of this, the boundary surface for the region of interest is clear. It can be understood as indicating a cross-section that can be made.

The apparatus 100 for improving region division performance according to an embodiment of the present disclosure may determine a cross-section 1 ′ in which a point satisfying the above-described pixel brightness change condition among a plurality of cross-sections 1 ′ exists as a selected cross-section. . In addition, according to an embodiment, when there are a plurality of cross sections 1 ′ having points satisfying the pixel brightness change condition, the apparatus 100 for improving the region division performance determines whether a point satisfying the pixel brightness change condition exists. All of the plurality of cross-sections 1 ′ may be determined as selected cross-sections. According to an embodiment, detailed conditions (eg, upper limit value, lower limit value, etc.) of a pixel brightness change condition for determining a candidate area and a pixel brightness change condition for determining a selection section may be determined differently.

As another example, the apparatus 100 for improving the region division performance may provide a pixel brightness change value associated with a candidate region in each of the plurality of cross-sections 1 ′ (a pixel brightness change at a point corresponding to the candidate region in the corresponding cross-section 1 ′). value) may be implemented to align a plurality of cross-sections 1 ′ in an order of increasing magnitude to determine a higher-order cross-section as many as a preset number of cross-sections as the selected cross-section.

As another example, the apparatus 100 for improving the region division performance may provide a pixel brightness change value associated with a candidate region in each of the plurality of cross-sections 1 ′ (pixel brightness at a point corresponding to the candidate region in the corresponding cross-section 1 ′). The change value) may be implemented to determine all cross-sections 1' that are equal to or greater than a preset threshold change value as selected cross-sections.

After the selected cross-section is determined, the apparatus 100 for improving the region division performance may correct the 3D model 2 ′ built based on the selected cross-section. Here, the meaning of being able to correct the constructed 3D model 2' means that the existing ambiguous boundary surface is changed to a clear boundary surface (outline) based on the determined selection cross-section, so that the 3D closer to the actual region of interest (organ, etc.) It can be understood that the model 2 can be created.

If the boundary point for the region of interest in the selection cross-section is located at a different point from the point in the boundary determined based on the preceding region division result (in other words, a candidate region), the apparatus 100 for improving region division performance may reset the boundary point in the selected cross-section to the boundary point on the 3D model 2 . In other words, the apparatus 100 for improving the region division performance may search for a boundary point closer to the actual boundary of the ROI from the selection cross-section. The resetting of the boundary point and setting the coordinates of the reset boundary point in the three-dimensional space may be performed by the following Equations 1-1 and 1-2.

[Equation 1-1]

Here, x , y , and z are the coordinates of the existing candidate area (coordinates of the current point), x' , y' , and z' are the coordinates of the changed boundary point, and Rx(θ) , Ry(θ) and Rz(θ) ) can be calculated by the following Equation 1-2.

[Equation 1-2]

Here, θ may be a rotation angle corresponding to the selected cross-section.

In addition, according to an embodiment of the present application, Equation 1-1 and Equation 1-2 generate a point matching the determined candidate area when a plurality of cross-sections 1' in a predetermined angular direction are generated. ) can also be used when converting to coordinates in

In summary, the apparatus 100 for improving the region segmentation performance generates cross-sections 1' in various angular directions with respect to the candidate area corresponding to the ambiguous boundary surface, and clearly defines boundary information on the region of interest among the generated cross-sections 1'. A 3D model 2 is created by selecting a cross-section 1' that can be provided according to a predetermined criterion, or by correcting a pre-established 3D model 2', the boundary surface for the selection area is clear even in sharp or angled parts. It is possible to provide a calibrated 3D model 2 that is clearly distinguished.

FIG. 8 is a diagram for explaining determining, as a selection cross-section, a cross-section having improved segmentation performance for a region of interest with respect to a determined candidate region.

Referring to FIG. 8, (a) of FIG. 8 shows a candidate region corresponding to an ambiguous boundary obtained as a result of region division, and (b) of FIG. 8 is generated in a predetermined angular direction to include at least a part of the candidate region. An example of a plurality of cross-sections 1' is shown, and FIGS. 3 (c1) and (c2) show a cross-section 1' determined as a selected cross-section with improved division performance based on a change in pixel brightness associated with a candidate area. is shown as an example.

For example, (c1) may be a selected cross section in an angular direction having a θ value of 20 degrees, and (c2) may be a selected cross section in an angular direction having a θ value of 110 degrees.

9 is a schematic configuration diagram of an apparatus for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure.

Referring to FIG. 9 , the apparatus 100 for improving the region segmentation performance of medical image data according to an embodiment of the present application includes a region divider 110 , a candidate region determiner 120 , a 3D transform unit 130 , and A cross-section selection unit 140 may be included.

The region divider 110 may perform region division on a region of interest for each of the plurality of input 2D medical image data 1 .

The candidate region determiner 120 may determine a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary with respect to the region of interest obtained as a result of region division.

The 3D transform unit 130 may construct a 3D model 2 ′ of the region of interest from the plurality of 2D medical image data 1 based on the region division result. Also, the 3D transform unit 130 may correct the 3D model based on the selected cross-section determined by the cross-section selection unit 140 .

The cross-section selection unit 140 may generate a plurality of cross-sections 1 ′ in a predetermined angular direction to include at least a portion of the candidate region from the 3D model 2 ′. Also, the cross-section selector 140 may determine a cross-section having improved segmentation performance for the region of interest as the selected cross-section based on a change in pixel brightness associated with the candidate region in each of the plurality of cross-sections 1 ′.

Hereinafter, an operation flow of the present application will be briefly reviewed based on the details described above.

10 is an operation flowchart of a method for improving region segmentation performance of medical image data according to an exemplary embodiment of the present disclosure.

The method for improving the region segmentation performance of medical image data illustrated in FIG. 10 may be performed by the above-described region segmentation performance improving apparatus 100 . Accordingly, even if omitted below, the description of the apparatus 100 for improving the region division performance may be equally applied to the description of the method for improving the region division performance of medical image data.

Referring to FIG. 10 , in operation S1010 , the region dividing unit 110 may perform region division of an ROI on each of a plurality of input 2D medical image data 1 .

Next, in operation S1020 , the candidate region determiner 120 may determine a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary with respect to the region of interest obtained as a result of region division.

Specifically, in operation S1020 , the candidate region determiner 120 may determine, as a candidate region, a point that does not satisfy the pixel brightness change condition in the vertical direction with respect to the ROI among points of the boundary. In particular, as for the pixel brightness change condition, according to an exemplary embodiment of the present application, at least a portion of a region in which a differential value of a pixel brightness change amount is close to the corresponding point by a predetermined level or more, a region exceeding a preset upper limit exists, and a region close to a predetermined level or more Outside of , it may be satisfied when the preset lower limit or less is maintained.

Also, in operation S1020 , the candidate region determiner 120 may determine the candidate region in consideration of the anatomical characteristics of the organ corresponding to the region of interest.

Also, according to an exemplary embodiment of the present disclosure, in operation S1020 , the candidate region determiner 120 may determine a candidate region at a point where a curvature exceeds a preset threshold curvature among boundaries with respect to the region of interest.

Next, in operation S1030 , the 3D transform unit 130 may build a 3D model 2 ′ of the region of interest from the plurality of 2D medical image data 1 based on the region division result.

Specifically, in step S1030 , the 3D transform unit 130 may derive point cloud data for the region of interest in the 3D space based on the plurality of 2D medical image data 1 .

Next, in step S1040, the cross-section selection unit 140 may generate a plurality of cross-sections 1' in a predetermined angular direction to include at least a portion of the candidate area from the 3D model 2' built in step S1030. have.

For example, in operation S1040 , the cross-section selection unit 140 may generate a plurality of cross-sections 1 ′ in all angular directions with respect to the candidate region.

As another example, in step S1040, the cross-section selection unit 140 selects a predetermined angular direction in which it is determined that the probability that the segmentation performance of the region of interest is improved is equal to or greater than a preset threshold probability based on the sampled three-dimensional data of the candidate region. By selection, a plurality of cross-sections 1 ′ may be generated.

As another example, in operation S1040 , the cross-section selection unit 140 may generate a plurality of cross-sections 1 ′ by selecting a predetermined angular direction in consideration of anatomical characteristics of an organ corresponding to the region of interest.

Next, in step S1050 , the cross-section selection unit 150 may determine, as the selected cross-section, a cross-section having improved segmentation performance for the region of interest based on a change in pixel brightness associated with a candidate region in each of the plurality of cross-sections 1 ′. have.

Next, in step S1060, the 3D transform unit 130 may correct the 3D model based on the selected cross-section.

In the above description, steps S1010 to S1060 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present application. In addition, some steps may be omitted as necessary, and the order between steps may be changed.

The method for improving region division performance of medical image data according to an embodiment of the present disclosure may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and carry out program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

Also, the above-described method for improving the region division performance of medical image data may be implemented in the form of a computer program or application stored in a recording medium and executed by a computer.

The foregoing description of the present application is for illustration, and those of ordinary skill in the art to which the present application pertains will understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present application. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and likewise components described as distributed may be implemented in a combined form.

The scope of the present application is indicated by the following claims rather than the above detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present application.

10: Medical image data analysis system
100: apparatus for improving the region segmentation performance of medical image data
110: region division part
120: candidate area determining unit
130: three-dimensional transformation unit
140: section selection unit
1: 2D medical image data
1': cross section
2': 3D model before calibration
2: 3D model after calibration

Claims (13)

A method for improving region segmentation performance of medical image data performed by an apparatus for improving region segmentation performance of medical image data, the method comprising:
performing region segmentation on a region of interest on each of a plurality of input 2D medical image data;
determining a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary with respect to the region of interest obtained as a result of the region division;
constructing a 3D model of a region of interest from the plurality of 2D medical image data based on the region division result;
generating a plurality of cross-sections in a predetermined angular direction to include at least a portion of the candidate region from the 3D model;
determining, as a selection cross-section, a cross-section having improved segmentation performance for a region of interest based on a change in pixel brightness associated with the candidate region in each of the plurality of cross-sections; and
correcting the 3D model based on the selected cross-section;
including,
The step of determining the candidate region comprises:
and determining, as the candidate region, a point that does not satisfy a pixel brightness change condition in a vertical direction with respect to the region of interest among the points of the boundary as the candidate region. delete According to claim 1,
The pixel brightness change condition is
At least some regions in which the differential value of the pixel brightness change exceeds the preset upper limit value in a region close to the corresponding point by a predetermined level or more, and maintains a preset lower limit value or less outside the region close to the predetermined level or more , how to improve zoning performance. According to claim 1,
The step of building the 3D model is,
and deriving point cloud data for a region of interest in a 3D space based on the plurality of 2D medical image data. According to claim 1,
The step of generating a plurality of cross-sections in the predetermined angular direction comprises:
generating a plurality of cross-sections in all angular directions with respect to the candidate region. According to claim 1,
The step of generating a plurality of cross-sections in the predetermined angular direction comprises:
A region that generates the plurality of cross-sections by selecting a predetermined angular direction in which it is determined that a probability that the segmentation performance of the region of interest is improved is equal to or greater than a preset threshold probability based on the sampled 3D data for the candidate region How to improve partitioning performance. According to claim 1,
The method for improving region segmentation performance, wherein the region of interest includes a predetermined intra-abdominal organ. 8. The method of claim 7,
The step of generating a plurality of cross-sections in the predetermined angular direction comprises:
and generating the plurality of cross-sections by selecting a predetermined angular direction in consideration of anatomical characteristics of an organ corresponding to the region of interest. 8. The method of claim 7,
The step of determining the candidate region comprises:
The method for improving region segmentation performance, characterized in that it is performed in consideration of anatomical characteristics of an organ corresponding to the region of interest. 10. The method of claim 9,
The step of determining the candidate region comprises:
and determining the candidate region at a point where a curvature exceeds a preset threshold curvature among boundaries of the region of interest. According to claim 1,
The method of claim 1, wherein the 2D medical image data includes CT image data. An apparatus for improving region segmentation performance of medical image data, the apparatus comprising:
a region divider performing region division on a region of interest on each of the plurality of input 2D medical image data;
a candidate region determiner configured to determine a candidate region among the boundaries based on a change in pixel brightness at each point of the boundary with respect to the region of interest obtained as a result of the region division;
a three-dimensional transformation unit for constructing a 3D model of a region of interest from the plurality of 2D medical image data based on the region division result; and
A plurality of cross-sections in a predetermined angular direction are generated from the 3D model to include at least a portion of the candidate region, and segmentation performance for a region of interest based on a change in pixel brightness associated with the candidate region in each of the plurality of cross-sections A section selection section that determines this enhanced section as a selected section,
including,
The three-dimensional transformation unit corrects the 3D model based on the selected cross-section,
and the candidate region determiner determines, as the candidate region, a point that does not satisfy a pixel brightness change condition in a vertical direction with respect to the region of interest among points of the boundary. 13. The method of claim 12,
The section selection unit,
generating a plurality of cross-sections in all angular directions with respect to the candidate region;
generating the plurality of cross-sections by selecting a predetermined angular direction in which it is determined that the probability of improving segmentation performance for the region of interest is equal to or greater than a preset threshold probability based on the sampled 3D data for the candidate region;
and generating the plurality of cross-sections by selecting a predetermined angular direction in consideration of anatomical characteristics of the region of interest.

Images