JP5155622B2 - Method for extracting specific region of medical image and medical image processing apparatus - Google Patents

Description

  The present invention relates to a method for extracting a specific region on an image, for example, a specific organ of a medical image, using a computer.

There is Patent Document 1 as a conventional specific organ extraction method.
JP 2005-296479 A

In Patent Document 1, organ extraction is performed not on a pixel basis but on an area unit having a certain spread. Here, a certain spread refers to a sense area determined by a preset circle (others, such as a sphere, a cone, etc.) figure (this is called a sense figure). For example, a point of an image that seems to be a specific organ on the screen (for example, coordinates near the center of the organ) is designated, and a circular figure with a radius R is set with this point as the center point. An attribute value Vi of a pixel group in the circular figure is obtained and compared with a reference attribute value Vth that is a predetermined extraction determination value. The attribute value is an average value or standard deviation of pixel density (pixel value), and the reference attribute value is a reference threshold value indicating whether or not the pixel is in a specific organ. If this reference attribute value is satisfied (Vi <Vth or conversely Vi> Vth or Vth1 <V <Vth2, where Vth, Vth1, and Vth2 are threshold values), the pixel group in the circle is a specific organ. It is determined that the area indicates. If it is determined that the region indicates a specific organ, all the pixel groups in the circular figure are recognized and extracted as pixels in the specific organ. Further, a circular figure slightly smaller than the circular figure is set, and the pixel group in the set small circular figure is sorted (extracted) into the pixel group of the specific organ. The reason for setting a slightly smaller circular figure is that the original circular figure itself is too coarse, and by setting a smaller circle, it can be surely selected as an intra-organ pixel group.
The above is a method for checking whether or not the radius R is a specific organ from the center point. The center point of this circle is moved, and the same circular figure is set for each movement point to determine whether or not the same organ is in the same organ. Check that.

  The movement of the center point is based on the area of the extracted circular figure as the extracted specific area. For example, a central point is set at one point on the circumference of a circular figure, and a specific organ is extracted in the above-described manner based on the circular figure at each central point. In this way, organ points are extracted by setting center points one after another on the circumference. Thus, when the first specific organ is extracted based on the first center point, the search area is expanded to the surrounding area based on the extracted organ. This technique is taken one after another, and organ regions are extracted in an expanded manner. If a region that is not a specific organ is hit, the reference attribute value is not satisfied, and the expansion stops.

  In addition to the radius R, there are also examples in which two radii R1 and R2 are given and the circle figure is R1 <R <R2.

  Patent Document 1 obtains an attribute value in a circular figure and compares it with a reference attribute value. However, the specific organ region has a certain size, and the pixel density used as the attribute value calculation source varies in the region. In the extraction method described above, the pixel group in the organ is changed to the organ due to the variation in the pixel concentration. It may occur that the pixel group is outside (so-called expansion leakage), or the pixel group outside the organ is made a pixel group inside the organ (so-called overexpansion).

  The organ to be extracted has a feature that even if there are various variations in pixel density, the variation method has relatively high continuity from adjacent pixel densities.

  Therefore, an object of the present invention is to provide an extraction method that enables highly accurate organ extraction that does not cause expansion omission or overexpansion by using nearby pixel densities.

The present invention sets an extraction judgment value when a computer is used to extend an already extracted area in a medical image to the periphery thereof, sets a predetermined figure near the boundary of the already extracted area, Or, the internal pixel group is divided into those belonging to the already extracted region and those not belonging to the unextracted region, and the pixel attribute values of the pixel group divided as belonging to the already extracted region and those belonging to the unextracted region are divided. A difference pixel attribute value that is a difference between the pixel attribute value of the selected pixel group and a pixel group that has been classified as belonging to the unextracted area based on a comparison between the difference pixel attribute value and the extraction determination value. It is characterized by performing an extension to be added.
Preferably, according to the movement of the predetermined shape on said predetermined graphic path straight set as the moving path of the extraction decision value based on the attempt to extend from the division is set.

Furthermore, the present invention provides an extraction determination value determination method for determining an extraction determination value.
Setting a temporary extraction judgment value;
Setting a trial movement range of the same sense figure in the medical image for determining the extraction determination value;
A difference pixel attribute value between a pixel group belonging to the already extracted region and each pixel attribute value is obtained for the sense figure at each position in the trial movement range, and the difference pixel attribute value is compared with the provisional extraction determination value. It is determined whether or not the difference pixel attribute value satisfies the provisional extraction determination value. If satisfied, the sense figure is moved to the next position in the trial movement range, and if not satisfied, the movement is stopped at that position. Steps,
By the above movement, it is monitored whether or not the sense figure is stopped at the boundary of the specific region to be extracted of the medical image for the extraction determination value, and if the stop is stopped at the boundary, the temporary extraction determination value is A decision value is output as a decision value, and the temporary extraction decision value is outputted as an erroneous value if it is stopped inside or outside without stopping at the boundary; and
An extraction determination value determination method in a specific region extraction method comprising:

Furthermore, the present invention uses a computer to determine whether or not to expand an already extracted extracted area in a medical image to the surrounding area, using a sense graphic having a plurality of pixel group sizes as an extraction determination value. In the extraction determination value determination method in the specific region extraction method to be compared,
Setting a trial movement range of a sense figure having the same multiple pixel group size as described above in a medical image for determining an extraction determination value;
Obtaining pixel attribute values for sense figures at each position within the trial movement range;
The obtained pixel attribute value is superimposed on the medical image for determining the extraction determination value, and a trend is displayed along the movement range.
Determining and inputting an extraction determination value for extracting a specific region from the trend content of the pixel attribute value along the displayed movement range;
An extraction determination value determination method in a specific region extraction method comprising:

  According to the present invention, it is possible to accurately determine whether or not an unextracted region belongs to an organ region based on a difference between a pixel value of an adjacent already extracted region and a pixel value of an unextracted region. Extraction is possible.

  FIG. 2 shows an example of a system diagram for realizing the organ region extraction method of the present invention. This system includes a CPU 1, a main memory 2, a magnetic disk 3, a display memory 4, a CRT (including a liquid crystal display device) 5, a controller 6, a mouse 7, a keyboard 8, and a common bus 9 connecting them, and a CT device 11. , LAN10. The devices 1 to 8 and the bus 9 constitute a so-called computer system, and the CPU 1 extracts a hand organ region using programs and data in the main memory 2 and the magnetic disk 3. The means for supporting them is 5 to 8 including the display memory 4. The display memory 4 temporarily stores data to be displayed on the CRT 5 and performs display on the CRT 5. The mouse 7 is an instruction on the screen on the CRT 5 and is connected to the bus 9 via the controller 6. The keyboard 8 performs various input settings, data input, and operation instructions.

  The CT apparatus 11 acquires the transmission data from the subject and obtains CT tomographic image data. The LAN 10 is connected to a common bus 9 and performs mutual communication between the CT apparatus 11 and the computer system via the common bus 9.

  Although the CT apparatus 11 obtains a CT tomographic image, there may be a processing distribution example in which the computer system connected to the common bus 9 has an image processing function for obtaining a CT tomographic image as a function until obtaining the shadow data. In this case, the LAN 10 transmits the shadow data.

  In FIG. 2, if an ultrasonic apparatus or an MRI apparatus is connected instead of the CT apparatus 11, an ultrasonic tomographic image and an MRI tomographic image are targeted.

  FIG. 1 is an example of a processing flow for organ region extraction, which corresponds to the processing by the computer system in FIG. In the flow F1, a difference reference attribute value base that is a determination value for performing extraction determination of an unextracted region is set. The unextracted region is a part on the CT tomographic image to be extracted as an organ from now on, and the reference attribute value is an attribute value to be used as a reference. Various values such as an average CT pixel value, a standard deviation value, a maximum pixel value, a minimum pixel value, and a setting example of one value or two or more values (for example, a CT pixel value) And standard deviation value). The difference is a difference between the attribute value of the unextracted area and the attribute value of the previously extracted area. In addition, there is an example in which the reference attribute value is given as a width with two values of large and small other than one value.

  The flow F2 is processing based on the premise that an already extracted region exists, and sets the center point of a circular figure near the boundary of the already extracted region, for example, on the boundary line. This setting is made by clicking the menu with the mouse 7 under the display state of displaying the CT tomographic image on the CRT 5 / displaying the extracted area on the image / displaying the center point setting menu on a part of the screen. In addition, it is realized by inputting one point designated by the operator on the boundary of the already extracted region as the center point.

  Upon receiving such center point setting input, the computer creates a circular figure centered on the center point on the boundary line of the extracted region as data.

  In the flow F3, since the circular figure extends over both the already extracted area and the non-extracted area outside the area in the data processing, the flow F3 is divided. Since the already extracted area is determined and the circular figure is determined, the overlapping area between the two is set as the already extracted area, and the remaining circular figure area is set as the unextracted area.

  The management of the already extracted area can be realized by providing an extraction management area having “1” for already extracted and “0” for not extracted in a memory area having an image size corresponding to the display image data. Alternatively, there is a method of adding a flag indicating whether or not the pixel value data has already been extracted.

In the flow F4, the attribute value of the pixel value in each area is obtained in each of the already extracted area E1 and the unextracted area E2 divided in the flow F3. Examples of attribute values are as follows.
(1) Average values a1 and a2 of the CT pixel values of the regions E1 and E2
(2) Standard deviation values b1 and b2 of the CT pixel values of the regions E1 and E2
(3) Maximum values c1 and c2 of the CT pixel values of the regions E1 and E2
(4) Minimum values d1 and d2 of the CT pixel values of the areas E1 and E2
It is a value indicating the characteristics of the extraction area.

  Next, the difference r between the attribute values of the areas E1 and E2 thus obtained is obtained. The average value is (c1-c2), the standard deviation value is (b1-b2), and so on. The reference attribute value r0 as the determination value in the flow F1 corresponds to the difference attribute value r. Therefore, in the flow F4, the calculated difference attribute value r and the reference attribute value r0 set in the flow F1. Compare with. As a result of the comparison, if the reference attribute value r as the determination value is satisfied, it is determined that the unextracted region E2 constitutes a part of the extracted organ. Here, satisfying is in accordance with a condition determined in advance such as (a1-a2)> r0 or (a1-a2) <r0. There is also an example in which two reference attribute values r1 and r2 are given and r1 <(a1-a2) <r2.

  In the area E2, all of the internal pixel groups (more precisely, all the pixel groups in the unextracted area other than the pixel group determined as the internal already extracted area) are regarded as a part of the extracted organ. In addition to this, when the circular figure as the sense area is large, a new circular figure with a radius smaller than the circular figure, for example, about 7 minutes or about half, is set, and this new circular figure area is specified as the extracted organ. .

  The calculation of the attribute value in the flow F4 is based on all the pixels in each of the areas E1 and E2, but in order to simplify the processing, some representative points are selected instead of all the pixels. There is also a method of obtaining the attribute value based on the representative point. Furthermore, as one of the ideas, how to calculate the attribute value based on each point on the circumferential line of the circular figure in the already extracted area and each point on the circumferential line of the circular figure as the unextracted area There is also.

  As a method of dividing the extracted area and the unextracted area in the flow F3, since the extracted area is fixed and the center point and the radius of the circular figure are fixed, the inside of the circular figure (including the circumference) is included. If all “1” is given as data and the logical product with the circular graphic data including the extraction area is taken, data of the already extracted area remains. The remaining area of the circular figure other than the already extracted area is automatically determined as the unextracted area.

  There is an example in which the radius R of the circular figure in the flow F2 is set to be smaller as it goes to the outside as compared with the radius at the first center point as the origin. This is a method of continuously subtracting −constant value−ΔR as it is away from the origin, fixing it to R up to a certain distance, and selecting and setting R ′ <R where R ′ <R after exceeding a certain distance. There is a way to make it gradually smaller as it gets farther away. By reducing the radius R of the circular figure as the extension to the outside progresses, the unevenness of the edge portion of the extraction region is reduced, and the organ region can be extracted with higher accuracy.

  Note that the processing flow in the figure is based on the premise of the existence of an already extracted region, and therefore the first extraction process has already been performed. In this first extraction process, a starting point is set at an arbitrary position in the CT image, for example, the center, and a circular figure is set around the starting point. Next, at least one of an average value, a standard deviation value, a maximum CT pixel value, and a minimum CT pixel value of the CT pixel values in the circular figure is obtained by data processing, and compared with a predetermined determination value. And determine whether it is inside the organ. If it is in an organ, it becomes the first extraction region. Do until you find this. Thus, the first extraction area is determined and the process proceeds to FIG.

  FIG. 3 is a flow chart for checking whether or not the reference attribute value as the determination value in the flow F1 in FIG. 1 is correct, or a processing flowchart for obtaining the correct reference attribute value in advance. The reference attribute value as a determination value is an important factor for organ region extraction, and if there is an error, it will cause extraction failure. Therefore, it is necessary to try and obtain it in advance. FIG. 3 shows the flow at that time.

  In the flow F1 of FIG. 3, first, a difference attribute value as a determination value, which the operator seems to be optimal or is empirically recognized, is set with a keyboard or a cursor. .

  The flow F2 sets a circular figure with a radius R as a central point at an arbitrary starting point in the organ of the image for extracting the judgment value, and moves the circular figure within the organ by shifting the central point. Move the figure so that it goes out of the organ. For example, the center point is linearly moved outward. Then, region extraction is performed for each center point in the flow F3, and the extracted region is displayed in the flow F4.

  FIG. 4 shows an example of linear movement of this center point. FIG. 4 shows a display screen of the CRT 5 and displays a trial CT image. In this example, the organ A is to be extracted, the organ B is not extracted, and both have a part in contact with each other.

  The circle figure initially set as the start point is the figure 21, and it is determined whether or not the circle figure 21 is in the organ A based on the start determination value. If it is within A, the first organ region has been extracted. Next, the center point is shifted along the linear movement direction to create the next circular figure, and this time, organ region extraction processing is performed according to the same processing as in FIG. The determination value at this time is an attribute value as a difference between the immediately extracted region and the unextracted region. If the organ region can be extracted, the next center point is moved and the same processing is performed. In this way, the determination is performed according to the difference attribute value as the determination value while moving the center point on the straight line one after another. If it is within the same organ and the difference attribute value as the determination value is the correct determination value, the circle graphic advances to the edge of the organ. Then, if the circular figure 23 at this edge stops and does not proceed any further, it can be seen that the difference attribute value as the determination value is a correct value.

  On the other hand, when the circular figure further advances and enters the organ B, it can be seen that the difference attribute value as the determination value is incorrect. At this time, a new difference attribute value is set as a determination value, and the processes shown in FIGS. 3 and 4 are performed again. Thus, such an attempt finds an attribute value that stops at the edge.

In FIG. 4, an operation screen (GUI) 24 is displayed and used to input processing parameters. In the processing parameter, two examples of a circle diameter (2R) or radius (R), a difference average CT value, and a difference standard deviation value (STD) can be set as difference attribute values. In addition, a test run instruction, setting and termination instruction menu is displayed. Various clicks on the display screen 24 are performed with the mouse 7 and parameters with the keyboard 8.
By using such an operation screen, trial processing can be realized quickly and appropriately.

  Although the circular figure as the sense figure is linearly moved, there are examples of circular movement and parabolic movement depending on the shape and part of the extracted organ.

  Although the attribute value as the determination value is the difference attribute value shown in FIG. 1, FIG. 3 and FIG. 4 also show the determination value for determining the attribute value obtained only from the circular figure, not the difference. Applicable.

  FIG. 5 shows a processing flow for obtaining a reference attribute value as a determination value when an attribute value obtained only from such a circular graphic is used.

  In the flow F1 of FIG. 5, the linear direction is set as in FIG. The difference from FIG. 3 is that the extraction area is not determined during the movement of the center point. Instead, for each moving center point, an attribute value (average value or standard deviation value) based on the pixels in the circular figure is obtained and displayed on the screen. An attribute value is obtained as a determination value for extracting an unextracted area from the display screen. Such processing is realized by the flows F2 and F3 in FIG.

  On the display screen of FIG. 6, a trend graph screen 25 associated with the linear movement of the circle is displayed. An average CT value standard deviation value (STD) was obtained and displayed for each circle figure at each center point. From this display screen, an attribute value is selected as a judgment value suitable for extraction of the organ A to be extracted. For example, a value in the range of m1 to m2 is selected as a judgment value for the extracted organ.

FIG. 7 is an example of a display screen displayed during the processing of the processing flow of FIG. 1. An already extracted area 30 already exists for the organ A, and a circular figure with a radius R at one point P1 of the boundary line of this area. This is an image when 31 is set. FIG. 7A shows an example of setting a circular figure 31, and FIG. 7B shows an image in which the entire area 32 of the circular figure 31 is recognized as an organ area.
As is apparent from FIG. 7B, a new extraction area is determined by the circular figure 31.

FIG. 8 is a schematic diagram of a display screen when the extraction region is expanded while forming a plurality of circular figures 31 one after another along the boundary line of the already extracted region 30. FIG. 8A shows an example of movement of the circular figure 31, and FIG. 8B shows an extension example 32 after it is determined that it is within the extraction region.
Thus, it can be seen that the extraction regions are obtained one after another by moving the circular figure on the boundary line.

  In the above embodiment, a circular figure is used for extraction expansion, but other figures such as a spherical figure and a cone can be used. Further, although an example of the X-ray CT apparatus is described, the present invention can also be used for MRI and ultrasonic tomographic images.

  Although the already extracted area is an internal image group including the peripheral line of the sense graphic, it may be an adjacent extracted area and outside the sense graphic.

It is a processing flow example figure of organ extraction of the present invention. It is an example figure of a processing system of organ extraction of the present invention. It is a processing flow example figure for determination value determination of the present invention. FIG. 4 is an example of a display screen in the process of FIG. 3. It is another example of processing flow for determination value determination of the present invention. FIG. 6 is an example of a display screen in the process of FIG. It is an example of a display screen showing organ region expansion extraction of the present invention. It is a display screen model example figure when organ area | region expansion extraction of this invention is performed continuously.

Explanation of symbols

1 CPU
2 Memory 3 Auxiliary memory 4 Display memory 5 CRT
11 CT equipment

Claims (8)

A method for extracting a specific area by using a computer to extend an already extracted area in a medical image to the periphery thereof,
Setting an extraction judgment value;
A setting step for setting a predetermined figure near the boundary of the extracted area;
A division step of dividing the surrounding or internal pixel group of the predetermined figure into those belonging to the already extracted region and those belonging to the unextracted region;
A calculation step for obtaining a difference pixel attribute value that is a difference between a pixel attribute value of the pixel group divided as belonging to the already extracted area and a pixel attribute value of the pixel group divided as belonging to the unextracted area When,
Based on the comparison between the difference pixel attribute value and the extraction determination value, an extension step of adding a pixel group that is divided as belonging to the unextracted region to the already extracted region;
In a method for extracting a specific region of a medical image comprising:
In the setting step, the extraction determination value is set based due to the movement of the predetermined shape in the set straight on the path, the attempts from the dividing step to said expansion step as the movement path of the predetermined graphic A method for extracting a specific area of a medical image. The path is a path that spans the first organ region and the second organ region,
In the setting step, the predetermined figure is moved on the route, the trial extraction determination value is changed, the trial is repeated, and the trial extraction determination value at which the expansion step ends at one region boundary is extracted. The method for extracting a specific region of a medical image according to claim 1, wherein the specific value is a determination value.   The method for extracting a specific area of a medical image according to claim 1, wherein the predetermined graphic is any one of a circular graphic, a spherical graphic, and a conical graphic.   4. Each pixel attribute value is at least one of an average value or standard deviation value of a pixel group, a maximum pixel value, and a minimum pixel value, and a difference pixel attribute value is a difference value thereof. The specific region extraction method for medical images according to any one of the above.   The method for extracting a specific area of a medical image according to any one of claims 1 to 4, further comprising a display step of displaying an area extracted by the trial. A medical image processing apparatus that extends an already extracted region in a medical image to the periphery thereof,
An input unit that accepts the setting of the extraction determination value and the setting of a predetermined graphic placed near the boundary of the extracted region;
A pixel attribute value of a pixel group divided around the predetermined figure or belonging to the already extracted area and divided to belong to the already extracted area, and the unextracted pixel group A difference pixel attribute value that is a difference between a pixel attribute value of a pixel group divided as belonging to an area and a difference between the difference pixel attribute value and the extraction determination value is calculated, and the difference attribute value belongs to the unextracted area A calculation unit that adds and expands the pixel group divided as
A medical image processing apparatus comprising:
The input unit, characterized in that the by the movement of the predetermined shape in the predetermined shapes on a linear path that is set as the travel route, extraction determination value based on the attempt to the extension from the division is input A medical image processing apparatus. The path is a path that spans the first organ region and the second organ region,
The input unit moves the predetermined figure on the path, changes the trial extraction judgment value, repeats the trial, and extracts the trial extraction judgment value at which the expansion ends at one region boundary. The medical image processing apparatus according to claim 6, which is input as a determination value.   The medical image processing apparatus according to claim 6, further comprising a display unit configured to display a region extracted by the trial.

Images