JP2006102125A - Method for identifying region of interest with time in medical image processing apparatus and medical image processing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a medical image processing apparatus and a method for identifying a region of interest with time in a medical image processing apparatus which are easy to operate and achieve a shorter processing time even when segmentation is performed on the basis of large volume of data. <P>SOLUTION: In volumes containing acquired with time three-dimensional image information of a subject, a point in a region of interest within an image acquired at a predetermined time is designated as a seed point. The method comprises a region of interest identifying step for identifying a region of interest in an organ containing the designated point at this point of time by carrying out segmentation, a region of interest enlarging step for generating a region by enlarging the region of interest identified by the region of interest identifying step, and a region of interest detecting step for setting the region enlarged by the region of interest enlarging step in a volume at a next point of time and for detecting the region of interest at the next point of time in the same. By repeating the region of interest growing step and the region of interest detecting step, the of the region of interest changing with time are identified. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a method for supporting diagnosis by a medical image using a time-series data set obtained from a subject, and in particular, a four-dimensional time series obtained using a medical image processing apparatus such as a CT apparatus or an MRI apparatus. The present invention relates to a region-of-interest specification method using a segmentation method of a region of interest (ROI) of a data volume and a medical image processing apparatus.

  Recently, an apparatus for reconstructing a large volume of data obtained from a subject using a computer and displaying an image has been used for diagnosis of the subject. Time-series volume data enables anatomical examination of details of a region of interest in a specific organ such as the heart. However, in order to specify such a region of interest, it is necessary to perform segmentation for each volume, and there has been a problem that it takes a very long time to perform the processing manually and accurately.

  Therefore, segmentation has been automatically performed recently, but anatomically varies depending on the subject, and it is difficult to perform segmentation accurately. Therefore, a method of segmenting an organ including the point by designating one point in the region of interest of the three-dimensional volume data set as a seed point has been considered (for example, see Patent Document 1). In this method, an organ in a region of interest in a time-series three-dimensional volume set is identified by a seed point, contracted, and contracted again by applying a filter expanded on the next three-dimensional volume set. By repeating the operation, there is a method of efficiently looking at the organ of the region of interest.

However, this method also requires processing time and requires considerable user operation.
JP 2002-282235 A

  In the above-described conventional segmentation method for medical images, there is a problem that if a large volume set is to be performed, the operation for that is troublesome and the processing time is also long.

  The present invention has been made in view of the problems of the conventional medical image segmentation method. Even when segmentation is performed based on a large amount of data, the operation is simple and the processing time is short. It is an object of the present invention to provide a method for specifying a region of interest over time in a medical image processing apparatus and a medical image processing apparatus.

  According to claim 1 of the present invention, when one point of interest is designated as a seed point in an image at a predetermined time of a volume obtained with time for three-dimensional image information in a subject, this designation is performed. A region-of-interest specifying step that identifies a region of interest at this time of the organ including the selected point by segment division, and a region-of-interest expanding step that generates a region in which the region of interest specified by the region-of-interest specifying step is expanded, A region of interest expanded by the region of interest expansion step is set as a volume at the next time point, and a region of interest detection step of detecting the region of interest at that time point therein, and the region of interest expansion step and the region of interest Medical image processing characterized by identifying the region of interest that changes over time by repeating detection steps Providing temporal specific method of region of interest in location.

  According to claim 2 of the present invention, in the temporal data acquisition step of obtaining three-dimensional image information in the subject as a volume over time, and in the image of the first volume obtained by this temporal data acquisition step, When one point of the region is designated as a seed point, a region of interest specifying step of segmenting the region of interest at this time of the organ including the specified one point by the segmentation, and the region of interest specifying step A region-of-interest expanding step for generating a region in which the region of interest is expanded; and the region expanded by this region-of-interest expanding step is set as an image of the volume at the second time point, and the region of interest is divided by segmentation at that point of time. A region of interest detection step for detecting the region of interest, and the region of interest expansion step and the region of interest detection By repeating the steps, to provide a time specific method of region of interest in the medical image processing apparatus characterized by identifying the region of interest change over time.

  According to the third aspect of the present invention, the temporal data acquisition step for obtaining the three-dimensional image information in the subject as a volume over time, and the 3 at a predetermined time point from the series of volumes obtained by the temporal data acquisition step. When a volume set generation step for generating a volume set of dimensions and a point of interest in the first volume image generated by the volume set generation step is specified as a seed point, the specified one point A region of interest specifying step of segmenting the region of interest at this time of the organ including the region of interest, a region of interest expanding step of generating a region expanded from the region of interest specified by the region of interest specifying step, and this region of interest expansion The area enlarged by the step is the image of the volume at the second time point. A region-of-interest detection step for setting and detecting the region of interest by segment division at that time, and repeating the region-of-interest expansion step and the region-of-interest detection step to change the region of interest changing over time Provided is a method for specifying a region of interest over time in a medical image processing apparatus, characterized by specifying a region.

  According to a fourth aspect of the present invention, in the method of specifying a region of interest over time in the medical image processing apparatus according to any one of the first to third aspects, the region-of-interest expansion step has the same dimension as the original volume. A volume generation step that generates a volume over time, a voxel initialization step that initializes all voxel values of the volume over time to zero, and a non-zero value for the voxel corresponding to the segmented region of interest. A voxel value setting step for setting a maximum segment value, a maximum / minimum value detection step for detecting respective maximum and minimum values in the three-dimensional coordinates of the segmented region of interest, and a size and shape of the region of interest by a specific segmentation method. The region of interest setting step for setting the region of interest and the subvolume of the region of interest obtained by this region of interest setting step And having a larger processing application step of applying a large processing, the.

  According to claim 5 of the present invention, in the method for temporally specifying a region of interest in the medical image processing apparatus according to any one of claims 1 to 3, the region of interest specifying step includes a search region generating step, a new region of interest Segmentation step for segmenting a region of interest and a region of interest separation step for separating a new region of interest, wherein the search region generation step obtains a maximum value and a minimum value of the three-dimensional coordinates of the expanded region of interest, respectively. A minimum value acquisition step, a volume generation step for generating a temporal volume of the same dimension as the original volume, a voxel initialization step for initializing all voxels of the temporal volume to zero, and a temporal volume Voxel value setting step for setting all of the voxels within the range of the next volume, The segmentation step includes a region-of-interest search step in which a region of interest containing seed voxels based on luminance values is continuously searched in a volume, and region expansion using a region expansion segmentation method to initialize seed points. The region-of-interest separation step comprises a step of setting a voxel other than the region of interest in the volume over time to zero, and a step of setting the voxel ratio to set the voxel of the region of interest of the volume over time to a non-zero value. It consists of steps.

  According to the sixth aspect of the present invention, in the temporal data acquisition means for obtaining the three-dimensional image information in the subject as a volume over time, and the image of the first volume obtained by the temporal data acquisition means When a point of interest is specified as a seed point, the region of interest specifying means for segmenting the region of interest including the specified one point at this time by segment division, and the region of interest specifying means A region-of-interest expanding means for generating a region expanded from the region of interest, and the region expanded by the region-of-interest expanding means is set as an image of the volume at the second time point, and the region of interest is divided by segment division at that time point. A region of interest detecting means for detecting a region, and a region of interest expanding process in the region of interest expanding means and the region of interest detecting means By kicking repeated detection processing of the region of interest, to provide a medical image processing apparatus characterized by identifying the region of interest change over time.

  According to the present invention, it is possible to obtain a method of specifying a region of interest over time in a medical image processing apparatus and the medical image processing apparatus that are easy to operate and require a short processing time.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a flowchart of a method for specifying a region of interest over time according to an embodiment of the present invention. FIG. 3 shows the overall configuration of a medical image processing apparatus 30 according to an embodiment of the present invention that performs this method.

  Each process and overall control control using a storage device 31, for example, a magnetic disk, in which a time-series data set of three-dimensional volume data obtained by a modality device (not shown) is stored, and a data set read from the storage device 31 are used. A central control unit (CPU) 32 for performing the processing, a main memory 33 used for processing of each process performed by the CPU 32, a keyboard 34 for performing direct input manually in each processing, and a mouse 35 for performing input by moving operation A controller 36 for converting the input from the mouse 35 into a control signal; a display memory 37 for temporarily storing the processed signal for display; and a display connected to the display memory 37 for displaying an image for medical diagnosis Device 38, storage device 31, central control unit (CPU) 32, main memory Li 33, keyboard 34, and a common bus 39. for exchanging signals between the controller 36 and the display memory 37.

  The method of specifying the region of interest over time in the medical image processing apparatus 30 is as follows: (1) selecting a seed point in the first volume, (2) segmenting the region of interest, (3) expanding the region of interest, (4) Overlay the region of interest expanded with the new volume, (5) Extract the region of interest in the new volume, (6) Repeat steps (3) to (5) for all volumes . (1) Selection of a seed point in the first volume is performed by moving and clicking the mouse 35 while viewing an image displayed on the display device 38, and thereafter (2) to (2) 6) is performed by inputting from the keyboard 34 while watching the screen of the display device 38, or by dragging or clicking the mouse 35.

  Hereafter, each said process is demonstrated based on FIG.

(1) Selecting a seed point in the first volume In step S101 in FIG. 1, a four-dimensional data set is loaded into the storage device 31 of the medical image processing apparatus 30. Here, the four-dimensional data set is a series of data sets obtained by taking a volume (a three-dimensional image by the three-dimensional data set) defining a space corresponding to each axis of x, y, and z every minute time Δt. This four-dimensional data set is a 4D data set of the heart obtained by an image diagnostic apparatus such as an X-ray CT scanner in order to segment the left ventricle of the subject in each data set, for example.

  Based on the four-dimensional data set loaded as described above, the medical image processing apparatus displays a three-dimensional image at the first time point. The user manually selects one seed point in the first volume in which the predetermined region of interest is clearly displayed. One voxel in the region of interest is selected by moving the cursor to an organ to be designated as the region of interest on the display screen and clicking the mouse button at that point (step S102).

  The selected voxel becomes the seed point of the initial value of the segment division process displayed in the volume at the next time point.

(2) Segmentation of Region of Interest In FIG. 2A, when the seed point 21 is selected by moving the cursor on the screen of the display device 38 with the mouse 35 and clicking, the region of interest 22 is Segmentation is performed by methods used, such as Rafael C. Gonzalez [1077] “Digital Image Processing”, Addison-Wesley Publishing Company, Inc. region growing. This region enlargement method is performed by bringing all voxels having a certain common characteristic, for example, a difference in luminance with respect to the seed point thereof, to a predetermined value or less into one lump.

  Regions of interest are successfully segmented by this method if the boundaries are defined in detail. When boundaries are not clearly defined, it is normal to first correct them as much as possible and process the volume.

  Segmentation divides all voxels of the volume into groups belonging to the region of interest and other groups (background) outside the region of interest. We performed segmentation of the left ventricle, for example, in the first volume in this way. Since this initial result is the source of subsequent segmentation of the region of interest, the user should verify it before processing the next step.

  In order to facilitate subsequent processing, the segmented region of interest generates a temporary binary mask of the same dimensions as the original volume (step S104), as shown in FIG. All of the voxels are separated from the rest of the volume and become background values except for the voxels (predefined non-background values) corresponding to the segmented region of interest.

(3) Enlarging the region of interest Once the first region of interest is segmented and divided, it is used to segment the region of interest in the next volume. For this reason, the first region of interest is shown in FIG. 2 (c) using a mathematical form operation as described, for example, in Jean Serra [1983], “Image Analysis and Mathematical Morphology”, Academic Press. As shown, it is enlarged by a predetermined amount (step S105).

  Mathematical morphologies are well known in the field of image processing, and the enlargement operation produces an “enlarged” version of the original volume. The volume is expanded three-dimensionally, but the shape of the region of interest remains the same. Since the next volume is a volume after a minute time Δt from the previous volume, it is assumed that the shape of one volume and the next volume has changed only slightly.

The reason for enlarging the volume is to generate a search area including a region of interest in the next volume. How much to enlarge depends on the specific application, and the region of interest in the next volume is taken large enough to be included in the enlarged volume.

  However, if the image is enlarged too much, a region not related to the same voxel luminance is included in the search region as a region of interest. One way to automatically determine the amount of enlargement and reduce the above problem is to start with the amount of enlargement from zero, and if the next region of interest touches the border of the search region, enlarge the enlargement region and perform the enlargement process. Is to repeat. These volume expansion amounts are controlled by inputting from the keyboard 34.

  The enlarged version becomes a binary mask with voxels in the potential region of interest component of the mask signal in the next volume.

(4) Overlaying the expanded region of interest with a new volume As shown in FIG. 2 (d), the expanded binary mask is used as a search region for the region of interest of the next temporal volume. This is done by obtaining maximum and minimum values in all three-dimensional (x, y, z) coordinates (step S106). These values are used as a search start point and a search end point in each search direction.

  If the assumptions as described above are satisfied, it is guaranteed that the region of interest in the next volume is included in the enlarged mask, and therefore voxels outside that mask can be ignored.

  In the heart data set, if the first region of interest was obtained in the left ventricle during diastole, then the ventricle in the next volume will be a small volume because it is going into systole.

  If the first region of interest is selected during systole, the amount of expansion is set sufficiently large so that the next region of interest is completely contained within the enlarged version of the previous region of interest. A search for a region of interest in the second and subsequent volumes is performed automatically.

(5) Extract the region of interest in the new volume Find the seed point in the next volume (step S107; this process is done automatically) and once the subvolume containing the region of interest is generated, the region of interest It can be easily and automatically segmented (step S108). One way to do this is to group all voxels that do not belong to the region of interest, i.e. voxels with different values from the background voxels and voxels of the region of interest. The remaining voxels correspond to the region of interest in the new volume. Another method is to continuously examine the voxels in the subvolume until a value corresponding to the region of interest is found and use it as a seed point for other region expansion processes.

  Voxels not included in the segmentation result are set to the background value. The final result of both methods above is a subvolume corresponding to the new region of interest.

  A binary mask is formed based on this subvolume (step S109).

(6) Repeat steps (3)-(5) for all of the volumes When a new region of interest is found, it is expanded and used to find the region of interest in the next volume over time. The process of enlarging a region of interest, using it as a search mask in the next volume, and finding a new region of interest in the search subvolume defined by that mask is repeated for each volume in the time-lapse data set.

In this way, the region of interest can be identified and observed over time on the screen of the display device 38 of FIG. We obtained final results for segmentation of the left ventricle at each heart volume.

  A series of volumes obtained at successive time points has the effect of giving doctors valuable information for diagnosis and treatment in order to visualize and analyze the interior of the subject. Segmentation of the region of interest in each volume is a monotonous and very time consuming task, but the present invention is a time-lapse data set with minimal user involvement, making the region of interest a fast and semi-automatic region of interest over time. A specific method was obtained.

The figure which shows the flowchart of the time-dependent identification method of the region of interest in the medical image processing apparatus in one Embodiment of this invention. The figure for demonstrating the specific process of the temporal identification method of the region of interest in the medical image processing apparatus in one Embodiment of this invention. The figure which shows the example of whole structure of the medical image processing apparatus in one Embodiment of this invention.

Explanation of symbols

21... Seed point,
22 ... Region of interest (ROI),
30 ... Medical image processing apparatus,
31 ... Storage device,
32... Central control unit (CPU),
33 ... main memory,
34 ... Keyboard,
35 ... mouse,
36 ... Controller,
37 ... display memory,
38 ... Display device,
39: Common bus.

Claims (6)

When one point of a region of interest is specified as a seed point in an image at a predetermined time of a volume obtained by obtaining three-dimensional image information in a subject over time, this organ of the specified one point is included. A region of interest specifying step of specifying a region of interest at a time by segmentation;
A region of interest expansion step for generating a region in which the region of interest specified by this region of interest specification step is expanded;
A region-of-interest detection step of setting the region expanded by the region-of-interest expansion step to a volume at the next time point and detecting the region of interest at that point in time; and
A method of specifying a region of interest over time in a medical image processing apparatus, wherein the region of interest changing over time is specified by repeating the region of interest expansion step and the region of interest detection step. A temporal data acquisition step for obtaining three-dimensional image information in the subject as a volume over time;
When one point of the region of interest is designated as a seed point in the image of the first volume obtained by the time data acquisition step, the region of interest at this time of the organ including the designated point is segmented A region of interest identification step identified by
A region of interest expansion step for generating a region in which the region of interest specified by this region of interest specification step is expanded;
A region of interest detection step of setting the region expanded by the region of interest expansion step to an image of a volume at a second time point and detecting the region of interest by segment division at that time point,
A method of specifying a region of interest over time in a medical image processing apparatus, wherein the region of interest changing over time is specified by repeating the region of interest expansion step and the region of interest detection step. A temporal data acquisition step for obtaining three-dimensional image information in the subject as a volume over time;
A volume set generation step for generating a set of three-dimensional volumes at a predetermined point in time from a series of volumes obtained by the time data acquisition step;
When one point of the region of interest is specified as a seed point in the image of the first volume generated by the volume set generation step, the region of interest at this time of the organ including the specified one point is segmented. A region of interest identification step to identify;
A region of interest expansion step for generating a region in which the region of interest specified by this region of interest specification step is expanded;
A region of interest detection step of setting the region expanded by the region of interest expansion step to an image of a volume at a second time point and detecting the region of interest by segment division at that time point,
A method of specifying a region of interest over time in a medical image processing apparatus, wherein the region of interest changing over time is specified by repeating the region of interest expansion step and the region of interest detection step. The region of interest expansion step includes
A volume generation step that generates a volume over time of the same dimensions as the original volume;
A voxel initialization step that initializes all of the voxel values of the volume over time to zero;
A voxel value setting step for setting a voxel value corresponding to the segmented region of interest to a non-zero value;
A maximum / minimum value detecting step of detecting each maximum value and minimum value in the three-dimensional coordinates of the segmented region of interest;
A region of interest setting step for setting the size and shape of the region of interest by a specific segmentation method;
An enlargement process applying step for applying an enlargement process to the sub-volume of the region of interest obtained by the region of interest setting step;
4. The method of specifying a region of interest over time in a medical image processing apparatus according to claim 1, comprising: The region-of-interest identification step includes a search region generation step, a segmentation step that segments a new region of interest, and a region-of-interest separation step that separates a new region of interest.
The search area generation step includes:
A maximum / minimum value acquisition step for acquiring the maximum value and minimum value of the three-dimensional coordinates of the enlarged region of interest, a volume generation step for generating a temporal volume of the same dimension as the original volume, and the temporal volume A voxel initialization step that initializes all of the voxels to zero, and a voxel value setting step that sets all of the voxels of the volume over time within the range of the next volume,
The segmenting step includes
A region-of-interest search step for continuously searching in the volume a region of interest that includes seed voxels based on luminance values;
A region expansion step using a region expansion segmentation method to initialize the seed points,
The region of interest separation step comprises:
A voxel zeroing step to zero voxels outside the region of interest in the temporal volume;
Voxel ratio zero setting step for setting the voxel of the region of interest of the temporal volume to a non-zero value,
4. The method of specifying a region of interest over time in a medical image processing apparatus according to claim 1, wherein Temporal data acquisition means for obtaining three-dimensional image information in the subject as a volume over time;
When one point of the region of interest is specified as a seed point in the first volume image obtained by the time data acquisition means, the region of interest at this time of the organ including the specified point is segmented A region of interest identifying means identified by
A region-of-interest expanding means for generating a region in which the region of interest specified by the region-of-interest specifying means is expanded;
A region of interest enlarged by the region of interest enlargement means is set in the image of the volume at the second time point, and the region of interest detection means detects the region of interest by segmentation at that time point,
A medical image processing apparatus that identifies a region of interest that changes over time by repeating a region-of-interest expansion process in the region-of-interest expansion unit and a region-of-interest detection process in the region-of-interest detection unit.

Images