JP4344825B2 - Irradiation position verification system - Google Patents

Description

  The present invention relates to an irradiation position verification system for assisting in aligning an irradiation site according to an irradiation plan in radiotherapy.

In recent years, the number of cancer patients has increased rapidly due to the arrival of an aging society. For this reason, the increase in the number of patients requiring radiotherapy is remarkable, and it is said that one in four people will receive radiotherapy.
Now, with the advancement of diagnostic imaging apparatuses, cancer detection accuracy has been improved. Advances in radiotherapy technology have also improved radiotherapy accuracy, and now it is possible to target cancer cells themselves by radiotherapy (eg, radiosurgery, gamma knife).
However, the position of the irradiated region is subjectively confirmed by visual observation, and therefore, individual differences due to the observer's experience and reproducibility are problematic. Therefore, it is desired to develop a system that supports confirmation of an irradiation site of radiotherapy.
Non-Patent Documents 1 and 2 are systems that support the confirmation of the irradiation site of radiation therapy that has been announced. Non-Patent Document 1 introduces a system that calculates a deviation by a computer after manually tracing a collation index. However, the problem is that the reproducibility is low due to manual tracing. Non-Patent Document 2 introduces that a collation index is automatically determined for a X-ray simulation image and a portal image, and a shift is calculated by a computer. However, the image (DRR) used for the treatment plan is not targeted, and it cannot be said that the handling is necessarily good.

John R. Van Sorensen de Koste, Hans CJ de Boer, M. Sc., Regine H. Schuchhard-Schipper, et al: Procedures for high precision setup verification and correction of lung cancer patients using CT-simulation and digitally reconstructed radiographs (DRR ), Int. J. Radiation Oncology Biol. Phys. 55 (3), 804-810, 2003 Hiroyuki Akazawa, Nobuyuki Nakamori, Atsuko Shiomoto, et al .: Development of an automatic portal image verification program for radiation therapy. Journal of Japanese Society of Radiological Technology, 60 (1), 101-110, 2003

  The present invention is to provide an irradiation position verification system for supporting position verification of an irradiation site according to an irradiation plan with excellent convenience and good reproducibility.

In order to achieve the above object, the present invention provides an irradiation position verification system for aligning an irradiation site using a radiation treatment plan image (DRR) and an irradiation field confirmation image (LG) photographed before treatment. The shape data for forming the DRR and irradiation field (FOV), the image acquisition means for acquiring the LG image, the pre-processing means for matching the size and position of the acquired DRR and LG image, Edge enhancement means for enhancing the edges of the DRR and LG images, FOV edge removal means for removing FOV edges in the LG using the acquired shape data, and binarizing the DRR and LG images 2 A binarizing means, a thinning processing means for thinning edges of binarized DRR and LG images, a collating means for detecting a deviation between the DRR and LG, and a display unit for displaying the collation result And a step.
Since the FOV edge in the LG, which is unnecessary for the collation, is removed, the collation can be performed accurately.

The image processing apparatus further includes a landmark determination unit that creates a mask by expanding edges in the image after the thinning process of the DRR, and removes an unnecessary edge of the LG using the mask, and the collating unit includes the landmark It is good to collate by LG and DRR which removed the unnecessary edge by the determination means. By this process, since the landmarks to be collated are limited, the collation can be performed appropriately.
Further, the image processing apparatus includes a noise removing unit that removes noise from the binarized LG image, and a unit that removes small edges from the DRR image and the LG image after noise removal. Since the edges are limited to appropriate ones, matching can be performed more appropriately.
The collation means may determine a range (template) in which LG collation is performed and obtain a collation position within a search range defined in the DRR.
Further, manual selection means for manually selecting an edge to be used for collation is provided, and the collation means can collate the selected edge of LG with the selected edge of DRR.
As the above-described irradiation position verification system, a computer program for causing a computer system to function and a recording medium storing the program are also the present invention.

  With the irradiation position verification system of the present invention described above, the position of the irradiation site can be verified with high accuracy and reproducibility, and appropriate radiotherapy can be performed.

Embodiments for carrying out the present invention will be described with reference to the drawings.
In the irradiation position collation system of the present invention, a radiotherapy plan image (digitally reconstructed radiographs), which is an image obtained by reconstructing a plurality of CT image data along the pyramid of the treatment apparatus under the same geometric conditions as those of the X-ray photograph. : DRR) and irradiation field confirmation image (lineacgraphy: LG) taken before treatment with the weight of the treatment device, and collating them by computer processing, supporting the position collation of the irradiated part . The DRR, which is a radiation treatment plan image, is an image created when a treatment plan is created. In addition, the irradiation field confirmation image (LG) includes an irradiation with a medical linear accelerator (Linear accelerator) used for actual radiotherapy for the purpose of confirming the isocenter position and the field of view (FOV) shape. Using a multi-leaf collimator (MLC) to form a field, the image is taken with the same geometric conditions as the actual radiation treatment, but with very little X-ray energy compared to the actual radiation treatment. Yes.

FIG. 1 is a flowchart showing the operation of the irradiation position verification system according to the embodiment of the present invention. In each process, there is also a process performed only on the LG image.
The process of the irradiation position verification system will be described with reference to the flowchart of FIG. 1 and FIGS.

  First, the DRR that is a reference image and the LG that is a collation image are read into the system (S102). The shape data of MLC is also acquired with DRR. Since the two image data usually have different sizes, the sizes are made the same as the preprocessing (S104). An example of DRR having the same size (FIG. 2A) and an example of LG (FIG. 2B) are shown. Since the processing time increases when the size of the image increases, it is desirable to perform the processing at the minimum size that allows the matching process.

  Next, in order to increase the contrast and make it easy to see the edge of the bone, LG is flattened (S106). Then, edge enhancement processing is performed for both DRR and LG (S108). The edge enhancement processing can be performed using, for example, a Sobel filter that uses the density multiplied by a coefficient to make the density of the central pixel. The Sobel includes a smoothing operation, and is different from general differentiation in that it is strong against noise. FIG. 3A shows DRR and FIG. 3B shows LG as an image after edge enhancement processing by a 7 × 7 sobel.

Here, the shift of the isocenter (center of image) between DRR and LG is corrected (S110). In DRR, MLC shape data and isocenter for irradiation field formation are recorded as numerical values (coordinate values on the image) in a part of the image data (DICOM header), and created so that the isocenter is the center of the image. The coordinate position is known. The irradiation field and isocenter reproduced from the DRCOM DICOM header are shown in FIG. For LG, the profile near the center of the LG image (see FIG. 4B) is analyzed to recognize the coordinate position of the LG isocenter (see FIG. 5).
For the recognition of the LG isocenter, a figure of the scale of the scale plate of the treatment apparatus (shown as a cross in FIG. 3B) that is copied in the LG image is used. The center of the cross is the isocenter of the LG image. First, only the central portion (shaded portion in FIG. 4B) is extracted from the LG image. This is the lower left image in FIG. Profiles created in the x direction (horizontal direction) and the y direction (vertical direction) are graphs on the upper and right sides of FIG. This profile reflects the characteristics of the scale image. Using this profile, the procedure for recognizing the isocenter, which is the center of the scale cross, will be described with reference to the y-direction profile graph (right side of FIG. 5A) (FIGS. 5B to 5C). d)). First, the maximum pixel value is searched (FIG. 5B). Pixel values at a position away from the position p1 at which the maximum pixel value is found by a certain range (3 pixels in this case) are compared, and a position p2 at which the second largest pixel value is obtained is obtained (FIG. 5C). An isocenter search range is set between p1 and p2, a minimum pixel value in the range is searched, and a position p3 at which the minimum pixel value is obtained is set as an isocenter.
Then, in order to coincide with the isocenter of the DRR, the LG image is moved by the difference in coordinates between the isocenter obtained by the LG and the isocenter of the DRR (see FIG. 6). Thus, the DRR and LG isocenter positions coincide.

The edge of the field of view (FOV) in the LG image is not in the DRR, and this edge becomes an obstruction factor for performing matching between the DRR and the LG image by the edge. For this purpose, the edge of the FOV in the LG image is removed (S112). This process will be described with reference to FIG.
First, an FOV shape image is created based on the MLC shape data taken together with the DRR image (see FIG. 7A). The created FOV image is contracted three times to create an FOV ′ image (see FIG. 7B). Further, the formed FOV image is expanded 10 times to create an FOV ″ image (see FIG. 7C). Taking the difference between FOV ″ and FOV ′, an image of a region of interest (ROI) is created (see FIG. 7D). An edge component in this ROI is set as an FOV edge, and a pixel value of this edge is set to 0 (= black: background color) to delete the edge. The image is shown in FIG. By using this ROI, as shown in FIG. 7E, only the FOV edge can be removed while leaving the image (human body structure edge) in the FOV.

Now, since matching is performed on a binarized image, histogram analysis processing is performed in order to automatically determine a threshold value for binarization (S114). As shown in the histogram graph of FIG. 8, counting is performed from the high pixel value side in each image, and a pixel value corresponding to 8% of the total number of pixels is obtained and set as a threshold value. In the example image, the DRR threshold is about 770 (arrow in FIG. 8A), and the LG threshold is about 600 (arrow in FIG. 8B). Using the obtained threshold value, binarization processing is performed on each image (S116). The binarized image shows DRR in FIG. 9A and LG in FIG. 9B.
As shown in FIG. 9B, when the LG image is binarized, a lot of noise scattered in the background appears. Therefore, the contraction process is repeated twice to remove the noise (S118). The LG image after noise removal is shown in FIG.

In order to select an edge to be matched, first, small ones are removed. In order to do this, a label for distinguishing is given to each connected edge (S120), and the area of each label is 50 pixels or less in the case of a DRR image, and in the case of an LG image. It is determined whether the pixel is 100 pixels or less, and the edge is deleted (S122). The image after the determination and deletion shows DRR in FIG. 11 (a) and LG in FIG. 11 (b).
The remaining connected edges are thinned (one pixel line) (S124). The image after the thin line processing shows DRR in FIG. 12A and LG in FIG.

Finally, a landmark to be matched is selected from the LG image (S126). For this purpose, the thinned DRR image (see FIG. 13A) is expanded 12 times to create a mask image (see FIG. 13B). This mask image is applied to the LG thinned image (see FIG. 13C), and only LG of the portion of the mask image is extracted, and the thinned edge in this portion is set as a matching target (FIG. 13). (See (d)).
FIG. 14 shows a screen of the present system that actually displays the landmarks determined by this processing. In FIG. 14, the LG image and the determined landmark are shown on the upper right side, the DRR image and the landmark are shown on the lower right side, and LG and DRR and the respective landmarks are overlapped in the center portion. Yes.

For the LG and DRR landmarks to be collated, a template (collation target range) is determined (S128), and collation by template matching is performed (S130). A method for determining and collating templates will be described with reference to FIGS.
As shown in FIG. 15A, when there are LG landmarks, the coordinates of the rightmost, leftmost, uppermost, and lowermost sides are found (see FIG. 15B), and these 4 are configured. A range surrounded by the coordinates of the points is defined as a template T (see FIG. 16A). As the search area S for this template, the same ROI (Region of Interest) size as the template and an appropriate margin (dx, dy) around the same are defined on the DRR (see FIG. 16B). Then, the most matching place is searched by moving the template as shown in the following formula (FIG. 16C).
Here, R is a residual, T (x, y): template, S (x, y): search area, N is the horizontal size of the template, M is the vertical size of the template, and 0 <x <M, 0 <Y <N, 0 <dx <50, 0 <dy <50. The time when R becomes the minimum value is the best match, and dx and dy at that time are values of deviation.
The shift amount (irradiation position shift) of the template in the x and y directions is displayed, for example, in mm (S132). After the shift amount is calculated, as shown in FIG. 17, a result image in which LG shifted by the detected shift amount is superimposed on the DRR is displayed. The shift amount is displayed as a numerical value in the horizontal window.
Thus, the difference (irradiation position deviation) between DRR and LG can be recognized, and it is understood how much movement should be indicated to the patient in the treatment apparatus.

  In this system, the operator selects an arbitrary landmark from the DRR and LG landmarks automatically determined by the above-described series of processes as shown in FIG. Can be automatically verified using the landmarks. The manual selection of the landmark may be performed after the landmark determination (S126) in the flowchart of FIG. This selection is performed by displaying DRR or LG with a landmark at the center in FIG. 14 and clicking the desired landmark with the mouse. The search range determination (S128) and collation (S130) are performed on the manually selected edges. The search range is a range of edges specified manually, and collation is performed only on the manually selected edges, so that accurate collation can be quickly performed without performing useless search.

  In the above description, the processing for DRR and the processing for LG have been described in parallel. However, in order to shorten the processing time, the DRR is acquired first, and the processing for the DRR (preprocessing, thinning processing, mask creation, ROI creation for FOV edge removal, etc.) is performed in advance. After photographing the LG, it is desirable to perform the processing on the LG using the processing result of the DRR.

It is a figure which shows the flowchart which shows the process of an irradiation position collation system. It is a figure which shows the image of DRR (a) and LG (b) after size adjustment. It is a figure which shows the image of DRR (a) and LG (b) after edge emphasis. It is a figure which shows the DRR irradiation field, isocenter (a), and LG profile analysis area (shaded part near the center) (b). It is a figure for demonstrating the recognition process of the center of the isocenter with respect to LG. It is an image displayed by superimposing DRR and LG, and is a diagram showing (a) before and after automatic correction of the isocenter shift (b). It is a figure for demonstrating the FOV edge removal process with respect to LG. It is a figure for demonstrating the histogram analysis with respect to DRR (a) and LG (b). It is a figure which shows the image of DRR (a) and LG (b) after a binarization process. It is a figure of LG image after noise removal. It is a figure which shows the image of DRR (a) and LG (b) after label determination. It is a figure which shows the image of DRR (a) and LG (b) after a thin line process. It is a figure for demonstrating landmark determination processing. It is a figure which shows the image of the landmark displayed on the screen of this system. It is a figure for demonstrating the determination process of a search range. It is a figure for demonstrating the process of collation. It is a figure which shows the screen which displays a collation result.

Claims (7)

An irradiation position verification system for aligning an irradiation site using a radiation treatment plan image (DRR) and an irradiation field confirmation image (LG) photographed before treatment,
Shape data for forming the DRR and irradiation field (FOV), image acquisition means for acquiring an image of the LG,
Preprocessing means for matching the size and position of the acquired DRR and LG images;
Edge enhancement means for enhancing edges of the DRR and LG images;
FOV edge removal means for removing the FOV edge in the LG using the acquired shape data;
Binarization means for binarizing DRR and LG images;
Thinning processing means for thinning edges of binarized DRR and LG images;
Collating means for detecting a deviation between the DRR and LG;
An irradiation position verification system comprising: display means for displaying a result of verification. In the irradiation position collation system according to claim 1,
Further, in the image after the thinning process of the DRR, a mask is created by expanding an edge, and landmark determination means for removing an unnecessary edge of the LG by the mask is provided,
The irradiation position collation system characterized in that the collation means collates with LG and DRR from which unnecessary edges are removed by the landmark determination means. In the irradiation position collation system according to claim 1 or 2, further,
Noise removing means for removing noise from the binarized LG image;
An irradiation position matching system comprising: means for removing small edges from the DRR image and the LG image after noise removal. In the irradiation position collation system in any one of Claims 1-3,
The irradiation means collation system characterized in that the collating means determines a range (template) in which LG collation is performed, and obtains a collation position within a search range defined in the DRR. In the irradiation position collation system in any one of Claims 1-4,
The irradiation position collating system further comprising manual selection means for manually selecting an edge to be used for collation, wherein the collation means collates the selected edge of LG with the selected edge of DRR. The computer program for functioning a computer system as an irradiation position collation system in any one of Claims 1-5. The recording medium which stored the computer program for functioning a computer system as an irradiation position collation system of Claim 6.