KR20230040860A - System And Method For Evaluating Multi-Leaf Collimator Performance - Google Patents

Abstract

A system for evaluating the performance of a multi-leaf collimator (MLC) according to an embodiment of the present invention comprises: a measurement unit which includes a flash plate for generating a scintillation signal by irradiating radiation emitted from a radiation source and passing through an MLC; a recording unit which records the scintillation signal generated from the measurement unit to generate a photographed image obtained by photographing the movement of the MLC, and generates a correction image based on the photographed image; and an analysis unit which analyzes the correction image received from the recording unit, and evaluates the position of the MLC.

Description

Multi-leaf collimator performance evaluation system and method {System And Method For Evaluating Multi-Leaf Collimator Performance}

The present invention relates to a multi-leaf collimator (MLC) performance evaluation system and method, and more particularly to a scintillator-based multi-leaf collimator performance evaluation system and method.

In order to improve the accuracy of the dose delivered to the patient during radiation therapy, high mechanical precision is required for the radiation therapy device. Among radiation treatments, intensity modulated radiation therapy (IMRT) consists of a radiation field generated by a combination of segments of a complex multi-leaf collimator (MLC). In radiation therapy, it is essential to precisely irradiate radiation to a tumor while protecting surrounding normal tissues, and a multi-lobed collimator (MLC) plays a central role. The multi-lobed collimator plays a role in protecting normal tissue by changing its size and shape to match the shape of the tumor whenever X-rays are delivered to the tumor. In the past, a lead alloy shielding block formed in the shape of a tumor was manufactured and used for each patient, but multi-leaf collimators are not separately manufactured for each treatment, so the treatment preparation time and cost are drastically reduced, and intensity-modulated radiation therapy is realized. Improve the precision of treatment.

Meanwhile, in Korea, weekly, monthly, and annual inspections are performed based on the TG-142 report published by the American Medical Physics Society for quality control of therapeutic devices. Positional accuracy evaluation of multi-layered collimators is recommended as a weekly inspection through the TG-142 report, and generally, the positional accuracy of multi-layered collimators is evaluated through analysis of the sensitized film after radiation is irradiated to the film. Positional analysis through the film is performed after scanning the film through a scanner, but has a disadvantage in that an accurate analysis is difficult due to errors caused by the resolution of the scanner. In addition, a time delay of several tens of minutes to several hours occurs depending on the type of film used to determine the accuracy of the position of the multi-layered collimator by scanning and analyzing the film after measurement, and since the film cannot be reused, a new film must be used for each evaluation. do. Also, it is not possible to evaluate the accuracy of the dynamic movement state of the multi-leaf collimator used in clinical practice.

In order to solve the above problems, embodiments of the present invention are intended to provide a reusable MLC performance evaluation system using a scintillator and a CMOS camera.

In addition, by applying a correction algorithm using a scintillation operator and a compensation map to a photographed image related to the motion of a video-based multi-lobe collimator (MLC), an MLC performance evaluation system capable of measuring the position of the MLC and simultaneously evaluating its performance even in the dynamic state of the MLC and to provide a method thereof.

A multi-leaf collimator performance evaluation system according to an embodiment of the present invention includes a measurement unit including a flash plate for generating a flash signal by irradiating radiation emitted from a radiation source and passing through a multi-leaf collimator (MLC); a recording unit for recording the scintillation signal generated from the measurement unit to generate a photographed image obtained by photographing the movement of the multi-leaf collimator, and to generate a correction image obtained by correcting geometric distortion based on the photographed image; and an analyzer configured to evaluate the position of the multi-lobe collimator by analyzing the corrected image received from the recorder.

The recording unit may record the flash signal in the form of a video using a CMOS camera module to obtain the photographed image in the form of a video.

The recording unit may include: an image conversion unit that converts the photographed image into a cumulative image including a plurality of images divided based on the changed shape of the multi-lobe collimator; and an image correcting unit configured to generate the corrected image by correcting geometrical distortion caused by the location of the recording unit with respect to the cumulative image.

The image correction unit may calculate a scintillation operator based on a luminance formula for the scintillation signal, and generate a compensation map compensating for attenuation of the scintillation signal caused by a distance between the recording unit and the flash plate.

The image correction unit may acquire the corrected image by applying an iterative reconstruction algorithm using the scintillation operator and the compensation map to the cumulative image.

The analysis unit generates a location map for learning based on a captured image for learning and location information for learning, which is multi-lobe collimator location information corresponding to the captured image for learning, and generates a location map for learning based on the captured image for learning, the location information for learning, and the location map for learning. a machine learning unit including a learning model to be learned; and a position prediction unit that outputs a position map displaying a boundary line of the shape of the multi-leaf collimator as a discrete signal using the corrected image as input data using the learning model.

The position prediction unit may output a binary image by correcting a position of the multi-lobe collimator with respect to the position map.

A multi-leaf collimator performance evaluation method according to an embodiment of the present invention includes generating a flash signal by irradiating radiation emitted from a radiation source and passing through a multi-leaf collimator (MLC) using a measuring unit including a flash plate. ; generating a photographed image obtained by photographing motions of the multi-leaf collimator by recording the scintillation signal generated from the measurement unit using a recording unit, and generating a correction image obtained by correcting geometric distortion based on the photographed image; and evaluating the position of the multi-leaf collimator by analyzing the corrected image using an analysis unit.

The photographed image may be obtained by recording the flash signal in the form of a video using a CMOS camera module.

The generating of the corrected image may include converting the photographed image into an accumulated image including a plurality of images divided based on the changed shape of the multi-lobe collimator; and generating the corrected image by calibrating geometrical distortion caused by the location of the recording unit with respect to the cumulative image.

The generating of the corrected image may include calculating a scintillation operator based on a luminance formula for the scintillation signal; and generating a compensation map that compensates for the attenuation of the scintillation signal caused by the distance between the recording unit and the flash plate.

The generating of the corrected image may include obtaining the corrected image by applying an iterative reconstruction algorithm using the scintillation operator and the compensation map to the cumulative image.

The step of evaluating the location of the multi-lobe collimator may include generating a location map for learning based on a captured image for learning and location information for learning, which is location information of the multi-lobe collimator corresponding to the captured image for learning, and generating the captured image for learning, the location information for learning, and learning a learning model based on the learning location map; and outputting a position map displaying a boundary line of the shape of the multi-leaf collimator as a discrete signal using the corrected image as input data using the learning model.

Evaluating the position of the multi-lobe collimator may include outputting a binary image by correcting the position of the multi-lobe collimator with respect to the position map.

According to embodiments of the present invention, it is possible to provide a reusable MLC performance evaluation system using a scintillator and a CMOS camera.

In addition, by applying a compensation algorithm using a scintillation operator and a compensation map to a photographed image of the motion of a video-based multi-lobe collimator (MLC), it is possible to measure the position of the MLC and simultaneously evaluate its performance even in a dynamic state of the MLC.

1 is a configuration diagram schematically showing the configuration of a multi-leaf collimator performance evaluation system according to an embodiment of the present invention.
2 is a design diagram showing a practical design form of a multi-leaf collimator performance evaluation system according to an embodiment of the present invention.
3 is a flowchart illustrating a multi-leaf collimator performance evaluation method according to an embodiment of the present invention.
4 is a configuration diagram schematically showing the configuration of a recording unit according to an embodiment of the present invention.
5 is a schematic diagram for explaining an iterative reconstruction algorithm among image correction units according to an embodiment of the present invention.
6 is a schematic diagram for explaining a scintillation operator of an image correction unit according to an embodiment of the present invention.
7 is a schematic diagram for explaining the creation of a compensation map by an image correction unit according to an embodiment of the present invention.
8 is a schematic diagram for explaining the operation of the analyzer according to an embodiment of the present invention. FIG. 9 is a schematic diagram for explaining the effect of a multi-leaf collimator performance evaluation method according to an embodiment of the present invention.
10A is an exemplary view of an image photographed using a film for radiometry. 10B is an exemplary diagram of a corrected image generated through a performance evaluation system according to an embodiment of the present invention.
11 is an exemplary view of a graph comparing dose distribution between an image captured using a film for radiometry and a corrected image.
12A is an exemplary view of a position map of the position of a multi-leaf collimator created by one embodiment of the present invention. 12B is an exemplary diagram of a binary image for position prediction of a multi-lobe collimator.
FIG. 13A is an exemplary view showing positional accuracy of a multi-leaf collimator (bank A) predicted from a film for radiometry. FIG. 13B is an exemplary view showing positional accuracy of a multi-leaf collimator (Bank B) predicted in a film for radiometry. 13C is an exemplary view showing the position accuracy of the multi-leaf collimator (Bank A) predicted from the position map of the analyzer. 13D is an exemplary view showing the position accuracy of the multi-leaf collimator (Bank B) predicted from the position map of the analyzer. 13E is an exemplary diagram illustrating the positional accuracy of a multi-leaf collimator (Bank A) predicted from a binary image of an analyzer. 13F is an exemplary diagram illustrating the positional accuracy of a multi-leaf collimator (Bank B) predicted from a binary image of an analyzer.
FIG. 14A is an exemplary diagram of a table comparing prediction results of multi-leaf collimators in the film image, position map, and binary image of FIGS. 13A, 13C, and 13E. FIG. 14B is an exemplary view of a table comparing prediction results of multi-leaf collimators in the film image, position map, and binary image of FIGS. 13B, 13D, and 13F.

Since the present invention can apply various transformations and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail in the detailed description. Effects and features of the present invention, and methods for achieving them will become clear with reference to the embodiments described later in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various forms.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, and when describing with reference to the drawings, the same or corresponding components are assigned the same reference numerals, and overlapping descriptions thereof will be omitted. .

In the following embodiments, terms such as first and second are used for the purpose of distinguishing one component from another component without limiting meaning. In the following examples, expressions in the singular number include plural expressions unless the context clearly dictates otherwise. In the following embodiments, terms such as include or have mean that features or components described in the specification exist, and do not preclude the possibility that one or more other features or components may be added. In the drawings, the size of components may be exaggerated or reduced for convenience of description. For example, since the size and shape of each component shown in the drawings are arbitrarily shown for convenience of description, the present invention is not necessarily limited to those shown.

Hereinafter, a multi-leaf collimator (MLC, hereinafter, simply referred to as 'MLC') performance evaluation system according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2 together. 1 is a configuration diagram schematically showing the configuration of a multi-leaf collimator performance evaluation system according to an embodiment of the present invention, and FIG. 2 shows a practical design form of a multi-leaf collimator performance evaluation system according to an embodiment of the present invention. it is a blueprint

The MLC performance evaluation system 10 may include a measurement unit 100 , a recording unit 200 and an analysis unit 300 .

The measuring unit 100 is a component for measuring the movement of the MLC that determines the area to which radiation is irradiated, and may include a flash plate 110 for generating a flash signal by irradiating radiation emitted from a radiation source and passing through the MLC. there is. Since the flash plate 110 generates a scintillation signal by stimulating radiation unlike a photosensitive film that is exposed by radiation, it has the advantage of being reusable in a multi-layered collimator performance evaluation system. In addition, when the MLC position must be analyzed and corrected using a conventional photosensitive film, the analysis must be performed after a certain period of time after the film is exposed, so there is a disadvantage in that the calibration is delayed for several hours or days, but in the present invention, the flash generated Since the signal is immediately obtained as an image using the recording unit 200 and the analysis unit 300 is used, it is possible to automatically analyze the signal almost simultaneously with the measurement.

The flash plate 110 may be disposed above the darkroom 22, but the arrangement of the flash plate 110 is designed in various ways within a range where the movement or position of the MLC can be easily measured through the recording of a flash signal. can be changed

The recording unit 200 may capture an image related to the movement of the MLC using the scintillation signal generated from the measurement unit 100 . The recording unit 200 may record the scintillation signal generated from the measuring unit 100 to generate a captured image of the movement of the MLC, and may generate a correction image obtained by correcting geometric distortion based on the captured image. .

The recording unit 200 may include a CMOS camera module 21 and a darkroom 22 as device configurations. Two CMOS camera modules 21 may be provided so as to be symmetrically disposed on both sides of the flash plate 110 disposed above the darkroom 22 . However, the number and arrangement of the CMOS camera modules 21 are not limited thereto, and in addition to the CMOS camera module 21, other photographing devices capable of easily recording a flash signal may be used. The CMOS camera module 21 may be installed in the darkroom 22 . The darkroom 22 may support the CMOS camera module 21 .

The recording unit 200 may record the flash signal generated from the flash plate 110 in the form of a video using the CMOS camera module 21 . The recording unit 200 may obtain an image of a flashing light signal in the form of a video file or the like. The recording unit 200 may generate a corrected image by correcting an image obtained by photographing a scintillation signal. As shown in FIGS. 10A and 10B , the recording unit 200 may acquire a captured image and correct the captured image (S10, see FIG. 10A) to generate a corrected image (S11, see FIG. 10B). A processor configuration of the recording unit 200 will be described in detail in related drawings to be described later.

The analyzer 300 may measure the position of the MLC by analyzing the correction image received from the recorder 200 . The analyzer 300 may evaluate the performance of the measured position by comparing the measured position of the MLC with the actual position. Alternatively, the analyzer 300 may evaluate the performance of the measured position by comparing the measured position of the MLC with the position measured by the film for radiometry. As shown in FIG. 11, the analysis unit 300 compares the MLC position and dose distribution measured with the EBT3 film, which is a film for radiometry, with the measured MLC position and the intensity of the correction image, so that the measured MLC performance can be evaluated. The operation of the analyzer 300 will be described in detail in related drawings to be described later.

The recording unit 200 and the analysis unit 300 are not shown in this drawing, but may include a communication unit, a processor, and a memory, respectively.

The communication unit may communicate with various types of external devices and servers according to various types of communication methods. Each communication unit of the recording unit 200 and the analysis unit 300 may be connected by a network (not shown) to exchange data with each other.

The processor may perform an operation of overall controlling each component 200 and 300 having each memory using various programs stored in the memory. Processors include microprocessors, central processing units (CPUs), processor cores, multiprocessors, application-specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs). It may include a processing device, but the present invention is not limited thereto.

The memory may temporarily or permanently store all kinds of data processed by each component 200 or 300 having each memory. The memory may include a random access memory (RAM), a read only memory (ROM), and a permanent mass storage device such as a disk drive, but the present invention is not limited thereto. Operations performed by each of the above-described recording unit 200 and analysis unit 300 may be performed by each processor communicating with a communication unit of another configuration.

In this figure, although the recording unit 200 and the analysis unit 300 are shown as being provided as separate modules, they are only functionally classified according to the operation performed by the multi-lobe collimator performance evaluation system 10 and must be independently distinguished from each other. It does not, and can be implemented by integrating into one module as needed. Depending on the embodiment, of course, the multi-lobe collimator performance evaluation system 10 may include other configurations not shown in the drawings.

3 is a flowchart illustrating a multi-leaf collimator performance evaluation method according to an embodiment of the present invention. A multi-leaf collimator performance evaluation method according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2 described above.

A scintillation signal may be generated by irradiating radiation emitted from a radiation source and passing through the MLC using the measurement unit 100 including the flash plate 110 (S100).

Thereafter, the recording unit 200 records the scintillation signal generated from the measuring unit 100 to generate a photographed image of the motion of the multi-leaf collimator, and a correction image obtained by correcting geometric distortion based on the photographed image. It can be created (S200). At this time, the captured image may be obtained by recording the flash signal in the form of a video using the CMOS camera module 21 . That is, since the captured image is obtained in the form of a moving picture, the performance of the MLC in a dynamic state can also be analyzed in real time, in contrast to the existing MLC performance evaluation system, in which only the performance of the MLC in a static state can be evaluated. The step of generating a corrected image (S200) will be described in detail with reference to FIGS. 4 to 7 to be described later.

Thereafter, the position of the multi-leaf collimator may be evaluated by analyzing the corrected image using the analysis unit 300 (S300). The step of evaluating the location of the MLC based on the corrected image (S300) will be described in detail with reference to FIGS. 8 and 9 to be described later.

4 is a configuration diagram schematically showing the configuration of a recording unit 200 according to an embodiment of the present invention.

The recording unit 200 may include an image conversion unit 210 and an image correction unit 220 .

The image conversion unit 210 may convert the captured image by the recording unit 100 into an accumulated image including a plurality of images divided based on the changed shape of the multi-lobe collimator. The accumulated image converted in this way may arrive with a scintillation signal generated from the measuring unit 100 being distorted due to the position of the camera module 21 of the recording unit 200 .

Accordingly, the image correction unit 220 may generate a corrected image by correcting the geometric distortion caused by the location of the recording unit 100 with respect to the cumulative image generated by the image conversion unit 210 . The function of the image correction unit 220 will be described in more detail using FIGS. 5 to 7 to be described later.

5 is a schematic diagram for explaining a correction image generation step (S200) corresponding to an iterative reconstruction algorithm among image correction units according to an embodiment of the present invention.

The operation of FIG. 5 may be performed by the recording unit 200 .

Referring to FIG. 5 , the recording unit 200 may generate a photographed image 51 related to the movement of the MLC using the scintillation signal generated from the measuring unit 100 (S210). The captured image 51 may be denoted as 'b' in an iterative reconstruction algorithm to be described later. Thereafter, the image conversion unit 210 (refer to FIG. 4 ) of the recording unit 200 may convert the captured image 51 into a cumulative image.

Parallel to this, the recording unit 200, particularly the image correction unit 220, may calculate the scintillation operator ALumi of the present invention (S220). In the past, the radiation transmitted through the MLC was directly sensitized to the film to acquire images and conduct analysis. However, in the present invention, radiation measurement was performed using the measuring unit 100 and the recording unit 200 instead of the photosensitive film, and used in reconstruction of computed tomography to correct geometric distortion caused by the position of the recording unit 200. An iterative reconstruction algorithm (S240) was used. Reconstruction of conventional computed tomography reconstructs an image using an image recorded by radiation emitted from a source penetrating a subject and a corresponding projection operator. However, since the present invention uses the scintillation image generated from the scintillation plate 110 of the measurement unit 100, the scintillation operator ALumi is obtained by replacing the projection operator calculated based on the transmission length through which the radiation photon passes through the subject. It is intended to correct the geometrical distortion caused by the position of the recording unit 200 by using it. Here, a description will be made with reference to FIG. 6 .

6 is a schematic diagram for explaining a step (S220) of calculating a scintillation operator of an image correction unit according to an embodiment of the present invention.

The image correction unit 220 may calculate a scintillation operator ALumi based on a luminance formula for a scintillation signal.

Referring to FIG. 6 , the luminance (B) of the flash signal is determined by the distance (r) between the recording unit 200 (here, it may mean the CMOS camera module 21) and the subject 120 and the recording unit 200. It can be defined by the brightness of a flash signal generated from an arbitrary area Lx of the object 120 to be sensed, and the luminance calculation formula can be expressed as [Equation 1] below. Hereinafter, the subject 120 may mean the flash plate 110 described above.

[Equation 1]

The scintillation operator (A _Lumi ) uses the luminance calculation formula of [Equation 1] as described above for x arbitrary areas (L) of the subject 120 corresponding to each of the N CMOS camera modules 21 provided in the recording unit 200. ₁ to L _x ) can be defined as a determinant calculated for each. The scintillation operator (A _Lumi ) defined in this way can be expressed as in [Equation 2] below.

[Equation 2]

At this time, rN,x of the denominator of [Equation 2] represents the distance from the N-th CMOS camera module 21 to an arbitrary area Lx of the subject 120, and LN,x of the numerator represents the N-th CMOS camera module 21. It represents the brightness of a flash signal generated from an arbitrary area Lx of the subject 120 sensed by the module 21 .

Thereafter, the image compensator 220 may map the calculated scintillation operator ALumi with each region corresponding to an arbitrary region Lx of the subject 120 in the photographed image b (x0 = ALumiT × b). In other words, the initial predicted value x0 may be calculated by multiplying each component of the scintillation operator ALumi matrix with each region corresponding to an arbitrary region Lx of the subject 120 of the photographed image b.

Meanwhile, the image compensator 220 may generate a compensation map 52 for compensating for attenuation of a scintillation signal generated by the distance between the recording unit 200 and the subject 120 based on the captured image 51. (S230). The compensation map 52 may be denoted as 'bcomp' in an iterative reconstruction algorithm to be described later. Hereinafter, 'bcomp' means a component corresponding to each of the arbitrary regions Lx of the subject 120 of the compensation map 52. can do. Here, a description will be made with reference to FIG. 7 .

7 is a schematic diagram for explaining a step (S230) of generating a compensation map of an image correction unit according to an embodiment of the present invention.

The image correction unit 220 may generate a compensation map 52 for compensating for attenuation of a scintillation signal generated by a distance between the recording unit 200 and the subject 120 . At this time, an operation of generating a compensation map 52 may be performed on an accumulated image generated by converting the above-described captured image.

Referring to FIG. 7 , the compensation map is based on the distance r from the recording unit 200 (which may mean the CMOS camera module 21) to an arbitrary area Lx of the subject 120. (52) can be produced. Each component (bcomp) of the compensation map 52 may be expressed as in [Equation 3] below, which corrects the distortion according to the distance between the camera module 21 and the subject 120, thereby correcting the final image to be analyzed. (correction image) is compensated so that it has uniform intensity or brightness as a whole.

[Equation 3]

Specifically, referring to the exemplary compensation map 521 shown in FIG. 7 , different shades are displayed according to distances along the second direction D2 . For example, since the compensation degree should be large for an area where the aforementioned distance r is long, a high compensation weight may be applied, and this may be represented by a relatively bright shade on the compensation map 52 . Conversely, since the compensation degree is small for an area where the above-mentioned distance r is short, a low compensation weight may be applied, and this may be represented by a relatively dark shade on the compensation map 52 . Referring to the example compensation map 521 , it can be seen that a higher compensation weight is assigned from top to bottom in the second direction D2 , and thus brightness is brighter.

Referring to FIG. 5 again, an iterative reconstruction algorithm (S240) is performed using the scintillation operator (A _Lumi ) generated in steps S220 and S230, the initial predicted value (x ₀ ) calculated based on the scintillation operator, and the compensation map (52). can do. Referring to [Equation 4] (Eq4) shown in the iterative reconstruction algorithm (S240), it can be seen that a scintillation operator (A _Lumi ) and a compensation map (b _comp ) are used to obtain the final corrected image 53. . That is, the use of the flash plate 110 not only overcomes the problem of a photosensitive film that can only be used once, but also compensates for the distance or interference between the flash signal generated from the flash plate 110 and the recording unit 200. There is an advantage in that a corrected image 53, which is an image to be analyzed accurately and uniformly, can be obtained.

8 is a schematic diagram for explaining the operation of the analyzer 300 according to an embodiment of the present invention.

The analysis unit 300 generates a corrected image 321 (the above-described corrected image ( 53)) may include a location prediction unit 320 outputting a location map 322 and/or a binary image 323 regarding the location of the MLC. A location map of the location of the MLC may be as shown in FIG. 12A. A binary image of the location of the MLC may be as shown in FIG. 12B.

The learning model (ML) generates a location map 313 for learning based on the captured image 311 for learning and the location information 312 for learning, which is multi-lobe collimator location information corresponding to the captured image 311 for learning, and the captured image 313 for learning. Based on the image 311, the location information 312 for learning, and the location map 313 for learning, learning may be performed through a Cycle Consistent Generative Adversarial Network (cycleGAN) (S310). In detail, the cycleGAN model uses two generators to convert data in one domain to another domain and back to the original domain through a recursive process that minimizes the loss function of the model and minimizes the gap between data in different domains. It is a method for predicting non-linear relationships.

The position predictor 320 uses the learning model ML to take the above-described corrected image 321 as input data and displays the boundary of the shape of the multi-leaf collimator with a discrete signal Position map 322 can be output (S320). Thereafter, the position predictor 320 may output a binary image 323 by correcting the position of the multi-lobe collimator with respect to the position map 322 using a binary image generator. The binary image generator may be learned through Unet based on the location map 322 for learning and the binary image 323 for learning (S320). In detail, the Unet model is a method of analyzing regional characteristics of input data through encoding and decoding of the input data and proceeding with segmentation learning derived as a binary image.

Compared to the location map 322, the binary image 323 refers to an image having a clear boundary of a region where a flashing light signal is recorded after passing through the MLC. The correction image 321, the location map 322, and the binary image 323 used in the analysis unit 300 will be described in detail in FIG. 9 described below.

9 is a schematic diagram for explaining the effects of a multi-leaf collimator performance evaluation system and method according to an embodiment of the present invention.

Referring to FIG. 9, an example of a correction image 321, a location map 322, and a binary image 323 used in the analysis unit 300 and the intensity (Intensity) of a scintillation signal according to the distance (position) of these images A graph 90 is shown. Referring to the graph 90, in the images 321, 322, and 323 of FIG. 9, the light-shaded portion where the flashing signal is recorded is the portion to which the radiation that passed through the MLC was irradiated, roughly in the first direction D1. It can be seen that it appears in the form of a square in the range of -70mm to 70mm. The intensity of the scintillation signal in the corresponding region (-70 mm to 70 mm) is close to about 1 (unit description), and in the case of the corrected image 321, the intensity was measured in the corresponding region with instability. On the other hand, in the case of the location map 322, the intensity was measured more uniformly than the correction image 321, but the intensity was measured relatively gently near the boundary between the part where the flash signal was recorded and the part where the flash signal was not recorded, indicating that the boundary is not clear. You can check. On the other hand, in the case of the binary image 323, the gradient of the intensity near the boundary is measured steeper than that of the location map 322, so it can be seen that the boundary line of the bright shaded part is more distinct. In other words, by measuring the shape (position) of the MLC with higher resolution, the performance of the MLC can be more accurately evaluated.

10A is an exemplary view of a captured image generated through a performance evaluation system according to an embodiment of the present invention. 10B is an exemplary diagram of a corrected image generated through a performance evaluation system according to an embodiment of the present invention.

11 is an exemplary view of a graph comparing dose distribution between an image captured using a film for radiometry and a corrected image.

12A is an exemplary view of a position map of the position of a multi-leaf collimator created by one embodiment of the present invention. 12B is an exemplary diagram of a binary image for position prediction of a multi-lobe collimator.

13A is an exemplary view showing positional accuracy of a multi-leaf collimator (bank A) measured on a film for radiometry. 13a shows position values (Trial 1 to Trial 4) of the multi-leaf collimator measured from the film for radiometry over four times. At this time, by changing the location, the position values of the multi-leaf collimator from Field 1 to Field 9 can be predicted, and the difference between them and the actual multi-leaf collimator can be graphed.

FIG. 13B is an exemplary view showing positional accuracy of a multi-leaf collimator (Bank B) predicted in a film for radiometry.

13C is an exemplary view showing the position accuracy of the multi-leaf collimator (Bank A) predicted from the position map of the analyzer. 13C shows position values (Trial 1 to Trial 4) of the multi-lobe collimator predicted from position maps predicted four times. At this time, by changing the location, the position values of the multi-leaf collimator from Field 1 to Field 9 can be predicted, and the difference between them and the actual multi-leaf collimator can be graphed.

13D is an exemplary view showing the position accuracy of the multi-leaf collimator (Bank B) predicted from the position map of the analyzer.

13E is an exemplary diagram illustrating the positional accuracy of a multi-leaf collimator (Bank A) predicted from a binary image of an analyzer. 13E shows position values (Trial 1 to Trial 4) of the multi-lobe collimator measured from binary images predicted four times. At this time, by changing the location, the position values of the multi-leaf collimator from Field 1 to Field 9 can be predicted, and the difference between them and the actual multi-leaf collimator can be graphed.

13F is an exemplary diagram illustrating the positional accuracy of a multi-leaf collimator (Bank B) predicted from a binary image of an analyzer.

Comparing the graphs of FIGS. 13A, 13C, and 13E, it can be seen that the difference between the position value of the multi-lobe collimator predicted from the binary image generated by the embodiments of the present invention and the actual value is the smallest. Similarly, comparing the graphs of FIGS. 13b, 13d, and 13f, it can be seen that the difference value between the position value of the multi-lobe collimator predicted from the binary image generated by the embodiments of the present invention and the actual value is the smallest. can It can be seen that the value predicted from the binary image according to the present invention is more accurate than the value predicted by the conventional radiation measurement film.

FIG. 14A is an exemplary diagram of a table comparing prediction results of multi-leaf collimators in the film image, position map, and binary image of FIGS. 13A, 13C, and 13E. FIG. 14B is an exemplary view of a table comparing prediction results of multi-leaf collimators in the film image, position map, and binary image of FIGS. 13B, 13D, and 13F. In FIG. 14A, the position values of the multi-leaf collimator (bank A) in the film image, the position map, and the binary image are measured from Field 1 to Field 9, and the resulting mean absolute error , MAE) can be compared. According to the table, the average absolute error value in the binary image is calculated as smaller (0.673) than the average absolute error value of the radiographic film (0.864) and the average absolute error value of the location map (0.997), which is multi-lobed in the binary image. This may mean that the position value of the collimator is the most accurate.

In FIG. 14B, the position values of the multi-leaf collimator (bank B) in the film image, position map, and binary image are measured from Field 1 to Field 9, and the mean absolute error , MAE) can be compared. According to the table, the average absolute error value in the binary image is calculated as smaller (0.566) than the average absolute error value (1.277) of the radiographic film and the average absolute error value (0.830) of the location map, which is multi-lobed in the binary image. This may mean that the position value of the collimator is the most accurate.

Embodiments according to the present invention described above may be implemented in the form of a computer program that can be executed on a computer through various components, and such a computer program may be recorded on a computer-readable medium. In this case, the medium may store a program executable by a computer. Examples of the medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROM and DVD, magneto-optical media such as floptical disks, and ROM, RAM, flash memory, etc. configured to store program instructions.

Meanwhile, the computer program may be specially designed and configured for the present invention, or may be known and usable to those skilled in the art of computer software. An example of a computer program may include not only machine language code generated by a compiler but also high-level language code that can be executed by a computer using an interpreter or the like.

In addition, although the preferred embodiments of the present invention have been shown and described above, the present invention is not limited to the specific embodiments described above, and the technical field to which the present invention belongs without departing from the gist of the present invention claimed in the claims. Of course, various modifications are possible by those skilled in the art, and these modifications should not be individually understood from the technical spirit or perspective of the present invention.

Therefore, the spirit of the present invention should not be limited to the above-described embodiments, and not only the claims to be described later, but also all scopes equivalent to or equivalently changed from these claims fall within the scope of the spirit of the present invention. would be considered to be in the category.

10: multi-leaf collimator performance evaluation system
100: measuring unit
110: flash plate
200: register
21: CMOS camera module
210: image conversion unit
220: image correction unit
52: reward map
53, 321: correction image
300: analysis unit
322: location map
323 Binary image

Claims (14)

a measuring unit including a flash plate generating a flash signal by irradiating radiation emitted from a radiation source and passing through a multi-leaf collimator (MLC);
a recording unit for recording the scintillation signal generated from the measurement unit to generate a photographed image obtained by photographing the movement of the multi-leaf collimator, and to generate a correction image obtained by correcting geometric distortion based on the photographed image; and
An analysis unit configured to evaluate the position of the multi-leaf collimator by analyzing the corrected image received from the recording unit; and a multi-leaf collimator performance evaluation system. According to claim 1,
Wherein the recording unit records the flash signal in the form of a video using a CMOS camera module to obtain the photographed image in the form of a video. According to claim 2,
the record,
an image conversion unit for converting the photographed image into a cumulative image including a plurality of divided images based on the changed shape of the multi-lobe collimator; and
and an image correction unit configured to generate the corrected image by calibrating geometric distortion caused by the position of the recording unit with respect to the cumulative image. According to claim 3,
The image correction unit,
Calculating a scintillation operator based on a luminance formula for the scintillation signal;
and generating a compensation map compensating for attenuation of the scintillation signal caused by a distance between the recording unit and the flash plate. According to claim 4,
The image correction unit,
The multi-lobe collimator performance evaluation system for obtaining the corrected image by applying an iterative reconstruction algorithm using the scintillation operator and the compensation map to the cumulative image. According to claim 3,
The analysis unit,
A learning model that generates a location map for learning based on a captured image for learning and location information for learning, which is multi-lobe collimator location information corresponding to the captured image for learning, and is learned based on the captured image for learning, the location information for learning, and the location map for learning. Machine learning unit including; and
A multi-lobed collimator performance evaluation system comprising a; position prediction unit that outputs a position map displaying a boundary line of the shape of the multi-lobed collimator as a discrete signal using the corrected image as input data using the learning model.
According to claim 6,
The position prediction unit,
The multi-lobe collimator performance evaluation system for outputting a binary image by correcting the position of the multi-lobe collimator with respect to the location map. generating a scintillation signal by irradiating radiation emitted from a radiation source and passing through a multi-leaf collimator (MLC) using a measuring unit including a flashing plate;
generating a photographed image obtained by photographing motions of the multi-leaf collimator by recording the scintillation signal generated from the measurement unit using a recording unit, and generating a correction image obtained by correcting geometric distortion based on the photographed image; and
Evaluating the position of the multi-leaf collimator by analyzing the corrected image using an analysis unit; including, a multi-leaf collimator performance evaluation method. According to claim 8,
The photographed image is obtained by recording the scintillation signal in the form of a video using a CMOS camera module. According to claim 9,
The step of generating the correction image,
converting the photographed image into a cumulative image including a plurality of divided images based on the changed shape of the multi-lobe collimator; and
and generating the corrected image by calibrating geometrical distortion caused by the location of the recording unit with respect to the cumulative image. According to claim 10,
The step of generating the correction image,
Calculating a scintillation operator based on a luminance formula for the scintillation signal; and
and generating a compensation map compensating for the attenuation of the scintillation signal caused by the distance between the recording unit and the flash plate. According to claim 11,
The step of generating the correction image,
And obtaining the corrected image by applying an iterative reconstruction algorithm using the scintillation operator and the compensation map to the cumulative image. According to claim 10,
Evaluating the position of the multi-leaf collimator,
A location map for learning is generated based on a captured image for learning and location information for learning, which is multi-lobe collimator location information corresponding to the captured image for learning, and a learning model is learned based on the captured image for learning, the location information for learning, and the location map for learning. step of doing; and
and outputting a position map displaying a boundary line of the shape of the multi-leaf collimator as a discrete signal using the corrected image as input data using the learning model. According to claim 13,
Evaluating the position of the multi-leaf collimator,
and outputting a binary image by correcting the position of the multi-lobe collimator with respect to the location map.

Images