A Novel ScanNet method for Real-Time Position ... - ACTA Press

Proceedings of the IASTED International Conference Advances in Computer Science (ACS 2013) April 1 0 - 1 2 , 2013 Phuket , Thailand

A NOVEL SCANNET METHOD FOR REAL-TIME POSITION MONITORING SYSTEM IN RADIOTHERAPY Yu-Feng Chien1, Chi-Heng Lu2, Sheng-Wen Lin3, Tsair-Fwu Lee4, Jiun-Jie Wang5, Mei-Chuan Chen6, Chiu-Ching Tuan7,* 1 Department of Electronics Engineering, National Taipei University of Technology, Taipei Taiwan, R.O.C 4F., No.20, Zhongxing Rd., Xinyi Dist., Keelung City 201, Taiwan R.O.C. 2 Department of Radiation Oncology, Chang Gung Memorial Hospital at Linkou, Taoyuan, Taiwan, R.O.C. 5 Fu-Hsing St., Kwei-Shan, Tao-Yuan 333, Taiwan, R.O.C. 3 Department of Electronics Engineering, National Taipei University of Technology, Taipei Taiwan, R.O.C 1, Sec. 3, Zhong-Xiao E. Rd., Taipei 10608 Taiwan, R.O.C. 4 Department of Radiation Oncology, Chang Gung Memorial Hospital at Linkou, Taoyuan, Taiwan, R.O.C. 415 Chien Kung Road, Sanmin District, Kaohsiung 80778, Taiwan, R.O.C. 5 Department of Radiation Oncology, Chang Gung Memorial Hospital at Linkou, Taoyuan, Taiwan, R.O.C. 5 Fu-Hsing St., Kwei-Shan, Tao-Yuan 333, Taiwan, R.O.C. 6 Department of Radiation Oncology, Chang Gung Memorial Hospital at Linkou, Taoyuan, Taiwan, R.O.C 5 Fu-Hsing St., Kwei-Shan, Tao-Yuan 333, Taiwan, R.O.C. 7 Department of Electronics Engineering, National Taipei University of Technology, Taipei Taiwan, R.O.C 1, Sec. 3, Zhong-Xiao E. Rd., Taipei 10608 Taiwan, R.O.C. 7, *[email protected] photon beams, proton could be more focused on the tumor location. However, this characteristic could make proton liable to the displacement of patient position or organ motion. In Fig. 1 Bragg peak in the graph, the red area on behalf of the tumor, the location of about 12 cm to 26 cm in the depth of the human body, The Photon-ray into the table about its energy peak in the cortex to a depth of 10 cm, the role of the tumor area less energy, not only the therapeutic effect is limited, located in table cortex to a depth of 10 cm of healthy tissue are subjected to higher energy radiation and produce severe side effects. Protonray energy peak concentrated in tumor depth, nearly three-quarters of the epidermis to reduce harm to healthy tissue between the tumor area, and deeper than the tumor area or even close to zero.

ABSTRACT Radiation therapy is a major modality of cancer treatment. To achieve good local control, high radiation dose must be precisely focused to the target region. A geometric missing during radiotherapy could result in tumor regrowth or severe damage of the neighboring normal tissues. To prevent geometric missing, a real-time monitoring the patient movement becomes very important. Using image or infrared technology to detect the displacement has been proposed. However, Promoting the current detection efficiency is too poor to fulfill clinical requirement due to the low resolution of 512 × 512 pixels and slow complex calculation. To overcome these problems, this study proposed ScanNet algorithm to monitor the patient breathing, cough, or other movement. ScanNet algorithm could detect 1 mm shift under the resolution of 1920 × 1080 pixels in realtime and to meet the quality requirement of modern radiotherapy for monitoring displacement. KEY WORDS Radiotherapy, Movement Detection, Contour, Patient Monitoring.

1. Introduction Radiation therapy is one of the important options for cancer treatment currently. New modalities of radiotherapy, such as proton or heavy ion therapy have Bragg's peak to prevent the damage to the distal normal organ, as shown in Fig. 1 Compared to the conventional

DOI: 10.2316/P.2013.801-081

Fig. 1. Bragg's peak

475

Restore the position of the patient before radiotherapy on bit many articles, and the effect is significant, while the motion detection in the course of treatment there are several methods have been proposed. As the laser triangulation point algorithm proposed by Yoshikazu Tsunashima et al [1] is a high-speed and highprecision laser machine as the light source, point-by-point scanning light source to skin distance to monitor the moving of the skin. Restore the three-dimensional image of the object to be measured at the same time he also Tsai stereo image correction method [2] by about two cameras will object to be measured with reference to the object coordinates to the Fourier transform, the detection rate, however, only two pictures per second and only for a small range of motion detection. James M. Parkhurst et al using the fringe projector will cosine stripes of light irradiated on the surface to be detected objects. In this case, the cosine stripes of light with this surface distortion using the camera on the surface deformation cosine light streaks shooting into the computer analysis of Fourier transform caused distortions change cosine light stripes representing the surface of the skin, because the surface degrees, it can be reconstructed through Fourier transform of a 3D stereoscopic image, and then tracking the changes in this stereoscopic image in order to achieve the detection of the displacement, which we call the cosine light projection method [3]. Qi-feng Yu et al proposed a three-dimensional dynamic optical sensing method [4] is a kind of precision calibrated rectangular grid, after specific objects marked characteristics than on the way to recognize objects, when two of the three-dimensional image of the images can be found in the object corresponding to complete the identification, take this object corresponding to the completion of the reconstruction of 3D position, the same track 3D object recognition to achieve detection of displacement. Hiroki Shirato real-time tumor tracking system [5] contains four sets of diagnostic X-ray TV systems, image processing unit, a gateway controller (Gating) and an image display unit. X-ray TV rate of 30 times per second identification tracking a 2.0mm metal mark, this mark is the image-guided tumor implantation or near the location of the tumor. If the flag in excess of the allowable displacement range of the switch, This system will alert to stop radiotherapy to avoid the damage caused by patient action displacement close ray treatment. Yulia Lyatskaya infrared image sensing method [6], the system by the circle of infrared emitter and reflector 180 degrees flag surrounded by a group of infrared cameras composition, the receiving reflected infrared from Infrared camera in order to calculate the to detect objects in a 2D plane, when using two sets of infrared cameras can obtain 3D coordinates information, then use continuous infrared signs Polaris tracking system to detect 3D coordinates. Nothing more than to use the above-mentioned motion detection and the following contour detection

technology, such as Canny [7] to find out the optimized function impending order Gaussian derivative Haralick [8]more zeros to identify the location of the edge of the gradient of the second countdown Torre and Poggio [9] and the Laplace transform and second derivative of the gradient function do insightful comparison and analysis. Pablo Arbela 'et al published the gPb and gPb-owt-ucm called boundary probability algorithm [10], this particular method For the analysis of image segmentation for a small disc, then this small disc cut into two semicircular calculated to calculate the value of each angle of the two semicircular square value, when the two discs of the histogram, to the angle of the centerline of the square value of the angle of the centerline of the two semicircular straight (histogram), and rotating the split semicircle the value of the maximum difference value, the current point is the direction of the contour, and gradually the whole image of the center line of the small disk connected together to form a contour detection. Therefore, the smaller the diameter of the disc divided image contours depicted more clear, but the time it takes the relative increase in terms of 2 frames per second calculation, the resolution can only be up to 512 × 512 pixels. These the contour recognition technology generally listed above require complex mathematical calculus, Fourier transform, the value of the square map and floating-point calculations, is not only time-consuming and can not achieve real-time detection, detection range is too small, it can not meet the present stage clinical radiotherapy quality requirements. Moreover, such a detection speed is only found in the patient breathing this slow and regular displacement. Rapid and transient displacement of pumped storage, cough, etc. need to be more rapid, effective motion detection method can meet. The proposed ScanNet detection method can achieve a resolution of 1920×1080 Proton radiation therapy to meet clinical needs, and a speed of 25 images per second to detect displacement.

2. Methods Most of the post is to identify the whole image to detect displacement, and then a long time to recognize the need to consume the entire image, it is difficult to simultaneously achieve high resolution and real-time identification and location of the true cancer treatment hot zone often only partial range (Fig. 2), only a small portion of the entire image. The other arm as shown in the displacement will not affect the red tumor field, identification entire image because the arm action and produce false warning message, or even interrupt the treatment. And our ScanNet method by operating interface motion detection range to focus on the treatment area to avoid misjudgment phenomenon. The our proposed ScanNet composed by the vertical and horizontal of the three scanning lines, and the position of the six scan lines are available to by Scroll Bar intuitive adjustments on the user interface, the location of the

476

scanning lines For monitoring means that region, that is the location of the software algorithms.

position is the same, can be calculated to detect the amount of displacement of the object, while monitoring the displacement of the reference point is beyond the set to one can know whether the displacement of the tumor area exceeded the allowable threshold range. This method only needs a simple expression, counting only the position of the scan line, up to 25 frames per second real-time detection. As shown in the marquee area in Fig. 3, the arm on the left is the brightness of the lighter areas,

Fig. 2. Proposed ScanNet method In order to achieve the rapid computation of contour position, we use to like Ali Habibi [9] described differential pulse code modulation (Differential Pulse Code Modulation, DPCM) algorithm, which is used in the data compression algorithm, and its purpose in the record before and after the point differences in luminance value, record the difference in value to replace all of the pixel data, so you can reduce the amount of data that has reached the purpose of data compression, the difference luminance value just right can be used to recognize a video contours. As shown in Fig. 3, the dotted black line marquee at the range to be analyzed, the blue line represents the location of the scan line through the place of the operation, i.e. ScanNet. Image data captured from a video camera came in, we first red, green, and blue colors add up to the average, this value is the value of the brightness of the pixels, our next operations are based on the brightness value to operators.

Fig. 4. Reference Point Algorithm Therefore, the higher values of the items of information, rough such as array in the hexadecimal C0h ~ C1h between the first three values (Fig. 4 series), began after the fourth value, the value plummeted to about 30h 35h. Next we will be behind the array values (Fig. 4, the first four) minus the current array value (Fig. 4, the first three) come to the third point for 8Fh (Fig. 4), that is the image contours where, it is also the position of the reference point we are looking for. Meanwhile, in the algorithm to find a reference point, to the back of the pixel point by subtracting the current pixel point, this step is exactly the DC component of the signal can be filtered out, this way can also dispense with the operation of the removal of the DC component steps. In the identification the resolution aspects as shown in Fig. 5, Images captured by Camera horizontal direction of the image point factors of up to 1920 pixels vertical direction up to 1080 points, in order to facilitate the calculation, we discard the left and right edges of the image just take the middle of 1800 effective pixels, normal shoulder width of about 90 cm, so our identification resolution of up to 90 cm / 1800 pixels = 0.5 mm / pixel. The sampling frequency must be twice the relationship, but also take into account the impact of the noise of the camera, we to take 1mm/pixel do know resolution proton radiation therapy required accuracy requirements. However, in the case of the same, each to capture images of the same object, their pixel values are not the same, i.e. the so-called Gaussian noise. The common solution is average consecutive number of images.

Fig. 3. Illustrated Scan Line As shown in Fig. 4, we select the luminance value of the dotted box on the array in order to later remove the operation, the value represented by the value of the position of the array, the array within the point value for the next pixel by subtracting the value of the current pixel point each point, and the difference value of the next point, when the amount of change of this value is expressed in hours before and after the point of this pixel is a smooth image. We set a threshold to determine the difference in the value of the change can be ignored, which represents the location of the image contours When this difference value is greater than the set threshold, we will define it as a reference point, we only need to remember live in the position of this reference point, this reference point and the next frame in the same position of the reference point

477

R[x]=0; Trigger[x]=0; Leave[x]=0;

A

Get R[x]

R[x] is Ref.?

NO

Yes

Fig. 5. ScanNet resolution Another noise in adjacent pixels in the sudden appearance of white spots or black spots, the pixel value 00h or ffh in randomly appear in the captured image. Black color similar to the salt, white pepper black and white image point is called the point of white pepper, black images known as salt. Such noise usually has a higher frequency, into the princess does not seem with the surrounding pixels, and therefore easy to distinguish. As long as we use a simple low-pass filter can be eliminated. Identify the reference point coordinates of each of three of the vertical and horizontal scanning line after the first filtered noise, noise filtering method is used herein, are filtered out from the timeline and the reference point exceeds the threshold, method shown in Fig. 6 shown. The reference point coordinates of the frame in the steps outlined in the reference point is described in the buffer array with the next frame in order to facilitate comparison, it is assumed could not within the threshold value before and after the current reference point position on a map the location of the box allowable position (that is, the amount of displacement is less than the allowable threshold) to find a reference point, tired by the variables in the Leave. Leave variable value exceeds the tolerance threshold set represents the true amount of displacement has exceeded radiotherapy allowed error value, the system will issue a warning signal to alert the radiologists' attention to patient dynamic or treatment reference. Conversely, if the reference point is the recurrence of the allowable threshold will Leave zero, and determine the previous Leave for noise, down an array value calculation, the calculation is completed until the reference point of the entire column.

Check last Frame exist Ref.?

No

Yes

Leave[x]=0;

Enter[x]++

No Next R[x]

Enter> Threshold? Yes

A

Check last Frame exist Ref.? No Yes

Enter[x]=0;

Leave[x]++

Leave > Threshold?

No

Yes A

Fig. 6. Noise reduction follow chart ScanNet the development environment is Windows XP, or Windows 7 under in C + + language to write and use Microsoft MFC programming framework to accelerate Windows program development, and Microsoft Visual Studio 2008 to compile complete the image data from the standard Image Class Loading the use of Open source code of the library.The main body of the program is a DiaglogBox Fig. 7 Shown: WM_INITDIAGLOG will be the first to receive the message which issued by the operating system. We will be all kinds of control components initialization done here. for example, ScrollBar, DropDownList. And initial default value,

478

Finally we to the update frequency SetTimer () to set ScanNet 4ms (25fps cycle is 1/25 = 4ms sec). The main program then enters a big loop waiting for parsing messages, when the information comes from the operating system, the main program to PeekMessage Translate Message the corresponding message sent to the corresponding subroutine, and we want to intercept processing WM_TIMER messages. Practice in the initial DialogBox to SetTimer () as a set Timer, when the Timer ID is equal to the set value of the object, then retrive an entire frame of image data for display and our ScanNet only from the scan line after the path to calculate the location of the reference point, and then detect the amount of displacement, displacement amount exceeds the set threshold, the alert message, otherwise it will be continued next time the Timer

In order to simulate the displacement movement and instrument calibration required, we need a verification system to ensure the system's reliability in ScanNet, many articles have been published to the simulated patients displacement writings published Mesic, S. et al. the method is based on a computer-controlled pneumatic pump to simulate lungs breathing movements [12], realistic action, as well as its advantages can simulate lung expansion and contraction

Fig. 8. ScanNet User Interface Kenneth H. Wong et al analog computer-controlled three-axis motor mechanically breathing action [13], the advantage is accurately simulate various displacement action. Two ways listed above, although its action is realistic and delicate, but the disadvantage is the high production costs, while our aim of a correction ScanNet system, simply and precisely control the amount of displacement, so we do not need to be realistic simulated lung change in shape of the portion respiration. For this purpose, we have produced a low-cost and can be changed at any time of the operation mode displacement simulator. We use Texas Instruments MSP430 for control unit, it is manufactured for sale in the simulator control unit and very popular on the market and a wide range of applications, l module, the stepper motor driver, stepper motor and reduction gearbox module. The simulator architecture is shown in Fig. 9 Shown. Fig. 7. ScanNet Main Program Flow Chart The program interface shown in Fig. 8 Shown, the screen resolution in general only up to 1920 x 1080, all images are displayed on the screen, there is no room for placement of the control elements, so in order to fit the screen resolution makes the operation more convenience, we will adjust the image size to half the pixels, but the program internal computing is still based on a pixel to calculate. Below H1, H2 and H3 for the scanning position of adjustment of the horizontal direction, the right of V1, V2 and V3 is the adjustment of the vertical direction in order to be adjusted for the monitoring location. Upper right compared to the threshold level to adjust the left in the figure underwear and arm junction point detected as a reference point.

Fig. 9. Movement Simulator System

479

The aim of our monitoring the displacement and displacement speed of the action, our simulation curve only simulate normal breathing, coughing and trembling of the three non-autonomous human action. Wherein the coughing action short and difficult to predict, the simulation curve cough we take Matthew Barlow [14] The data obtained in the experiments, their cough length of time of about 0.1 seconds, the maximum amplitude is about 15mm, in Fig. 10 Shown .

Displacement simulator interface as shown in Fig. 13, The left panel is divided into breathing, coughing, and pumped storage action simulation, Delay lever Adjustable actions Adjustment speed Movement NC Joining Module can adjust the action steps, the right panel can display control panel word sent to the stepper motor control, continuous operation of options for continuing to simulate the action.

Fig. 13. Movement simulator system user interface Fig. 10. Cough Curve Gearbox output is a rocker, shown in Fig. 14 in the orange part. First, we define the reduction ratio of the gear box to 1:10, assuming the gear box rocker required displacement distance is X = 1mm, the length of the rocker arm 30 mm in the amount of displacement of the simulator rocker desired rotational angle as θ.

Mesic, S.[12], a graph of the normal breathing, the experimental values in Fig. 11 Shown, the length of time is about 2 seconds, the Y axis coordinates represented in the figure is the lung pressure, units kPal, we try to figure out the actual amplitude is about 10mm.

Fig. 14. Rocker of Gear box

Fig. 11. Respiration Curve

Calculation formula such as (1) shows, the number of steps of the stepper motor rotation of the step 500, so that each step rotation of 360/500 = 0.72 degrees, converted into a rotary rocker angle of 1.9 degrees, the stepping motor required walking The number of steps equal to 1.9 / 7.2 * 10 = 26 steps.

Trembling movements and the actual literature available, the length of time we assume that its action is about 0.1 seconds, the maximum amplitude of about 5 mm. These three actions curve simulation shown in Fig. 12 Shown, gray is representative of normal breathing, coughing red represents Purple displacement simulation trembling. MSP430 convert these curve to control code to the stepping motor driver to driving the stepping motor, and then through the speed reduction gear in order to achieve more delicate movements.

if ( x = 1mm), θ =

20 10 0 0.5

1

1.5

2

2π

1 30 × 3600 = 1.90

(1)

In the development stage, the simulation curve is controlled by the computer according to the formula (1) is calculated to be simulated action stepper motor steps after the calculation results sent to the MSP430, for quick fixes, pending future systems mature and then the simulation curve move the code to the MSP430 specifically calculated, and this way control parameters such as the distance and speed of the computer simply transfer the

(mm)

0

tan −1

2.5

Fig. 12. Movement Simulator Curve

480

displacement movement of the action to the MSP430 can control computer can concentrate on the calculation and monitoring of displacement, in order to control in the same computer manipulation multiple cameras. The MSP430 firmware process as shown in Fig. 15. The firmware by the C++ language drilled write and compile the completion of the Microsoft Visual Studio 6.0. The program is started by the main (). First, we will only use MSP430 internal Clock as the clock 1MHz, then will P1SEL, and P1DIR with the UART TXD transmission termination, the P1IES and P1IFG and RXD together the UART control computer is used to set communication GPIO's PORT2 divided into two fourtuple high byte reserved for other control signals output pin, four low byte bit just as stepper motor drive control four output word. When the main () to initialize these registers setting into a main loop, loop constantly monitors the input port of the UART, once the data from the control of the client computer passed them resolve parse the command into the corresponding equations to calculate modeled as stepper motor control byte port2 output current amplification module to drive the stepper motor, and then the program back to the main loop, waiting for the next command. In addition MSP430 Watch Dog function, once the program can not be updated within a set time, the Watch Dog Timer MSP430 self restart in order to avoid the crash phenomenon, our firmware only perform a single hardware architecture of the Thread, the future will be Timer for time division analog operation of the multi-program execution in order to complete side of the output, operation, and the other side can also respond to the control side of the command.

simple addition and subtraction and multiplication and division shift register instead part, not the complex calculation of floating-point arithmetic or calculus, so much faster, compared to the prior art from the entire image to identify the contours of the implementation costs due to the use of desktop video camera, did not using precision or special equipment, so the overall cost can be controlled in less than $ 3,000. Its previously published monitoring is relatively cheap, and easy to operate without default any reflective metal objects in the patient body, this feature is also increased clinical Applicability. And ScanNet method can detect a precision of 1 mm, speeds of up to 25 frames per second, without placing any object on patients who will be the future additional ZOOM IN lens can detect three meters 70 cm shoulder width can be meet the diverse needs of the Proton radiation therapy. Due ScanNet required computation time is extremely short-lived, have been much larger than the image from the camera returns control the speed of the computer, so we can be with a control the computer operating simultaneously three-axis direction ScanNet signal, we plan these three cameras were installed located in the left side, right side, with the upper end of the radiation therapy bed, so that when the treatment of mechanical arm (Gantry) not in rotation when blocked ScanNet the operation to avoid the camera dead ends or treatment arm blocking, so we will proceed in the future tri-axial displacement detection in order to improve the feasibility of clinical applications. ScanNet pixel is red, blue and green pixel values of the three channels together averaged, known as the luminance value to do the subsequent operations, the future Reflections make judgments in order to increase the ability to identify the single color respectively. Various threshold settings will affect ScanNet judgment, we have to proceed with clinical trials program, and hope to determine the optimal threshold set down in the actual environment of the treatment room. The number of scanning lines also determines the operation of the entire system, the target can be more representative of the scan line can be detected more delicate, but the computation time is relatively increased, so find the optimum number of scan lines in order to optimize the system is also our next one. Photography lens USB2.0 interface transfer to the control computer, however, the USB2.0 bandwidth can only reach 450Mb/sec, unable to meet the bandwidth of 30 frames / sec full HD (1920x1080), when the future USB3.0 camera popularity, its bandwidth will be able to meet the higher image resolution and frame number has reached higher detection accuracy required. Actual displacement simulator to verify ScanNet displacement detection system to allow a the displacement simulator repeat Table 2. Listed three actions hundred times each, displacement of ScanNet detection results are as follows: in the displacement of about 0.5mm action simulation successful identification rate of about 25% only, this may be the noise caused by the relationship, and

START

Set Clock=1MHz P2=output P1SEL=TXD P1DIR=TXD No Data Received P1IES=RXD P1IFG=RXD

Yes Calculate Curve Output to P2

Fig. 15. MSP430 Firmware Flow Chart

3. Conclusion ScanNet detection method to calculate the treatment of hot zone of the horizontal and vertical directions 9 scan lines to replace the whole image and the algorithm only

481

the displacement of more than 1mm can be 100% successful identification. The ScanNet reliability up to 100%.

Cybernetics, 1996., IEEE International Conference on , vol. 1, no. 1, pp.484-486 vol.1, 14-17 Oct 1996 [6] LYATSKAYA, Y.. Performance and characteristics of an IR localizing system for radiation therapy. Journal of Applied Clinical Medical Physics, North America, 7, may. 2006. Available at: . Date accessed: 23 Nov. 2012. [7] Canny, John; , "A Computational Approach to Edge Detection," Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.PAMI-8, no. 6, pp.679-698, Nov. 1986 [8] R. M. Haralick, “Zero-crossing of second directional derivative edge operator,” IEEE Trans. Pattern Anal. Machine Intell., vol. PAMI-6, no. 1, pp. 58-68, Jan. 1984. [9] V. Torre and T. A. Poggio, “On edge detection,” IEEE Trans. Pattern Anal. Machine Intell., vol. PAMI-8, no. 2, pp. 147-163, Mar. 1986. [10] Arbeláez, P.; Maire, M.; Fowlkes, C.; Malik, J.;,"Contour Detection and Hierarchical Image Segmentation," Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol. 33, no. 5, pp.898-916, May 2011. [11] Habibi, A.; Hershel, R.; , "A Unified Representation of Differential Pulse-Code Modulation (DPCM) and Transform Coding Systems," Communications, IEEE Transactions on , vol. 22, no. 5, pp. 692- 696, May 1974. [12] Mesic, S.; Babuska, R.; Hoogsteden, H.C.; Verbraak, A.F.M.; , "Computer-controlled mechanical simulation of the artificially ventilated human respiratory system," Biomedical Engineering, IEEE Transactions on , vol. 50, no. 6, pp.731-743, June 2003 [13] Wong KH; Tang, J.; Dieterich, S.; Zhang, H.; Zhou, T.; Cleary, K.; , "Respiratory motion compensation studies using a 3D robotic motion simulator and optical/electromagnetic tracking technologies," Nuclear Science Symposium Conference Record, 2004 IEEE , vol. 4, no. 1, pp. 2652- 2655 Vol. 4, 16-22 Oct. 2004 [14] Barlow, M.; Varadan, V.V.; Jia Di; Khoa Phan; , "A Compact, Self-Contained Cough Monitoring System," Region 5 Technical Conference, 2007 IEEE , vol. 1, no. 1, pp.167-170, 20-22 April 2007

Table 2. The results of motion detection in breathing simulation. Test Cycle

Simulator

Scan Net

1

0.5 mm

× (25%)

2

1.0 mm

 (100%)

3

1.5 mm

 (100%)

4

2.0 mm

 (100%)

Displacement simulator now plastic model gear stepper motor deceleration, but the gap is too large plastic gear error caused the future, we also consider the shift to a precision screw gear gap is too big to solve in order to simulate a more precise displacement action. Changed significantly due to modern radiation therapy for patient privacy and warmth demand, patients who will cover a thin blanket, so that the image contour is not obvious, we consider the production of lattice cloth blanket Instead, in order to improve the S / N ratio, in order to identification, but will also allow easier threshold setting.

References [1] Yoshikazu Tsunashima, M.SC., Takeji Sakae, PH.D. and YOSHIYUKI SHIOYAMA, "Correlation between the respiratory waveform measured using a respiratory sensor and 3D tumor motion in gated radiotherapy," In Proc. Information Engineering (ICIE), 2010 WASE International Conference on , vol. 3, no. 1, pp.30-33, 1415 Aug. 2010 [2] J. Tsai Roger Y., “A versatile camera calibration technique for high-accuracy 3D machine vision metrology using off the shelf TV cameras and lenses”, IEEE J. of Robotics and Automation, V0l. M-3, No. 4, 8(1987), 323344 [3] M. Takeda, and K. Mutoh, "Fourier-Transform Profilometry for the Automatic-Measurement of 3-D Object Shapes," App. Opt., vol. 22, no. 24, pp. 3977-3982, Dec. 1983. [4] Parkhurst, J.M.; Price, G.J.; Sharrock, P.J.; Marchant, T.E.; Moore, C.J.; , "Next generation optical surface sensing for real-time measurement in radiotherapy," in Proc. Engineering in Medicine and Biology Society, EMBC, 2011 Annual International Conference of the IEEE , vol. 1, no. 1, pp.6170-6173, Aug. 30 2011-Sept. 3 2011 [5] Qifeng Yu; Dongsheng Zhang; Zhihui Lei; Tiehan Quan; , "Accurate measurement of 3D coordinate of an object with subpixel technique," Systems, Man, and

482