|Year : 2017 | Volume
| Issue : 5 | Page : 50-58
AFOMP Best Paper
|Date of Web Publication||24-Oct-2017|
Source of Support: None, Conflict of Interest: None
|How to cite this article:|
. AFOMP Best Paper. J Med Phys 2017;42, Suppl S1:50-8
| AB-1: Prototype Development of Artificial Intelligence-Based Proton Therapy Planning System|| |
Yong-Jin Kim, Do-kun Yoon, Sunmi Kim, Han-Back Shin, Moo-Sub Kim, Tae Suk Suh
Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea. E-mail: [email protected]
Introduction: People are being helped of an artificial intelligence (AI) in many areas of life. And the AI is also can be used to progress a radiation therapy, It helps to establish a more accurate and faster treatment plan. In this study, we have studied the application of the AI to treatment planning acquisition for proton therapy, which requires more accuracy in dose calculation, to establish the treatment plan faster and accurately.
Objectives: The purpose of this study is to develop and operate an artificial neural network system of AI based radiation treatment planning system for proton therapy.
Materials and Methods: For a design of an artificial neural network, we establish the radiation treatment planning method by using convolution neural network model with some modifications. When a new input (patient data) data has been uploaded, it assigns a proper treatment plan through a supervised learning technique from a database. And it progresses the optimization for treatment plan through weight decay process. This mechanism leads to faster and more correct treatment plan creation than the conventional dose planning based on the either algorithm or Monte Carlo. This database was constructed by using a lot of results from Monte Carlo simulation operation. And it was interlocked with the deep learning algorithm.
Results and Discussion: In this study, we confirmed the target volume from convenient image using the developed algorithm. We successfully created the treatment plan by accessing the Monte Carlo database. In comparison with the treatment plan generation through direct Monte Carlo simulation, we deducted faster and equivalent therapeutic performance analysis results from the study. In terms of time, we succeed greatly fast acquisition of the proton therapy plan. For the aspect of accuracy, we have to improve the performance through the additional development of better optimization algorithms. In the future, we will progress the study to develop a complete engine.
| AB-2: Graphics Processing Unit-Based Fast Imaging Technique During Boron Neutron Capture Therapy: Monte Carlo Simulation Study for Single-Positron Emission Tomography Operation|| |
Hye Jeong Yang, Do-Kun Yoon, Han-Back Shin, Moo-Sub Kim, Sunmi Kim, Tae Suk Suh
Department of Biomedical Engineering, College of Medicine, The Catholic University of Korea, Seoul, South Korea. E-mail: [email protected]
Introduction: Boron neutron capture therapy (BNCT) is an effective radiation treatment technique based on the nuclear reaction between an epi-thermal neutron beam and boron particles. An alpha particle which is main therapeutic factor is emitted from the boron neutron capture reaction. Simultaneously as the boron neutron capture reaction was progressed, the prompt gamma ray of 478 keV is emitted from the same reaction point. The emitted prompt gamma ray can be used to gauge tumor status during the radiation treatment. However, it is difficult to monitor the therapeutic effect of BNCT in real time.
Purpose: Purpose of this study was to show the graphics processing unit (GPU) based fast prompt gamma ray imaging technique during BNCT using the Monte Carlo simulation.
Materials and Methods: To acquire the 478 keV prompt gamma ray image in a single step, positron emission tomography (PET) with an insertable specific collimator for single photon detection (S-PET) was simulated as shown in Figure 1. And then in order to perform a fast prompt gamma ray imaging, we were attempted to reconstruct the prompt gamma ray image using a modified reconstruction algorithm with GPU computation during BNCT Figure 1.
|Figure 1: Simulation configuration of the acquisition of both the prompt gamma ray image and the positron emission tomography image. Neutron accelerator, water phantom including boron uptake regions, and positron emission tomography detector were simulated. In the right figure, the insertable collimator has been added (circular violet structure)|
Click here to view
Results and Discussion: Single photon detection from PET gantry with the insertable collimator was available during BNCT and prompt gamma ray of 478 keV had a capacity to obtain a tomographic image as shown in Figure 2. Moreover, prompt gamma ray images according to the treatment fraction were acquired almost immediately after the acquisition of projection data owing to GPU based image reconstruction. As a result, we could confirm the application feasibility of the fast prompt gamma ray technique for BNCT.
|Figure 2: Diagram of the original pattern and reconstructed image depending on boron neutron capture therapy procedure. The original pattern shows the tomographic of the virtual water phantom. The BURs are labeled clockwise as A-E. On the right column, the positron emission tomography and prompt gamma ray images were reconstructed with a modified reconstruction algorithm|
Click here to view
Conclusion: Due to the fast GPU computation, GPU application for BNCT S-PET has a large effect on reduction of reconstruction time. Therefore, it was possible to reconstruct GPU-based fast prompt gamma ray image during BNCT.
| AB-3: TLD Correction Factors for Field Sizes Used in Lung Stereotactic Body Radiation Therapy Dosimetric Measurements|| |
Roger Cai Xiang Soh1,2, Guan Heng Tay3, Wen Siang Lew1, Shaun Baggarley2, James Cheow Lei Lee1,3
1Division of Physics and Applied Physics, School of Physical and Mathematical Sciences, Nanyang Technological University, 2Department of Radiation Oncology, National University Cancer Institute Singapore, National University Hospital, 3Division of Radiation Oncology, National Cancer Centre Singapore, Singapore. E-mail: [email protected]
Introduction: TLD dosimetric verification for Lung Stereotactic Body Radiation Therapy (SBRT) may be a challenge due to small field TLD dose perturbation within the low-density lung medium. Accurate measurement can be achieved by correcting the perturbed dose using TLD correction factors calculated by Monte Carlo (MC) methods. The current study investigates the MC calculated TLD correction factors against depth and field size within a proposed lung phantom.
Materials and Methods: 14 TLD-100 measurements were done along the central axis within a lung phantom, which consists of CIRS plastic water slabs sandwiching 14 pieces of composite cork slabs. TLD correction for the positions in composite cork was determined by taking the ratio of the Monte Carlo simulated dose response of TLD in cork to the Monte Carlo simulated dose response of TLD in water. BEAMnrc and DOSXYZnrc user codes were used for simulation (EGSnrc, National Research Council of Canada, Ottawa, ON).
Results: It was found that the greatest TLD correction was for the smallest field size investigated, 2 x 2cm2, where electronic disequilibrium was found to be the greatest as shown in Figure 1. Corrections at inhomogeneous interfaces were dependent on the medium before the interface. In addition, MC corrected TLD measurements agree with Acuros External Beam (AXB) 2 x 2 cm2 predicted depth doses to within 2% as shown in Figure 2.
|Figure 1: Monte Carlo simulated TLD correction factors with respect to depth|
Click here to view
|Figure 2: Monte Carlo TLD corrected measurements against AXB 2 x 2 cm2 predicted dose profile|
Click here to view
Discussion: TLD perturbation is dependent on three main aspects, namely, lateral charge particle disequilibrium, over response of TLD in broad beam and partial volume averaging. Due to the small physical size of TLD-100, partial volume averaging was deemed negligible in this study. Lateral charge particle disequilibrium was found to be dependent on field size, the TLD material, and the density of the phantom. The electron range in low-density materials, such as composite cork, is longer than the range in water. Hence, lateral electronic disequilibrium occurs predominantly in composite cork. Furthermore, under small field conditions, the Compton electronic range is greater than the distance between the point of measurement to the field edge, increasing the probability for an electron to transfer its energy outside the field, resulting in greater electronic disequilibrium.
Lastly, the over response of TLD in broad beam was due to TLDs having a higher atomic number than that of water. 6 MV photon beam consists of an extensive amount of low-energy photons as compared to higher energies photon beam. Due to larger cross-section for photoelectric effect in high atomic number materials within TLDs, low-energy photons will be absorbed to a high extent. This low energy photon absorption increases with increasing field sizes.
Conclusion: The results presented in this study indicate that corrections will be required for in-vivo TLD measurements in the proposed lung phantom for Lung SBRT dosimetry measurements. Monte Carlo simulated TLD correction factors calculated were dependent on the depth of measurement and field sizes. The greatest TLD correction was for the smallest field size investigated, 2 x 2cm2. Further studies may include the effect of TLD angular dependence on TLD correction factors.
| AB-4: Investigation of Optimal Shutter Scan Acquisition Parameters in Digital Tomosynthesis System|| |
Dohyeon Kim1, Byungdu Jo2, Haenghwa Lee2, Donghoon Lee1, Sunghoon Choi2, Hyemi Kim2, Zhen Chao1, Seungyeon Choi1, Hee-Joung Kim1,2
Departments of 1Radiation Convergence Engineering and 2Radiological Science, Yonsei University, Seoul, South Korea. E-mail: [email protected]
Introduction: Digital tomosynthesis system has been studied to reduce the exposure dose. Nevertheless, many studies have suggested that it is still needed to reduce the exposure dose in digital tomosynthesis system. Diagnostically important information is often concentrated on a region of interest (ROI) in a reconstructed 3D image. For this reason, ROI imaging techniques are considered to be a reasonable dose reduction method. If the ROI reconstruction method is applied to digital tomosynthesis system, it could expect a better dose reduction effect. But most studies do not focus on improving the image quality of the overall anatomy particularly in the outside ROI. Therefore, we proposed shutter scan acquisition for region of interest (ROI) imaging to reduce the patient exposure dose in digital tomosynthesis system. The purpose of this study was to investigate the effect of composition ratio of truncated and non-truncated projections and to determine the optimal set of acquisition parameters for the proposed shutter scan acquisition in digital tomosynthesis system.
Materials and Methods: Projections obtained by shutter scan acquisition consist of truncated and non-truncated projections as shown in Figure 1. A prototype chest digital tomosynthesis (CDT) system (LISTEM, Korea) and the LUNGMAN phantom (Kyoto Kagaku, Japan) with 8 mm lung nodule were used for this study. In this study, we call the number of truncated projections divided by the number of non-truncated projections as shutter weighting factor. The shutter scan acquisition parameters were optimized using 5 different acquisition sets with the shutter weighting factor (0.1, 0.3, 1, 3 and 7). A total of 81 projections with shutter scan acquisition were obtained in 5 sets according to shutter weighting factor. The image quality was investigated using the contrast noise ratio (CNR). We also calculated figure of merit (FOM) to determine optimal acquisition conditions for the shutter scan acquisition. The total effective dose was used as the dose value for calculating the FOM.
|Figure 1: (a) Schematic diagram of the shutter scan acquisition, (b) obtained projections in shutter scan acquisition|
Click here to view
Results: The ROI of the reconstructed image with shutter scan acquisition showed enhanced contrast as shown in Figure 2. The CNR value was highest when the shutter weighting factor was 1. The highest CNR value, shutter weighting factor 1, is the acquisition set consisting of 41 truncated projections and 40 non-truncated projections. On the other hand, the result of the FOM value was the highest value when the shutter weighting factor was 3. The highest FOM value, shutter weighting factor 3, is the acquisition set consisting of 61 truncated projections and 20 non-truncated projections.
|Figure 2: Reconstructed images: (a) conventional acquisition in 37th slice, (b) conventional acquisition in 25th slice, (c) conventional acquisition in 28th slice, (d) shutter scan in 37th slice, (e) shutter scan in 25th slice, (f) shutter scan in 28th slice|
Click here to view
Discussion: It is expected that the patient exposure dose can be reduced by limiting field of view (FOV) to focus on region of interest in proposed shutter scan acquisition. We investigated the effects of composition ratio of the truncated and non-truncated projections on reconstructed images through the shutter scan acquisition.
Conclusion: The optimal acquisition conditions for the shutter scan acquisition were determined by deriving the figure of merit (FOM) values. Our results suggest possible directions for further improvements in digital tomosynthesis system for shutter scan acquisition method.
| AB-5: Study of Feasibility of Using a Complementary Metal Oxide Semiconductor Based Mobile Phone Camera as a Radiation Dosimeter|| |
Josmi Joseph, I. Rabi Raja Singh, S. Amalan, E. Winfred Michael Raj, B. Paul Ravindran
Department of Radiotherapy, Christian Medical College, Vellore, Tamil Nadu, India.
E-mail: [email protected]
Introduction: Complementary metal oxide semiconductor (CMOS), is the fundamental sensor used in almost all digital imaging devices including smartphones. Some of the studies show that this sensor responds to ionizing radiation also and it is directly proportional to the exposure of incident radiation. There are already some downloadable applications (apps) available for detecting the stray ionizing radiation but not for measuring the radiation. To use CMOS as a radiation dosimeter which is not fully energy independent, individual calibration is necessary to account for the variation in response for different energies of radiation and also for different sensor. As a preliminary study, a separate Android application was developed and with the same, the characteristics such as buildup effect, energy dependence, dose linearity, consistency, angular dependence and relative dose factor measurement were examined and analyzed to check whether this sensor can be used as radiation dosimeter.
- To develop an android application to use the CMOS sensor as radiation dosimeter
- To study the following characteristics using the developed app, And to compare the same with the standard ion chamber values, Thermoluminescent dosimeter (TLD), Optically stimulated luminescence dosimeter (OSL) and semiconductor diodes.
- Percentage Depth Dose
- Dose linearity
- Energy dependence
- Angular dependence
- Relative dose factor measurement
Materials and Methods: When a CMOS is exposed to ionizing radiation, specks proportional to the number of incident photons will be formed on the frame. An android application has been developed to quantify the specks. It includes,
- Calibration - to determine the threshold pixel value above which the pixel is to be considered as a radiation signal
- Noise filtration - by high-delta method which will also help in determining the signal from each frame.
Using the developed app the above mentioned characteristics were studied and compared with that of standard ion chamber data.
Results and Discussion: It has been proved that the CMOS based smartphone have shown good linear response to the dose in the clinical range (100 cGy to 500 cGy). There is no need for an additional buildup material to be added to achieve the maximum ionization as the sensor has got inherent buildup due to its fabrication. It also shows good consistency. Even though the orientation of phone didn't affect the measured value, significant energy dependence was observed. This can be eliminated by performing individual calibration. The measured values from the for 100 to 500 cGy is found to be comparable with the data measured with other dosimeters such as Ionization chamber, Thermoluminescent dosimeter (TLD), Optically stimulated luminescence dosimeter (OSL) and semiconductor diodes. It is promising, that the smartphone with CMOS sensor can be used as a dosimeter for relative dose measurement even though it is found to be energy dependent.
| AB-6: Improving Image Quality in Contrast Enhanced Dual Energy Mammography by Noise Compensation Techniques|| |
Minjae Lee, Donghoon Lee, Seungyeon Choi Junyoung Son, H. J. Kim
Department of Radiation Convergence Engineering, Yonsei University, Seoul, South Korea. E-mail: [email protected]
Introduction: The incidence and the number of death due to breast cancer are continuously increasing. Conventional mammography (CM) is generally used for diagnosing breast, but it is well known that diagnosing in the dense breast with CM is difficult. Contrast enhanced dual energy mammography (CEDM) is one of the latest developments in breast diagnosis, and it is especially effective in dense breast. However, CEDM has inherent risk of side effects due to iodine contrast agents, and there is serious quantum noise problem due to dual energy imaging techniques. The purpose of this study was to confirm the feasibility of CEDM imaging using a diluted iodine contrast agent to reduce the possibility of adverse effects. In addition, various denoising algorithms were applied to CEDM in order to improve image quality.
Materials and Methods: We used a Mammographic x-ray system (SOUL, MEDI-FUTURE, Inc. South Korea) and, soft tissue equivalent slab phantom with 5cm thickness and 70% glandularity, and diluted iodine contrast agent with 5% and 10% Ioversol (320 mg iodine per milliliter). To acquire CEDM images, we used energy weighted log subtraction algorithms with low and high energy projection which were acquired with 26 kVp and 40 kVp, respectively. In addition, to reduce quantum noise in dual energy images we applied total variation denoising and non-local means (NLM) algorithms to CEDM images. We also measured the average glandular dose (AGD) of both CM and CEDM. To quantitatively evaluate the results, contrast to noise ratio (CNR) and figure of merit (FOM) were measured.
Results: The CM and CEDM images are shown in Figure 1. The CNR and FOM of the CM and the CEDM images were measured as shown in [Table 1]. Despite the use of a diluted iodine contrast agent, the signal of iodine was relatively well detected in CEDM compared to CM. Also, the results of using the denoising algorithms showed improved results. For example, CNR of CEDM with TV and NLM were about 2.4 times higher than CEDM without denoising algorithm. Although, AGD of CEDM was slightly higher than CM due to double exposures of different X-ray spectra, FOM, which considers both patient dose and image quality, was high in following order CEDM with NLM, CEDM with TV, CEDM without denoising and CM.
|Figure 1: The conventional mammography image and contrast enhanced dual energy mammography without denoising and contrast enhanced dual energy mammography images with total variation, non-local means|
Click here to view
|Table 1: The contrast to noise ratio and figure of merit of the conventional mammography image and contrast enhanced dual energy mammography images without denoising and with denoising|
Click here to view
Discussion: The CNR and FOM of CEDMusing dual energy subtraction technique are higher than that of CM, but the image quality is reduced by increasing quantum noise. The image quality of CEDM using TV and NLM is improved because the CNR and FOM are increased by reducing quantum noise.
Conclusion: We confirmed the possibility of acquiring CEDM images using a diluted iodine contrast agent. With a diluted iodine contrast agent, CEDM images combined with denoising algorithm proved to obtain relatively superior iodine signal than CM.
Acknowledgment: This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2017R1A2B2001818).
| AB-7: Multi-Modality Image Fusion Based on Improved Fuzzy Neural Network Method|| |
Z. Chao1, H. Lee2, S. Choi2, D. Kim1, P. Jeon1, H. J. Kim1,2
Departments of 1Radiation Convergence Engineering and 2Radiological Science, Yonsei University, Seoul, South Korea. E-mail: [email protected]
Introduction: In clinical applications, single modality images do not often provide enough diagnostic information. For instance, CT images clearly show bone tissue information, but for soft tissue, especially invasive tumors, CT images cannot correctly display boundaries. On the contrary, MRI images are more sensitive in soft tissue, which are helpful to determine the range of lesions, but the effects on bone tissue are not obvious. Thus it is necessary to correlate one modality of medical images to another to obtain more useful information. In recent years, neural network technique has been applied to medical image fusion by various researchers, meanwhile, many training methods for neural networks have been implemented, but the effects are not obvious. Therefore we produced a new system that based on fuzzy theory and Radial Basis Function Neural Network (RBFNN), then we proposed a new method (the combination of gravity search algorithm (GSA) and error back propagation algorithm (EBPA)) to train neural network to correct the parameters of the network.
Materials and Methods: New fuzzy-RBFNN system includes 5 layers: input layer, fuzzy partition layer, front combination layer, inference layer and output layer. In input layer, the neurons represent the images that need to be fused; In fuzzy partition layer, the membership functions are used to represent the neurons in this layer; In front combination layer and inference layer, the neurons represent the formulation of fuzzy rules and the output of fuzzy rules respectively; In output layer, the neuron represents fused image. Through this system, we firstly obtained actual output data after inputting two images to be fused. Ideal output data (training samples of fuzzy neural network) were then obtained by maximum value method. Next we trained the system by the combination of gravity search algorithm and error back propagation algorithm: original data required by GSA were obtained through EBPA and trained through GSA (we took the parameters of membership functions and the weighted factors between inference layer and output layer as original data). Finally we acquired fused images by the trained system. In this experiment, we used two sets of head images (CT, MRI and CT, SPECT). Entropy (H: the richness of image information) and Mutual Information (MI: the degree of correlation with the original variables) were used to quantitatively compare error backpropagation and particle swarm method.
Results: The results showed that error backpropagation method and particle swarm method generated worse results by showing lower H and MI values. On the contrary, medical image fusion based on our new method presented much better results with the highest performance H and MI.
Discussion: In this study, by adding more hidden layers, our improved fuzzy-RBFNN system can more clearly express the structure of input data and the function of neural network compared with conventional Radial Basis Function neural network. In addition, we adopted a new method to train the neural network.
Conclusion: The experiments of image fusion prove that our new system can fuse medical images better than the other two methods, and a new training method can effectively improve the function of fuzzy-RBFNN system.
| AB-8: Estimation of Virtual Source Position for Electrons with Regular And Irregular Fields from a Medical Linear Accelerator|| |
Doke Hanuman Bajirao, K. M. Ganesh, M. Ravikumar
Department of Radiation Physics, Kidwai Memorial Institute of Oncology, Bengaluru, Karnataka, India. E-mail: [email protected]
Introduction: The superficial and shallow tumors are usually treated with cerrobend shielded electron beams. The irregular field defining inserts are placed at the distal end of the electron applicators. When the treatment is delivered at an extended distance, the electron output and the percentage depth dose needs to be corrected based on inverse square law from the electron source position. As the electron beam emerging from an accelerator exit window undergoes complex multiple scattering in the scattering foil, the beam monitor chambers, the X-ray collimators, electron applicators, field defining inserts and air column. The position of scattering foil cannot be considered as a nominal source position and the output need to be corrected accordingly. In such cases the output can be predicted accurately assuming the effective or virtual source position.
Objectives: This study determines the virtual source position for different energies using different field sizes in Clinac DHX linear accelerator. This study also attempts to find the difference between the virtual source positions of regular and irregular field sizes for different energies using 10×10 cm2, 15×15 cm2 and 20×20 cm2.
Materials: Dual energy linear accelerator Clinac DHX (Varian Medical Systems) with electron cones for different electron energy (6, 9, 12, 16, 20 MeV) and secondary standard ionization chamber, Electrometer, solid phantom used in this study.
Methods: Measurement of Virtual source position (VSP) for regular fields-For this study, electron energies ranging from 6-20 MeV in Clinac DHX linear accelerator with water equivalent solid phantom, parallel plate type chamber and Dose1 Electrometer.VSP for Electron cones of 4×4 cm2, 6×6 cm2, 10×10 cm2, 15×15 cm2, 20×20 cm2 and 25×25 cm2 of regular fields sizes were evaluated. The VSP were derived by acquiring the reading at central axis at Dmax depth for each electron energy. The readings were taken for four different SSDs such as 100cm, 105 cm, 110 cm and 120 cm in each radiation field. The measurements were repeated for all the available electron cones. The readings were taken with 100 monitor unit (MU) at Dmax with dose rate of 400 MU/min. Measurement of virtual source position for irregular fields-To measure the VSP using cut out blocks with different sizes and shapes was the VSP at each electron beam energy (6, 9, 12, 16 and 20 MeV) was used. In addition, the VSP obtaining using the method used in experiment when each cut out block (10×10, 15×15, 20×20) was used.
Results and Discussion: It is observed that the virtual source distance increases with increase in energy and field size. The estimation of virtual source distance for irregular field size using electron cones of 10x10 cm2, 15x15 cm2 and 20x20 cm2 showed larger deviation with that of regular field for lower energies and subsequently narrows down with increase in electron energy. The results indicate that actual estimation of virtual source position for particular energy and field size is necessary when irregular field is used for treatment and in calculating the accurate doses. The diffusion phenomenon made with uniform dose distribution at the depth of the reference points significantly reduces the lateral scattering effect with the thickness of the cut-out block when compared to regular field. Due to the scattering of electron beam in its path by the scattering foils, collimator, and applicator, etc. Hence, for dose calculation with electron beams by direct SSD should not be used for dose estimation which will lead to error in dose estimation.
| AB-9: Development of Voxel Based Method for Evaluation of the Radiotherapy Treatment Plans: A Novel Approach|| |
Gaganpreet Singh, Arun S. Oinam1, Vivek Kumar
Centre for Medical Physics, Panjab University, 1Department of Radiation Therapy, Regional Cancer Center, Postgraduate Institute of Medical Education and Research, Chandigarh, India. E-mail: [email protected]
Introduction: Radiotherapy treatment plans are evaluated by different methods i.e. slice by slice evaluation, DVH based evaluation and plan metrics methods etc. All the methods involves the three dimensional (3D) dose matrices which contain voxels having different information tagged with them. In this study, information of the location of voxel of different 3D dataset of computed tomography (CT) images, RT structures set and RT dose are used for the evaluation of the plan and further this approach is used for the radiobiological evaluation of the treatment plan.
Objectives: Development and implementation of the voxel based method for evaluation of the radiotherapy treatment plan incorporating radiobiological parameters.
Materials and Methods: In this study, Digital imaging and communication in medicine in radiotherapy (DICOM-RT) standard files are used to extract the CT Images, RT Structures set (RTSS), RT Dose and RT Plan information. Patient's plan is exported from the Eclipse TPS of Varian Medical System, Palo Alto, USA and imported into the in-house build program which is written in Matlab® software (Mathworks Inc.). In this program, Dose Cube and Structure set Cube are constructed by interpolating with respect to CT Cube in such a way that resultant structure Cube and Dose Cube are of the same dimensions. Voxels mapping have been used to reconstruct dose volume histograms (DVH) of different organs and target volumes of the plan. Further, Voxel based biologically equivalent dose (BED) and equivalent dose of 2 Gy (EQD2) have been calculated using the standard mathematical formulism of BED and EQD2 doses. BED and EQD2 based colorwash and isoeffect dose curves have also been constructed incorporating linear quadratic (LQ) model parameter alpha by beta ratios of assigned values of different normal tissues and tumor.
Results and Discussion: This study shows the reconstruction of the patient plan on an independently developed program based on DICOMRT standard files. It provides the visualization of contours and dose colorwash over the CT images same as that of represented by Eclipse TPS. Reconstructed DVH depends on the grid size of the CT-, Structure set-and Dose – Cube. Change in the voxel size effect the DVH. In this study, DVH and colorwash construct from the voxel based approach compared with the Eclipse TPS. Results have shown a good agreement between the DVH and colorwash obtained from the TPS and from the developed program. BED and EQD2 doses at different voxel locations have been calculated and used for the display of radiobiologically equivalent dose colorwash and isoeffective dose curves for the evaluation of the plans which provides better understanding of the radiobiological equivalent doses.
Conclusion: This study shows voxel based novel approach for evaluation of patient treatment plan which can be further extended for the radiobiological evaluation of treatment plans which incorporate radiosensitivity of different organs and target organs. This study can be further extended for the evaluation of complex radiotherapy treatment plan of the patients.
| AB-10: 3D Silicon Microdosimetry for Boron Neutron Capture Therapy: A simulation study|| |
N. Hu, R. Uchida, L. T. Tran1, A. Rosenfeld1, Y. Sakurai2
Division of Nuclear Engineering, Graduate School of Engineering, Kyoto University, Kyoto, 2Kyoto University Research Reactor Institute, Osaka, Japan,1Centre for Medical Radiation Physics, University of Wollongong, Wollongong, Australia. E-mail: [email protected]
Introduction: Boron neutron capture therapy (BNCT) is an emerging radiotherapy modality using a neutron beam collectively with boron-10 containing pharmaceutical to treat patients with cancer. In contrast to conventional radiotherapy, the types of radiation present in BNCT consists of many distinct radiation components, each having a different biological weighting factor. Microdosimetry is an effective dosimetry technique in a mixed radiation environment. Using this technique, it is possible to derive the relative contributions of different radiation modalities. This paper presents the feasibility study of a novel 3D mesa bridge microdosimeter in BNCT, developed by University of Wollongong (UOW).
Materials and Methods: This bridge microdosimeter is comprised of an array of 4248 individual silicon cells fabricated on a 10 μm thick n-type silicon-on-insulator substrate. The performance of the microdosimeter was studied using Monte Carlo simulation. Different boron converter and silicon-on-insulator substrate thickness was modelled and the energy deposition within the detector was simulated using the Particle and Heavy Ions Transport Code System (PHITS). The T-Deposit tally in PHITS was used to calculate the energy deposited per event inside the sensitive volume of the bridge microdosimeter. The lineal energy was calculated by dividing the deposited energy per event by the average chord length of the detector. The clinical BNCT field at Kyoto University Reactor (KUR) using both thermal and epithermal irradiation modes were used in this study.
Results: A thinner boron converter resulted in more reaction particles reaching the sensitive volume of the detector. Approximately double the number of particles reached the detector for a 1 μm thick boron converter as compared to a 0.5 mm thick boron converter. For the alpha particle, a peak at 120 keV/μm was observed with both the 0.5 mm and 1 μm boron converter and a peak at 200 keV/μm was observed with no boron converter. A peak in the no boron converter spectrum arises from the boron p+ dopant in the device. Figure 1 shows the microdosimetric spectrum obtained from the bridge microdosimeter for the KUR epithermal beam.
|Figure 1: Microdosimetric spectrum of the Kyoto University Reactor epithermal beam generated by Particle and Heavy Ions Transport Code System|
Click here to view
Discussion: The range of the alpha particles produced during BNCT is approximately 5 μm. For a 10 μm thick detector, the alpha particle will come to a full stop inside the sensitive volume, resulting in an inaccurate lineal energy deposition. To account for this, a 2 μm thick detector was simulated and tested. Results showed the lineal energy deposition was improved with the use of the 2 μm thick sensitive volume detector. Simulation results showed that the thermal irradiation mode seemed the most appropriate mode to perform the measurements, because of the high thermal neutron flux, which resulted in high production of reaction particles, and lower epithermal and fast neutron flux, which resulted in lower recoil silicon particles produced.
Conclusion: The microdosimetric spectrum showed each particle can be separated by calculating the lineal energy.The simulation results show that this microdosimeter can be utilised as an effective tool for dosimetry in BNCT field and experimental validation will follow once KUR is operational.
| AB-11: Peripheral Dose Measurements with FFF Beams from Linear Accelerator and Tomotherapy for Sbrt of Ca Prostate - Phantom Study|| |
Pragya Shree, R. A. Kinhikar, A. Dheera, C. M. Tambe, D. D. Deshpande
Department of Medical Physics, Tata Memorial Hospital, Mumbai, Maharashtra, India. E-mail: [email protected]
Purpose: To quantify the relative peripheral doses (PD) for the unflattened beams for advanced treatment techniques with SBRT for Ca Prostate cases delivered with linear accelerator and tomotherapy machines.
Materials and Methods: Varian Truebeam linear accelerator was used to provide 6 and 10MV unflattened photon beams and Accuray Tomotherapy Hi-ArtII was used to provide 6MV unflattened photon beam. SBRT (for Ca prostate) treatment plans were generated for 6FFF and 10FFF photon beams using Eclipse and Tomo TPS (five patients for each). All treatment plans were delivered to the relevant anatomical region of a body phantom (30×90×20 cm3, width × length × depth, stack of solid water slabs). Dosimetric measurements were performed with TLD-100 (LiF:MgTi) which were placed on surfaceat 5, 10, 15, 20 and 25cm from the field edge and at depth (5 and 10 cm) as well. The TLDs were read using Rexon reader and the readings at the respective distance and depth were recorded. The Peripheral Doses were normalized to the prescribed dose (14Gy in 2 fractions). The measured values of peripheral doses for Tomotherapy 6FFF photon beams was compared with the Truebeam 6FFF photon beam and the values for Truebeam 10FFF and 6FFF photon beams were also compared with each other.
Results: The measured mean peripheral doses (PDs) at 20cm beyond the field edge at surface (depth=0) with linac 10FFF and 6FFF was 0.32%±0.16 (ranging from 0.16% to 0.52%) and 0.30%±0.14 (0.17% to 0.49%) respectively and at a distance of 25 cm, 0.23%±0.1 (0.11% to 0.36%) and 0.22%±0.1 (0.12% to 0.35%) respectively. And at the depth of 10 cm, the measured mean PDs at 20 cm beyond the field edge with linac 10FFF and 6FFF was 0.23%±0.08 (0.13% to 0.33%) and 0.25%±0.1 (0.14% to 0.38%) respectively and at a distance of 25 cm, 0.17%±0.07 (0.1% to 0.27%) and 0.17%±0.07 (0.11% to 0.27%) respectively. The measured mean PDs at 20 cm beyond the field edge at surface with tomotherapy 6FFF was 0.38%±0.06 (0.30% to 0.46%) and at a distance of 25 cm, 0.25%±0.03 (0.22% to 0.29%). And at the depth of 10 cm, the measured mean PDs at 20 cm beyond the field edge with tomotherapy 6FFF was 0.33%±0.13 (0.22% to 0.56%) and at a distance of 25 cm, 0.22%±0.07 (0.17% to 0.34%).
Conclusions: The PDs for both the 6 FFF and 10 FFF photon beams were found to decrease with increasing distance from the radiation field edge. The relative deviation between PDs of 6FFF and 10FFF was found to decrease with increasing distance from the field edge and this deviation became almost negligible beyond 20 cm distance from the field edge. The measured PD from Tomotherapy 6FFF was found higher than Linac 6FFF by the maximum factor of 2.26.
| AB-12: GPU-ACCELERATED 4D Image Reconstruction Using ON-BOARD KV CONE-BEAM CT|| |
Sunghoon Choi, Chang-Woo Seo, Hee-Joung Kim
Department of Radiological Science, Yonsei University, Wonju, Korea. E-mail: [email protected]
Introduction: On-board kV cone-beam CT (CBCT) imaging has recently become available based on a flat-panel detector mounted in a linear accelerator (LINAC) for radiotherapy guidance. Respiration induces motion artifacts which could be reduced by retrospectively sorting projection images and reconstructing multiple three-dimensional (3D) CT dataset at different respiratory phases. Unfortunately, multiple breathing phases over a full gantry rotation involve huge number of projections to reduce aliasing artifact at each phase. Therefore, a real-time patient monitoring in treatment planning using 4D CBCT is quite challenging because it demands high computation time during image reconstruction.
Objectives: The main focus of this work is to present a GPU-accelerated 4D CBCT image reconstruction technique in order to diagnose a patient with a fast speed, thereby provides a real-time patient monitoring in treatment planning.
Materials and Methods: For a respiratory-correlated 4D CBCT reconstruction, the authors acquired total 1,262 projections of a moving phantom (QUASAR Programmable Respiratory Motion Phantom Figure 1, Modus Medical Device Inc., London, ON) using a flat-panel based on-board imager (Varian Medical Systems, Palo Alto, CA). The breathing signal for determining respiratory correlation was directly measured from a real-time position management (RPM) respiratory gating system. Projection data of the phantom moving with simulated sinusoidal motion were acquired and reconstructed into 3D and 4D CBCT datasets. The number of retrospectively rearranged projection images in four respiratory phase bins was 445, 235, 236, and 413 for peak-exhale (PE), mid-exhale (ME), mid-inhale (MI), and peak-inhale (PI) state, respectively, while total of 335 projection images with equally spaced angle step around full 360 gantry rotation was involved in 3D reconstruction. All image reconstructions were conducted by using the filtered back-projection scheme. For image quality analysis, the reconstructed axial images of 3D and 4D data were visually compared with each other. The authors also verified the GPU-accelerated computing power over CPU programming by checking the elapsed time required for reconstructing both 3D and 4D volumetric data.
|Figure 1: (a) The programmable motion phantom used in this study and (b) its cylindrical inserts with sphere target|
Click here to view
Results and Discussion: The reconstructed axial images in each respiratory phase of 4D data provided little geometric error of moving structure (e.g., little blurring artifact) compared to 3D data Figure 2. However, aliasing artifacts represented by the streaking lines were substantially highlighted in 4D images due to small angular coverage over full gantry rotation. The reconstruction times for producing 4D dataset were 2.69, 0.91, 0.94, and 1.56 sec (total 6.14 sec) for PE, ME, MI, and PI phase bin, respectively, on GPU programming, whereas the elapsed times on CPU programming were 287.70, 155.03, 147.08, and 256.69 sec (total 846.53 sec). This could demonstrate that GPU computing time shows an enhanced speed of ~137 times than CPU programming, which could match-up a clinically feasible time. In conclusion, the authors observed reasonable outcomes of GPU-accelerated 4D CBCT image reconstruction scheme for real-time patient monitoring through GPU-acceleration in the current work.
|Figure 2: The reconstructed axial images of (a) conventional 3D dataset and (b) peak-exhale phase bin|
Click here to view
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8], [Figure 9], [Figure 10]
| Article Access Statistics|
| Viewed||330 |
| Printed||2 |
| Emailed||0 |
| PDF Downloaded||66 |
| Comments ||[Add] |