• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.1159-1166
• /
• 2009

The use of cone-beam computed tomography(CBCT) has been proposed for guiding the delivery of radiation therapy. A kilovoltage imaging system capable of radiography, fluoroscopy, and cone-beam computed tomography(CT) has been integrated with a medical linear accelerator. A standard clinical linear accelerator, operating in arc therapy mode, and an amorphous-silicon (a-Si) with an on-board electronic portal imager can be used to treat palliative patient and verify the patient's position prior to treatment. On-board CBCT images are used to generate patient geometric models to assist patient setup. The image data can also, potentially, be used for dose reconstruction in combination with the fluence maps from treatment plan. In this study, the accuracy of Hounsfield Units of CBCT images as well as the accuracy of dose calculations based on CBCT images of a phantom and compared the results with those of using CT simulator images. Phantom and patient studies were carried out to evaluate the achievable accuracy in using CBCT and CT stimulator for dose calculation. Relative electron density as a function of HU was obtained for both planning CT stimulator and CBCT using a Catphan-600 (The Phantom Laboratory, USA) calibration phantom. A clinical treatment planning system was employed for CT stimulator and CBCT based dose calculations and subsequent comparisons. The dosimetric consequence as the result of HU variation in CBCT was evaluated by comparing MU/cCy. The differences were about 2.7% (3-4MU/100cGy) in phantom and 2.5% (1-3MU/100cGy) in patients. The difference in HU values in Catphan was small. However, the magnitude of scatter and artifacts in CBCT images are affected by limitation of detector's FOV and patient's involuntary motions. CBCT images included scatters and artifacts due to In addition to guide the patient setup process, CBCT data acquired prior to the treatment be used to recalculate or verify the treatment plan based on the patient anatomy of the treatment area. And the CBCT has potential to become a very useful tool for on-line ART.)

• The Journal of Korean Society for Radiation Therapy
• /
• v.17 no.2
• /
• pp.125-131
• /
• 2005

Purpose : For stereotactic radiosurgery (SRS) of a tumor in the region whose movement due to respiration is significant, like Lung lower lobe, the gated therapy, which delivers radiation dose to the selected respiratory phases when tumor motion is small, was performed using the Respiratory gating system and its clinical effectiveness was evaluated. Materials and Methods : For two SRS patients with a tumor in Lung lower lobe, a marker block (infrared reflector) was attached on the abdomen. While patient' respiratory cycle was monitored with Real-time Position Management (RPM, Varian, USA), 4D CT was performed (10 phases per a cycle). Phases in which tumor motion did not change rapidly were decided as treatment phases. The treatment volume was contoured on the CT images for selected treatment phases using maximum intensity projection (MIP) method. In order to verify setup reproducibility and positional variation, 4D CT was repeated. Results : Gross tumor volume (GTV) showed maximum movement in superior-inferior direction. For patient #1, motion of GTV was reduced to 2.6 mm in treatment phases ($30{\sim}60%$), while that was 9.4 mm in full phases ($0{\sim}90%$) and for patient #2, it was reduced to 2.3 mm in treatment phases ($30{\sim}70%$), while it was 11.7 mm in full phases ($0{\sim}90%$). When comparing two sets of CT images, setup errors in all the directions were within 3 mm. Conclusion : Since tumor motion was reduced less than 5 mm, the Respiratory gating system for SRS of Lung lower lobe is useful.

• The Journal of Korean Society for Radiation Therapy
• /
• v.16 no.1
• /
• pp.51-56
• /
• 2004

The aim of this study is to evaluate the effect of skin dose and PDD by using wounds protecting gauzes or Vaselinespread gauzes. And it was studied that the possibility to substitute custom bolus into gauzes. 4MV photon (CL600C, varian, US), Polystyrene Phantom (30(W) X30(L) X 30(H)) with Markus chamber(PTW, US) were used for dose measurement. This study was distinguished natural gauzes and spread over Vaseline gauzes. We gave variety to the gauze thickness at 5, 10 and 15 sheets respectively. For comparison between using bolus and not that, we had used 1.0 cm thickness bolus so that analyzed surface dose and PDD at the same conditions above mentioned. When maximum point was defined as reference point, surface dose was measured as $35\%$ in open beam. When the gauzes were attached to surface as 5, 10 and 15 sheets, surface dose were increased as 69, 80 and $91\%$ respectively according to thickness of gauzes. When spread over Vaseline gauzes were attached to surface as 5, 10 and 15 sheets, surface dose were increased respectively as 98, 100 and $98\%$ according to thickness of gauzes. Also when 0.5 cm bolus and 5 sheets gauzes were composed, surface dose was measured as $98\%$. The gauzes that were attached to skin surface in radiation therapy had been scattering material and contributed increasing surface dose without variation of percentage depth dose. However, if we want to delivery much dose to skin surface then we have to apply many sheets of gauzes to skin surface. Although we get easy that result by bolus or spread over Vaseline gauzes, we have to revise percentage depth dose at calculation. Therefore, if we find pertinent conditions based on measured data that are considered skin dose and patient setup efficiency, to replace custom bolus with gauzes will be helpful to efficient treatment.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.9 no.4
• /
• pp.34-44
• /
• 2006

Simulation models allow researchers to model large hydrological catchment for comprehensive management of the water resources and explication of the diffuse pollution processes, such as land-use changes by development plan of the region. Recently, there have been reported many researches that examine water body quality using Geographic Information System (GIS) and dynamic watershed models such as AGNPS, HSPF, SWAT that necessitate handling large amounts of data. The aim of this study is to develop a watershed based water quality estimation system for the impact assessment on stream water quality. KBASIN-HSPF, proposed in this study, provides easy data compiling for HSPF by facilitating the setup and simulation process. It also assists the spatial interpretation of point and non-point pollutant information and thiessen rainfall creation and pre and post processing for large environmental data An integration methodology of GIS and water quality model for the preprocessing geo-morphologic data was designed by coupling the data model KBASIN-HSPF interface comprises four modules: registration and modification of basic environmental information, watershed delineation generator, watershed geo-morphologic index calculator and model input file processor. KBASIN-HSPF was applied to simulate the water quality impact by variation of subbasin pollution discharge structure.

• Journal of the Korea Concrete Institute
• /
• v.22 no.4
• /
• pp.535-546
• /
• 2010

Recently, the probability of collision accident between vehicles or vessels and infrastructures are increasing at alarming rate. Particularly, collision impact load can be detrimental to sub-structures such as piers and columns. The damaged pier from an impact load of a vehicle or a vessel can lead to member damages, which make the member more vulnerable to impact load due to other accidents which. In extreme case, may cause structural collapse. Therefore, in this study, the vehicle impact load on concrete compression member was considered to assess the quantitative design resistance capacity to improve, the existing design method and to setup the new damage assessment method. The case study was carried out using the LS-DYNA, an explicit finite element analysis program. The parameters for the case study were cross-section variation of pier, impact load angle, permanent axial load and axial load ratio, concrete strength, longitudinal and lateral rebar ratios, and slenderness ratio. Using the analysis results, the performance based resistance capacity evaluation method for impact load using satisfaction curve was developed using Bayesian probabilistic method, which can be applied to reinforced concrete column design for impact loads.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.27 no.2
• /
• pp.105-114
• /
• 2007

Two-dimensional imaging of $V_s$ profile becomes more important in Korea because of the large horizontal variation of soil stiffness. To obtain a shear-wave velocity profile in geotechnical practice, various seismic nondestructive investigation methods are being frequently used. In this study, harmonic wavelet analysis of wave (HWAW) method is applied to the determination of $V_s$ profile to overcome some of weaknesses in the existing surface wave methods. HWAW method which is based on time-frequency analysis using harmonic wavelet transform has been developed to determine phase and group velocities of waves. Field testing of this method is relatively simple and fast because one experimental setup which consists of one pair of receivers is needed to determine $V_s$ profile of site. The proposed method uses the signal portion of the maximum local signal/noise ratio to evaluate the phase velocity to minimize the effects of noise, and uses single array inversion which considers receiver locations. Field tests were performed in 2 sites in order to evaluate accuracy of test method and estimate the applicability of 2-D imaging by HWAW method. Through field applications and comparison with other test results, the good accuracy and applicability of the proposed method were verified.

• Ocean and Polar Research
• /
• v.25 no.spc3
• /
• pp.361-371
• /
• 2003

The effect of bottom friction on the steady wind-driven circulation in the Yellow Sea and the East China Sea (YSECS) has been studied using a two-dimensional numerical model with and without tidal forcing. Upwind flow experiment in YSECS has also been carried out with a schematic time variation in the wind field. The surface water setup and circulation pattern due to steady wind forcing are found to be very sensitive to the bottom friction. When the effects of tidal currents are neglected, the overall current velocities are overestimated and eddies of various sizes appear, upwind flow is formed within the deep trough of the Yellow Sea, forming a part of the topographic gyre on the side of Korea. When tidal forcing is taken into account, the wind-induced surface elevations are smoothed out due to the strong tide-induced bottom friction, which is aligned almost normal to the wind stresses; weak upwind flow is farmed in the deep trough of the Yellow Sea, west and south of Jeju. Calculation with wind forcing only through a parameterized linear bottom friction produces almost same results from the calculation with $M_2$ tidal forcing and wind forcing using a quadratic bottom friction, supporting Hunter (1975)'s linearization of bottom friction which includes the effect of tidal current, can be applied to the simulation of wind-driven circulation in YSECS. The results show that steady wind forcing is not a dominant factor to the winter-time upwind flow in YSECS. Upwind flow experiment which considers the relaxation of pressure gradient (Huesh et al. 1986) shows that 1) a downwind flow is dominant over the whole YSECS when the northerly wind reaches a maximum speed; 2) a trend of upwind flow near the trough is found during relaxation when the wind abates; 3) a northward flow dominates over the YSECS after the wind stops. The results also show that the upwind flow in the trough of Yellow Sea is forced by a wind-induced longitudinal surface elevation gradient.

• International Journal of Highway Engineering
• /
• v.8 no.4 s.30
• /
• pp.125-133
• /
• 2006

In this study, the Korean Industrial Standard(KS) penetration grade system was examined to evaluate its statistical validity. It was found from this limited study that the system has no gap between each grade, the binder falling out of limit trespasses into other binder grade. Therefore, many products can have similar property level with an adjacent grade even though the mean value is within the specification limit. It was also found out that the equal range setup of each grade, such as 20 for 40-60, 60-80, 80-100, has no statistical foundation. Since KS penetration grade is defined without gap between each grade, the producer must maintain the coefficient of variation below the level by the ASTM system to satisfy the normal quality control limit of products. Due to its lack of a gap between grades, higher percentage of products will be duplicated even if the mean is at the median of the specification limits due to material's quality uncertainty. Especially if the mean moves toward the upper limit, a half or more of the binder grade will be overlapped with the upper binder grade. Therefore, KS penetration specification should be immediately modified by statistically valid methodologies.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.23 no.4
• /
• pp.313-325
• /
• 2011

Finite element grid system using h-refinement on the Yellow Sea was constructed based on previous study (Suh, 1999b) from 14 K to 210 K and special attention was concentrated on refining the coastal zone. In grid generation, depth change between adjacent points and non-dimensional tidal wave length ratio were considered. As a result approximately a quarter of the total nodes are located nearby 5 m of shallow area. Accurate bathymetry data using 30's and ETOPO1 with open boundary conditions of 8 major tidal constituents extracted automatically from FES2004 have been applied. In tidal simulation a 3-dimensional nonlinear harmonic model was setup and tidal amplification due to changes in vertical turbulent and bottom friction were simulated. In this study not only 8 major tidal constituents but also nonlinear shallow tides $M_4,$, $MS_4$ and long period $M_f,$, $M_{sf}$ were reproduced. It is found that implication of spatial variation of friction coefficient plays a very important role in reproduction of astronomical and shallow tides which are computed by iterative computation of nonlinear terms. Also it should be considered differently with respect to tidal periods. To understand the distribution of tidal asymmetry, amplitude ratio of $M_4/M_2$ and phase differences $2g(M_2)-g(M_4)$ were calculated. Tidal distortion ratio marks up to 0.2 on the west coast showing shallow coastal characteristics and somewhat wide range of ebb-dominances in front of Mokpo area are reproduced.

• Journal of Intelligence and Information Systems
• /
• v.23 no.2
• /
• pp.1-17
• /
• 2017

The deep learning framework is software designed to help develop deep learning models. Some of its important functions include "automatic differentiation" and "utilization of GPU". The list of popular deep learning framework includes Caffe (BVLC) and Theano (University of Montreal). And recently, Microsoft's deep learning framework, Microsoft Cognitive Toolkit, was released as open-source license, following Google's Tensorflow a year earlier. The early deep learning frameworks have been developed mainly for research at universities. Beginning with the inception of Tensorflow, however, it seems that companies such as Microsoft and Facebook have started to join the competition of framework development. Given the trend, Google and other companies are expected to continue investing in the deep learning framework to bring forward the initiative in the artificial intelligence business. From this point of view, we think it is a good time to compare some of deep learning frameworks. So we compare three deep learning frameworks which can be used as a Python library. Those are Google's Tensorflow, Microsoft's CNTK, and Theano which is sort of a predecessor of the preceding two. The most common and important function of deep learning frameworks is the ability to perform automatic differentiation. Basically all the mathematical expressions of deep learning models can be represented as computational graphs, which consist of nodes and edges. Partial derivatives on each edge of a computational graph can then be obtained. With the partial derivatives, we can let software compute differentiation of any node with respect to any variable by utilizing chain rule of Calculus. First of all, the convenience of coding is in the order of CNTK, Tensorflow, and Theano. The criterion is simply based on the lengths of the codes and the learning curve and the ease of coding are not the main concern. According to the criteria, Theano was the most difficult to implement with, and CNTK and Tensorflow were somewhat easier. With Tensorflow, we need to define weight variables and biases explicitly. The reason that CNTK and Tensorflow are easier to implement with is that those frameworks provide us with more abstraction than Theano. We, however, need to mention that low-level coding is not always bad. It gives us flexibility of coding. With the low-level coding such as in Theano, we can implement and test any new deep learning models or any new search methods that we can think of. The assessment of the execution speed of each framework is that there is not meaningful difference. According to the experiment, execution speeds of Theano and Tensorflow are very similar, although the experiment was limited to a CNN model. In the case of CNTK, the experimental environment was not maintained as the same. The code written in CNTK has to be run in PC environment without GPU where codes execute as much as 50 times slower than with GPU. But we concluded that the difference of execution speed was within the range of variation caused by the different hardware setup. In this study, we compared three types of deep learning framework: Theano, Tensorflow, and CNTK. According to Wikipedia, there are 12 available deep learning frameworks. And 15 different attributes differentiate each framework. Some of the important attributes would include interface language (Python, C ++, Java, etc.) and the availability of libraries on various deep learning models such as CNN, RNN, DBN, and etc. And if a user implements a large scale deep learning model, it will also be important to support multiple GPU or multiple servers. Also, if you are learning the deep learning model, it would also be important if there are enough examples and references.