• Progress in Medical Physics
• /
• 제16권2호
• /
• pp.82-88
• /
• 2005

In this study, the physical compensator made with the high density material, Cerrobend, and the electronic compensator realized by the movement of a dynamic multileaf collimator were analyzed in order to verify the properness of a design function in the commercial RTP (radiation treatment planning) system, Eclipse. The CT images of a phantom composed of the regions of five different thickness were acquired and the proper compensator which can make homogeneous dose distribution at the reference depth was designed in the RTP. The frame for the casting of Cerrobend compensator was made with a computerized automatic styrofoam cutting device and the Millennium MLC-120 was used for the electronic compensator. All the dose values and isodose distributions were measured with a radiographic EDR2 film. The deviation of a dose distribution was $\pm0.99 cGy\;and\;\pm1.82cGy$ in each case of a Cerrobend compensator and a electronic compensator compared with a $\pm13.93 cGy$ deviation in an open beam condition. Which showed the proper function of the designed compensators in the view point of a homogeneous dose distribution. When the absolute dose value was analyzed, the Cerrobend compensator showed a $+3.83\%$ error and the electronic compensator showed a $-4.37\%$ error in comparison with a dose value which was calculated in the RTP. These errors can be admtted as an reasonable results that approve the accuracy of the compensator design in the RTP considering the error in the process of the manufacturing of the Cerrobend compensator and the limitation of a film in the absolute dosimetry.

• Progress in Medical Physics
• /
• 제17권4호
• /
• pp.212-223
• /
• 2006

Positional accuracy of the on-board imager (OBI) isocenter with gantry rotation was presented in this paper. Three different type of automatic evaluation methods of discrepancies between therapeutic and OBI isocenter using digital image processing techniques as well as a procedure stated in the customer acceptance procedure (CAP) were applied to check OBI isocenter migration trends. Two kinds of kV x-ray image set obtained at OBI source angle of $0^{\circ},\;90^{\circ},\;180^{\circ},\;270^{\circ}$ and every $10^{\circ}$ and raw projection data for cone-beam CT reconstruction were used for each evaluation method. Efficiencies of the methods were also estimated. If a user needs to obtain an isocenter variation map with full gantry rotation, a method taking OBI image for every $10^{\circ}$ and fitting with 5th order polynomial was appropriate. However for a mere quality assurance (QA) purpose of OBI isocenter accuracy, it was adequate to use only four OBI Images taken at the OBI source angle of $0^{\circ},\;90^{\circ},\;180^{\circ}\;and\;270^{\circ}$. Maximal discrepancy was 0.44 mm which was observed between the OBI source angle of $90^{\circ}\;and\;180^{\circ}$ OBI isocenter accuracy was maintained below 0.5 mm for a year. Proposed QA program may be helpful to Implement a reasonable routine QA of the OBI isocenter accuracy without great efforts.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• 제21권12호
• /
• pp.715-732
• /
• 2009

This article reviews the papers published in the Korean Journal of Air-Conditioning and Refrigeration Engineering during 2008. It is intended to understand the status of current research in the areas of heating, cooling, ventilation, sanitation, and indoor environments of buildings and plant facilities. Conclusions are as follows. (1) Research trends in thermal and fluid engineering have been surveyed in the categories of general fluid flow, fluid machinery and piping, new and renewable energy, and fire. Well-developed CFD technologies were widely applied in developing facilities and their systems. New research topics include fire, fuel cell, and solar energy. Research was mainly focused on flow distribution and optimization in the fields of fluid machinery and piping. Topics related to the development of fans and compressors had been popular, but were no longer investigated widely. Research papers on micro heat exchangers using nanofluids and micro pumps were also not presented during this period. There were some studies on thermal reliability and performance in the fields of new and renewable energy. Numerical simulations of smoke ventilation and the spread of fire were the main topics in the field of fire. (2) Research works on heat transfer presented in 2008 have been reviewed in the categories of heat transfer characteristics, industrial heat exchangers, and ground heat exchangers. Research on heat transfer characteristics included thermal transport in cryogenic vessels, dish solar collectors, radiative thermal reflectors, variable conductance heat pipes, and flow condensation and evaporation of refrigerants. In the area of industrial heat exchangers, examined are research on micro-channel plate heat exchangers, liquid cooled cold plates, fin-tube heat exchangers, and frost behavior of heat exchanger fins. Measurements on ground thermal conductivity and on the thermal diffusion characteristics of ground heat exchangers were reported. (3) In the field of refrigeration, many studies were presented on simultaneous heating and cooling heat pump systems. Switching between various operation modes and optimizing the refrigerant charge were considered in this research. Studies of heat pump systems using unutilized energy sources such as sewage water and river water were reported. Evaporative cooling was studied both theoretically and experimentally as a potential alternative to the conventional methods. (4) Research papers on building facilities have been reviewed and divided into studies on heat and cold sources, air conditioning and air cleaning, ventilation, automatic control of heat sources with piping systems, and sound reduction in hydraulic turbine dynamo rooms. In particular, considered were efficient and effective uses of energy resulting in reduced environmental pollution and operating costs. (5) In the field of building environments, many studies focused on health and comfort. Ventilation. system performance was considered to be important in improving indoor air conditions. Due to high oil prices, various tests were planned to examine building energy consumption and to cut life cycle costs.

• Journal of Intelligence and Information Systems
• /
• 제22권2호
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.