• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.10
• /
• pp.303-308
• /
• 2020

The ReLU(Rectified Linear Unit) function has been dominantly used as a standard activation function in most deep artificial neural network models since it was proposed. Later, Leaky ReLU, Swish, and Mish activation functions were presented to replace ReLU, which showed improved performance over existing ReLU function in image classification task. Therefore, we recognized the need to experiment with whether performance improvements could be achieved by replacing the RELU with other activation functions in the super resolution task. In this paper, the performance was compared by changing the activation functions in EDSR model, which showed stable performance in the super resolution task. As a result, in experiments conducted with changing the activation function of EDSR, when the resolution was converted to double, the existing activation function, ReLU, showed similar or higher performance than the other activation functions used in the experiment. When the resolution was converted to four times, Leaky ReLU and Swish function showed slightly improved performance over ReLU. PSNR and SSIM, which can quantitatively evaluate the quality of images, were able to identify average performance improvements of 0.06%, 0.05% when using Leaky ReLU, and average performance improvements of 0.06% and 0.03% when using Swish. When the resolution is converted to eight times, the Mish function shows a slight average performance improvement over the ReLU. Using Mish, PSNR and SSIM were able to identify an average of 0.06% and 0.02% performance improvement over the RELU. In conclusion, Leaky ReLU and Swish showed improved performance compared to ReLU for super resolution that converts resolution four times and Mish showed improved performance compared to ReLU for super resolution that converts resolution eight times. In future study, we should conduct comparative experiments to replace activation functions with Leaky ReLU, Swish and Mish to improve performance in other super resolution models.

• Progress in Medical Physics
• /
• v.12 no.2
• /
• pp.113-124
• /
• 2001

For veterinary imaging diagnosis, we obtained MR images of the canine brain, spine, kidney and pelvis from 3T MRI system which was equipped with the world first 3T active shield magnet. Spin echo (SE) and fast Spin Echo (FSE) images were obtained from the canine brain, spine, kidney and pelvis of normal and sick dogs using a homemade birdcage and transverse electromagnetic (TEM) resonators operating in quadrature and tuned to 128 MHz. In addition, we employed a homemade saddle shaped RF coil. Typical common acquisition parameters were as follows: matrix=512$\times$512, field of view (FOV)=20cm, slice thickness=3 w, number of excitations (NEX)=1. For T1-weighted MR images, we used TR=500 ms, TE=10 or 17.4 ms. For T2-weighted MR images, we used TR=4000 ms, TE=108 ms. Signal to noise ratio (SNR) of 3T system was measured 2.7 times greater than that of prevalent 1.57 system. The high resolution images acquired in this study represent more than a 4-fold increase in in-plane resolution relative to conventional images obtained with a 20 cm field of view and a 5 mm slice thickness. MR images obtained from 3T system revealed numerous small venous structures throughout the image plane and provided reasonable delineation between gray and white matter The present results demonstrate that the MR images from 3T system could provide better diagnostic quality of resolution and sensitivity than those of 1.5T system. The elevated SNR observed in the 3T high field magnetic resonance imaging can be utilized to acquire images with a level of resolution approaching the microscopic structural level under in vivo conditions. These images represent a significant advance in our ability to examine small anatomical features with noninvasive imaging methods. Moreover, MRI technique could begin to apply for veterinary medicine in Korea.

• Progress in Medical Physics
• /
• v.14 no.2
• /
• pp.131-140
• /
• 2003

As the importance of accuracy in measurings of 3-D anatomical structures continues to be stressed, an objective and quantitative of assessing image quality and accuracy of 3-D volume-rendered images is required. The purpose of this study was to evaluate the quantitative accuracy of 3-D rendered images obtained with MDCT, scanned at various scanning parameters (scan modes, slice thicknesses and reconstruction slice thickness). Twelve clinically significant points that play an important role for the craniofacial bone in plastic surgery and dentistry were marked on the surface of a dry human skull. The direct distances between the reference points were defined as gold standards to assess the measuring errors of 3-D images. Then, we scanned the specimen with acquisition parameters of 300 mA, In kVp, and 1.0 sec scan time in axial and helical scan modes (pitch 3:1 and 6:1) at 1,25 mm, 2.50 mm, 3.75 mm and 5.00 mm slice thicknesses. We performed 3-D visualizations and distance measurements with volumetric analysis software and statistically evaluated the quantitative accuracy of distance measurements. The accuracy of distance measurements on the 3-D images acquired with 1.25, 2.50, 3,75 and 5.00 mm slice thickness were 48%, 33%, 23%, 14%, respectively, and those of the reconstructed 1.25 mm were 53%, 41%, 43%, 36% respectively. Meanwhile, there were insignificant statistical differences (P-value<0.05) in the accuracy of the distance measurements of 3-D images reconstructed with 1.25 mm thickness. In conclusion, slice thickness, rather than scan mode, influenced the quantitative accuracy of distance measurements in 3-D rendered images with MDCT. The quantitative analysis of distance measurements may be a useful tool for evaluating the accuracy of 3-D rendered images used in diagnosis, surgical planning, and radiotherapeutic treatment.

• Proceedings of the Korea Society for Industrial Systems Conference
• /
• 2002.06a
• /
• pp.148-155
• /
• 2002

Multithreaded models improve the efficiency of parallel systems by combining inner parallelism, asynchronous data availability and the locality of von Neumann model. This model executes thread code which is generated by compiler and of which quality is given by the method of generation. But multithreaded models have the demerit that execution model is restricted to a specific platform. On the contrary, Java has the platform independency, so if we can translate from threads code to Java bytecode, we can use the advantages of multithreaded models in many platforms. Java executes Java bytecode which is intermediate language format for Java virtual machine. Java bytecode plays a role of an intermediate language in translator and Java virtual machine work as back-end in translator. But, Java bytecode which is translated from multithreaded models have the demerit that it is not secure. This paper, multhithread code whose feature of platform independent can execute in java virtual machine. We design and implement translator which translate from thread code of multithreaded code to Java bytecode and which check secure problems from Java bytecode.

• Progress in Medical Physics
• /
• v.15 no.4
• /
• pp.202-209
• /
• 2004

Flat panel based digital radiography (DR) systems have recently become useful and important in the field of diagnostic radiology. For DRs with amorphous silicon photosensors, CsI(TI) is normally used as the scintillator, which produces visible light corresponding to the absorbed radiation energy. The visible light photons are converted into electric signal in the amorphous silicon photodiodes which constitute a two dimensional array. In order to produce good quality images, detailed behaviors of DR detectors to radiation must be studied. The relationship between air exposure and the DR outputs has been investigated in many studies. But this relationship was investigated under the condition of the fixed tube voltage. In this study, we investigated the relationship between the DR outputs and X-ray in terms of the absorbed energy in the detector rather than the air exposure using SPEC-l8, an X-ray energy spectrum model. Measured exposure was compared with calculated exposure for obtaining the inherent filtration that is a important input variable of SPEC-l8. The absorbed energy in the detector was calculated using algorithm of calculating the absorbed energy in the material and pixel values of real images under various conditions was obtained. The characteristic curve was obtained using the relationship of two parameter and the results were verified using phantoms made of water and aluminum. The pixel values of the phantom image were estimated and compared with the characteristic curve under various conditions. It was found that the relationship between the DR outputs and the absorbed energy in the detector was almost linear. In a experiment using the phantoms, the estimated pixel values agreed with the characteristic curve, although the effect of scattered photons introduced some errors. However, effect of a scattered X-ray must be studied because it was not included in the calculation algorithm. The result of this study can provide useful information about a pre-processing of digital radiography.

• Progress in Medical Physics
• /
• v.25 no.4
• /
• pp.271-280
• /
• 2014

Recently, digital breast tomosynthesis (DBT) has been investigated to overcome the limitation of conventional mammography for overlapping anatomical structures and high patient dose with cone-beam computed tomography (CBCT). However incomplete sampling due to limited angle leads to interference on the neighboring slices. Many studies have investigated to reduce artifacts such as interference. Moreover, appropriate filters for tomosynthesis have been researched to solve artifacts resulted from incomplete sampling. The primary purpose of this study is finding appropriate filter scheme with FBP reconstruction for DBT system to reduce artifacts. In this study, we investigated characteristics of various filter schemes with simulation and prototype digital breast tomosynthesis under same acquisition parameters and conditions. We evaluated artifacts and noise with profiles and COV (coefficinet of variation) to study characteristic of filter. As a result, the noise with parameter 0.25 of Spectral filter reduced by 10% in comparison to that with only Ramp-lak filter. Because unbalance of information reduced with decreasing B of Slice thickness filter, artifacts caused by incomplete sampling reduced. In conclusion, we confirmed basic characteristics of filter operations and improvement of image quality by appropriate filter scheme. The results of this study can be utilized as base in research and development of DBT system by providing information that is about noise and artifacts depend on various filter schemes.

• Journal of Korean Academy of Oral and Maxillofacial Radiology
• /
• v.28 no.1
• /
• pp.111-126
• /
• 1998

In this study, radiographic evaluation was made using panoramic radiography and cross-sectional tomography of SCANORA/sup (R)/ in male and female adults in their 20's on the relationship between the maxillary sinus floor and the apex of the maxillary molar, to test the accuracy and effectiveness of the cross-sectional tomography, and to use this information in the assessment of preop. and postop. root canal treatment, apical surgery, extraction and implantology. Forty-one adults with an average age of 24.4 years were studied using panoramic radiography and cross-sectional tomography. In panoramic view and cross-sectional view, the position of the apices of maxillary molars were classified as separated, contacted, or protruded type; the general shape of the maxillary sinus floor was evaluated horizontally and vertically from cross-sectional tomography. The accuracy of each radiography was tested using maxilla from 5 fresh cadavers from the Anatomy Lab at Yonsei University Dental College, and panoramic view and cross-sectional tomography were taken in the same condition as with the patients. The results were as follows: 1. Panoramic view and cross-sectional view were taken in the maxilla specimen, and the actual distance between the maxillary sinus floor and the tooth apices were measured in the specimen; the median values of the distance from the tooth apices to the maxillary sinus floor in the panoramic view, cross-sectional view and in the actual maxilla specimen were 2.83 mm, 4.51mm, and 4.l5mm, respectively. In the cross-sectional view, the measured distance was close to the actual distance but in the panoramic view, the measured distance was far from the actual distance. 2. When the results of the panoramic view and cross-sectional view were compared, 40.5％ of the results agreed with each other in the two radiographic methods and buccal roots of the 2nd molar were the closest to the maxillary sinus floor in the cross-sectional tomography. 3. In cross-sectional view, when the vertical relationship of the maxillary sinus floor and maxillary roots was assessed, in 1st molars, type II (the sinus floor that extends down to the buccolingual furcation area) was predominant, while in 2nd molars, type I (the sinus floor located above the level connecting the buccal and lingual apices) was predominant. In the horizontal relationship, in 1st molars, type II (the lowest floor of the maxillary sinus located in between the buccal and lingual roots) was predominant; in 2nd molars, type I (the lowest floor of the maxillary sinus located on the buccal side of the buccal roots) and type II appeared in similar frequency. In conclusion, the SCANORA/sup (R)/ cross-sectional tomography was more effective than the frequently used panoramic view, in that the relationship of the maxillary molars and maxillary sinus floor can be evaluated more accurately and the buccolingual cross-sectional view can also be observed. And maxillary sinus floor that was close to maxillary 2nd molar had tendency to be located on buccal side than that close to 1st molar. Therefore, cross-sectional tomography is an effective and accurate method to evaluate the position of the teeth in relation to the sinus floor preoperative and can be easily used to diagnose localized periapical lesions. Also, the image quality obtained was quite satisfactory.

• Nuclear Medicine and Molecular Imaging
• /
• v.41 no.2
• /
• pp.172-180
• /
• 2007

Among the nuclear medicine imaging methods available today, $H_2^{15}O-PET$ is most widely used by cognitive neuroscientists to examine regional brain function via the measurement of regional cerebral blood flow (rCBF). The short half-life of the radioactively labeled probe, $^{15}O$, often allows repeated measures from the same subjects in many different task conditions. $H_2^{15}O-$ PET, however, has technical limitations relative to other methods of functional neuroimaging, e.g., fMRI, including relatively poor time and spatial resolutions, and, frequently, insufficient statistical power for analysis of individual subjects. However, recent technical developments, such as the 3-D acquisition method provide relatively good image quality with a smaller radioactive dosage, which in turn results in more PET scans from each individual, thus providing sufficient statistical power for the analysis of individual subject's data. Furthermore, the noise free scanner environment $H_2^{15}O$ PET, along with discrete acquisition of data for each task condition, are important advantages of PET over other functional imaging methods regarding studying state-dependent changes in brain activity. This review presents both the limitations and advantages of $^{15}O-PET$, and outlines the design of efficient PET protocols, using examples of recent PET studies both in the normal healthy population, and in the clinical population.

• Journal of Korean society of Dental Hygiene
• /
• v.5 no.2
• /
• pp.201-218
• /
• 2005

The purpose of this study was to examine how dentists perceived dental coordinators including their education, hiring criteria, working condition and job. It's basically attempted to help define the job and role of 5 and to suggest how they should be nurtured. The subjects in this study were dentists at dental hospitals and clinics where dental coordinators were employed among approximately 200 dental institutions in Seoul, Cyeonggi province and Incheon. After a survey was conducted in June 2005, answer sheets from 99 respondents were analyzed. The findings of the study were as follows: 1. Regarding education for dental coordinators, 99.9% of the dentists investigated felt the need for separate education programs for dental coordinators, 42.4% knew what would-be dental coordinators learned about, and 81.8% considered it necessary for them to take intermediate or higher courses. An organization affiliated with the Korea Dental Hygienists Association was viewed as the best institute to educate dental coordinators, and educational institutes that included a department of dental hygiene was looked upon as the second best one. 68.7% believed that dental coordinators should take an official examination to test their qualifications, and concerning educational subsidy, the largest group of the dentists thought that a certain amount of subsidy should be provided. 2. As for coordinator hiring, the top priority was the impression(look) of applicants(55%), followed by adjustability to existing employees(24.5%) and professional competency(17.3%). As to the route of hiring, 41.4 percent, the largest group, reeducated some of existing employees, and dental hygienists were regarded as the best personnels to serve as a coordinator. Concerning job performance, they put the most emphasis on interpersonal relationship, which was followed by executive ability, impression and career, 58.6% the largest group, believed that dental coordinators should have a three-year or higher career to work at a dental institute. 3. As to working conditions, 75.7%, the largest group, paid dental coordinators based on their job performance, and 23.2%, the second largest group, had their pay equal to that of dental hygienists, 88.9% allowed them to determine their own retirement age. 4. In regard to their perception of dental coordinators, the largest number of the dentists considered it necessary for them to keep receiving education(4.29), and the second largest group felt that they served to enhance the image of dental institutes(4.18). The third largest group thought that they contributed to letting patients more satisfied with the quality of dental services. But they tended not to agree that their turnover rate was low(3.04), and they didn't find them to receive appropriate education, either(3.10). 5. The current major job of coordinators associated with customer services was handling appointments with customers(91.9%), treating unsatisfied customers(85.9%), and controling waiting time(84.8%). Regarding self-management, coordinators directed their energy into having good manners(89.9%), acquiring fundamental dental knowledge(84.8%), and learning how to treat customers(83.8%). Concerning hospital affairs handling, they were most responsible for information desk(87.9%), which was followed by receiving(86.9%). As a result of investigating what type of job the dentists hoped dental coordinators to fulfill in consideration of their career, their age and type of investment, the dentists wanted them the most to speak a foreign language, which belonged to the field of self-management.

• Journal of radiological science and technology
• /
• v.39 no.1
• /
• pp.99-105
• /
• 2016

The noise power spectrum (NPS) is one of the most general methods for measuring the noise amplitude and the quality of an image acquired from a uniform radiation field. The purpose of this study was to compare different NPS methodologies by using megavoltage X-ray energies. The NPS evaluation methods in diagnostic radiation were applied to therapy using the International Electro-technical Commission standard (IEC 62220-1). Various radiation therapy (RT) devices such as TrueBeam$^{TM}$(Varian), BEAMVIEW$^{PLUS}$(Siemens), iViewGT(Elekta) and Clinac$^R$ iX (Varian) were used. In order to measure the region of interest (ROI) of the NPS, we used the following four factors: the overlapping impact, the non-overlapping impact, the flatness and penumbra. As for NPS results, iViewGT(Elekta) had the higher amplitude of noise, compared to BEAMVIEW$^{PLUS}$ (Siemens), TrueBeam$^{TM}$(Varian) flattening filter, Clinac$^{R}$iXaS1000(Varian) and TrueBeam$^{TM}$(Varian) flattening filter free. The present study revealed that various factors could be employed to produce megavoltage imaging (MVI) of the NPS and as a baseline standard for NPS methodologies control in MVI.