• Journal of the Korean Society of Radiology
• /
• v.14 no.7
• /
• pp.991-1001
• /
• 2020

Supervised deep learning technologies for improving the image quality of computed tomography (CT) need a lot of training data. When input images have different characteristics with training images, the technologies cause structural distortion in output images. In this study, an imaging model based on the deep reinforcement learning (DRL) was developed for overcoming the drawbacks of the supervised deep learning technologies and reducing noise in CT images. The DRL model was consisted of shared, value and policy networks, and the networks included convolutional layers, rectified linear unit (ReLU), dilation factors and gate rotation unit (GRU) in order to extract noise features from CT images and improve the performance of the DRL model. Also, the quality of the CT images obtained by using the DRL model was compared to that obtained by using the supervised deep learning model. The results showed that the image accuracy for the DRL model was higher than that for the supervised deep learning model, and the image noise for the DRL model was smaller than that for the supervised deep learning model. Also, the DRL model reduced the noise of the CT images, which had different characteristics with training images. Therefore, the DRL model is able to reduce image noise as well as maintain the structural information of CT images.

• Journal of the Korean Society of Radiology
• /
• v.18 no.1
• /
• pp.21-26
• /
• 2024

The one of the famous deep learning models for object detection task is you only look once version 5 (YOLOv5) framework based on the one stage architecture. In addition, YOLOv5 model indicated high performance for accurate lesion detection using the bottleneck CSP layer and skip connection function. The purpose of this study was to evaluate the performance of YOLOv5 framework according to various hyperparameters in position emission tomogrpahy (PET) phantom images. The dataset was obtained from QIN PET segmentation challenge in 500 slices. We set the bounding box to generate ground truth dataset using labelImg software. The hyperparameters for network train were applied by changing optimization function (SDG, Adam, and AdamW), activation function (SiLU, LeakyRelu, Mish, and Hardwish), and YOLOv5 model size (nano, small, large, and xlarge). The intersection over union (IOU) method was used for performance evaluation. As a results, the condition of outstanding performance is to apply AdamW, Hardwish, and nano size for optimization function, activation function and model version, respectively. In conclusion, we confirmed the usefulness of YOLOv5 network for object detection performance in nuclear medicine images.

• Journal of the Korean earth science society
• /
• v.32 no.2
• /
• pp.161-169
• /
• 2011

Focal mechanism solutions in the central and western areas of the Korean Peninsula (36-37.8$^{\circ}N$, 126-128$^{\circ}E$) were obtained from the analysis of the recent seventeen earthquakes (M${\geq}$2.2) which occurred from January, 2005 to May, 2010. The spatial differences between the epicenters recalculated by this study and those announced by the Korea Meteorological Administration are less than $0.03^{\circ}$, indicating a small deviation. Focal mechanism solutions were obtained from the analysis of P wave polarities, SH wave polarities and SH/P amplitude ratios. The focal mechanism solutions show dominant strike-slip faulting or oblique slip faulting with strike-slip components. The P-axes trends are mainly ENE-WSW or E-W directions. The direction of fault plane and auxillary fault plane with NNE-SSW and WNW-ESE are almost parallel to the general trends of lineaments in the study area. The results also show that focal mechanism solutions and the main axis of stress field in the Kyonggi massif and Okchon belt are almost same.

• The Korean Journal of Nuclear Medicine
• /
• v.35 no.3
• /
• pp.142-151
• /
• 2001

Purpose: This study was performed to search the relatively specific brain regions related to the semantic processing of Korean and English words on the one hand and the regions common to both on the other. Materials and Methods: Regional cerebral blood flow associated with different semantic tasks was examined using $[^{15}O]H_2O$ positron omission tomography in 13 healthy volunteers. The tasks consisted of semantic tasks for Korean words, semantic tasks for English words and control tasks using simple pictures. The regions specific and common to each language were identified by the relevant subtraction analysis using statistical parametric mapping. Results: Common to the semantic processing of both words, the activation site was observed in the fusiform gyrus, particularly the left side. In addition, activation of the left inferior temporal gyrus was found only in the semantic processing of English words. The regions specific to Korean words were observed in multiple areas, including the right primary auditory cortex; whereas the regions specific to English words were limited to the right posterior visual area. Conclusion: Internal phonological process is engaged in performing the visual semantic task for Korean words of the high proficiency, whereas visual scanning plays an important role in performing the task for English words of the low proficiency.

• The KIPS Transactions:PartD
• /
• v.14D no.5
• /
• pp.501-508
• /
• 2007

Simulation system was introduced and used a lot in the fields of aviation, vessel, and medical treatment. 3D Simulation system has been used quite insufficiently as it requires a lot of system resource and huge amount of computer calculation. As the graphic card performance and simulation function developed, however, PC based simulation has been activated and is verified of its possibility as an educational software. However, educational institutions need to invest huge amount of budget and manpower to purchase and maintain CT Equipment. For such a reason, educational institutions entrust their students to hospitals for indirect experience of operation or for mere observation. This study, therefore, developed a CT Virtual reality education system with which medical CT Equipment can be directly operated in PC based 3D Virtual environment.

• The Korean Journal of Nuclear Medicine Technology
• /
• v.25 no.1
• /
• pp.21-28
• /
• 2021

Purpose Brown fat, or brown adipose tissue (BAT), is involved in non-shivering thermogenesis and creates heat through glucose metabolism. BAT activation occurs stochastically by internal factors such as age, sex, and body mass index (BMI) and external factors such as temperature and environment. In this study, as a retrospective, electronic medical record (EMR) observation study, statistical analysis is conducted to confirm BAT activation and various factors. Materials and Methods From January 2018 to December 2019, EMR of patients who underwent PET/CT scan at the National Cancer Center for two years were collected, a total of 9155 patients were extracted, and 13442 case data including duplicate scan were targeted. After performing a univariable logistic regression analysis to determine whether BAT activation is affected by the environment (outdoor temperature) and the patient's condition (BMI, cancer type, sex, and age), A multivariable regression model that affects BAT activation was finally analyzed by selecting univariable factors with P<0.1. Results BAT activation occurred in 93 cases (0.7%). According to the results of univariable logistic regression analysis, the likelihood of BAT activation was increased in patients under 50 years old (P<0.001), in females (P<0.001), in lower outdoor temperature below 14.5℃ (P<0.001), in lower BMI (P<0.001) and in patients who had a injection before 12:30 PM (P<0.001). It decreased in higher BMI (P<0.001) and in patients diagnosed with lung cancer (P<0.05) In multivariable results, BAT activation was significantly increased in patients under 50 years (P<0.001), in females (P<0.001) and in lower outdoor temperature below 14.5℃ (P<0.001). It was significantly decreased in higher BMI (P<0.05). Conclusion A retrospective study of factors affecting BAT activation in patients who underwent PET/CT scan for 2 years at the National Cancer Center was conducted. The results confirmed that BAT was significantly activated in normal-weight women under 50 years old who underwent PET/CT scan in weather with an outdoor temperature of less than 14.5℃. Based on this result, the patient applied to the factor can be identified in advance, and it is thought that it will help to reduce BAT activation through several studies in the future.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2016.02a
• /
• pp.277.2-277.2
• /
• 2016

현재 화합물 반도체 나노구조는 광학적, 전기적 특성을 기반으로 하는 단전자 트랜지스터, 적외선 검출기, 레이저, 태양전지와 같은 분야에 응용하기 위한 많은 연구가 진행되고 있다. 특히 양자점은 3차원으로 구속되어 있는 상태 밀도를 갖고 있어 레이저 응용 시 낮은 문턱 전류 밀도, 높은 이득, 높은 열적 안정성을 기대되고 있지만 양자점의 운반자 수집과 열적 안정성의 한계가 여전히 존재한다. 이러한 문제를 해결하기 위해 다양한 방법이 연구되고 있으며, 그 중 단층 양자점에 비해 운반자 수집과 열적 안정성이 뛰어난 다층 양자점이 결합된 구조에 대한 연구가 활발히 이루어지고, 다층으로 성장된 양자점 구조는 양자점의 크기 분포 조절이 용이하고 양자점 층간의 전기적 결합력이 강한 특성이 있다. 본 연구에서는 분자 선속 에피 성장법(Molecular Beam Epitaxy; MBE)과 원자 층 교대 성장법(Atomic Layer Epitaxy; ALE)으로 CdTe/ZnTe 다층 양자점을 ZnTe 장벽층의 두께를 변화하면서 성장 후 광학적 특성을 연구하였다. 저온 광루미네센스 측정(Photoluminescence; PL)을 통하여 ZnTe 장벽층 두께가 증가할수록 양자점의 PL 피크가 높은 에너지로 이동함을 알 수 있었는데, 이는 ZnTe 장벽층의 두께가 증가할수록 양자점 층간의 결합력이 감소하면서 양자점의 크기가 작아졌기 때문이다. 그리고 ZnTe 장벽층의 두께가 증가할수록 PL 세기가 커지는 것을 알 수 있었는데, 이는 ZnTe 장벽층의 두께가 증가할수록 더 많은 운반자가 양자점으로 구속되기 때문이다. 또한 온도 의존 광루미네센스 측정 결과 ZnTe 장벽층의 두께가 증가할수록 열적 활성화 에너지가 커지는 것을 관찰하였고, 시분해 광루미네센스 측정을 통해 ZnTe 장벽층의 두께에 따른 운반자 동역학에 대해 연구하였다. 이와 같은 결과 CdTe/ZnTe 다층 양자점 구조에서 장벽층의 두께에 따른 광학적 특성에 대해 이해 할 수 있었다.

• Applied Microscopy
• /
• v.30 no.3
• /
• pp.311-319
• /
• 2000

The gill morphology and ultrastructure of the fleshy shrimp, Penaeus chinensis were investigated by light and electron microscopy. Fleshy shrimp has dendrobranchiate gills. Gill has a longitudinal septum dividing them into afferent and efferent channel. Each gill lamella is covered by multi-layered thin cuticle of different electron density. The lamella basal cell is squamous and contains cytoplasm of electron dense. Simple epithelial layer consists of squamous epithelium contained large nucleus. The lamella pillar structures are characterized by the axial microtubules and lateral membrane interdigitations Secretory cells of AB-PAS negative are multicellular gland. In active gland each cell boundary is not apparent and the cytoplasm contains smooth endoplasmic reticula, mitochondria, membrane-bounded secretory vesicles of low electron density and granular resettes. In inactive gland each cell boundary is apparent and the cytoplasm is occupied with numerous small granules of electron dense. The well-developed rough endoplasmic reticula and Golgi apparatus are observed in the unicellular gland of alcian blue positive.

• The korean journal of orthodontics
• /
• v.38 no.6
• /
• pp.437-455
• /
• 2008

Condylar hyperplasia is a pathologic condition showing 3-dimensional skeletal hyperplasia of the mandible. The reason for condylar hyperplasia is not yet known, but the effects of hormone, trauma, infection, genetics, fetal condition, and hypervascularity are known as possible reasons. When we diagnose a patient as having condylar hyperplasia, it is important to decide if it is in progress or not. Treatment for facial asymmetry due to condylar hyperplasia are decided accordingly, including condylectomy, that is removal of growth site of the affected condyle, and conventional orthognathic surgery only or condylectomy with orthognathic surgery after the completion of growth. Therefore, it is important to determine the growth state of condylar hyperplasia in treatment stability. This is verified through bone scan and regular check-ups with 3D CT or PA cephalogram. This case report introduces an improved case of facial asymmetry with condylectomy together with orthognathic surgery.

• The Korean Journal of Nuclear Medicine
• /
• v.32 no.3
• /
• pp.238-249
• /
• 1998

Purpose: To localize and compare the neural basis of verbal and visual human working memory, we performed functional activation study using $H_2^{15}O$ PET. Materials and Methods: Repeated $H_2^{15}O$ PET scans with one control and three different activation tasks were performed on six right-handed normal volunteers. Each activation task was composed of 13 match-ing trials. On each trial, four targets, a fixation dot and a probe were presented sequentially and subject's task was to press a response button to indicate whether or not the probe was one of the previous targets. Short meaningful Korean words, simple drawings and monochromic pictures of human faces were used as matching objects for verbal or visual memory. All the images were spatially normalized and the differences between control and activation states were statistically analyzed using SPM96. Results: Statistical analysis of verbal memory activation with short words showed activation in the left Broca's area, promoter cortex, cerebellum and right cingulate gyrus. In verbal memory with simple drawings, activation was shown in the larger regions including where activated with short words and left superior temporal cortex, basal ganglia, thalamus, prefrontal cortex, anterior portion of right superior temporal gyrus and right infero-lateral frontal cortex. On the other hand, the visual memory task activated predominantly right-sided structures, especially inferior frontal cortex, supplementary motor cortex and superior parietal cortex. Conclusion: The results are consistent with the hypothesis of the laterality and dissociation of the verbal and visual working memory from the invasive electrophysiological studies and emphasize the pivotal role of frontal cortex and cingulate gyrus in working memory system.