http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
General Anesthesia of a Patient with Charcot-Marie-Tooth Disease
( Junbum Lee ),( Gahee Kim ),( Sangho Lee ),( Jeong-hyun Choi ) 경희대학교 경희의료원 2022 慶熙醫學 Vol.37 No.1
Charcot-Marie-Tooth (CMT) disease is a hereditary motor and sensory neuropathy of the peripheral nervous system. The initial presentation is usually foot drop or muscle wasting of the distal part of the leg. The symptom eventually progresses to atrophy of respiratory muscles, vocal cord and diaphragm. CMT patients require special caution during the general anesthesia. CMT disease can cause malignant hyperthermia, respiratory insufficiency and hyperkalemia during anesthesia. This report presents a 50-year-old man who has been diagnosed CMT disease and scheduled to undergo anterior cervical discectomy and fusion to treat cervical spinal stenosis. Propofol and remifentanil were used for the total intravenous anesthesia, and rocuronium was used for muscle relaxant purpose. Several intraoperative monitoring methods were performed to proceed the anesthesia safely, and the patient was discharged without any complications 5 days after the surgery. This case report with comprehensive literature review investigates possible adverse events and special cautions during the general anesthesia of CMT patients, and discusses adequate intraoperative monitoring to avoid such problems.
Deep Learning-based Human Action Recognition to Prevent Industrial Workplace Accidents
Jongmok Lee(이종목),Changyun Choi(최창윤),Junbum Park(박준범),Sungmin Kim(김성민),Seokman Sohn(손석만),Seungchul Lee(이승철) 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.4
A significant number of accidents at industrial workplaces such as power plants have always been a serious social issue. Various solutions were sought to prevent such problems, but these accidents constantly occurred at unexpected times and places. This paper proposes a human action recognition method to resolve this problem using images taken from the surveillance and security cameras (or CCTV) at the industrial workplace. Video frames are continuously gathered in real-time from CCTV, which is installed throughout the workplace. Then extracted skeleton data from workers feeds into a deep learning model to predict workers’ action. The goal is to check if the predicted behavior belongs to the predetermined risky class and then determine whether that worker is safe or not. In detail, the OpenPose method was applied to extract 2D skeleton data from human, and object detection and super-resolution were employed to improve the performance by enhancing the visibility of human within image frames. The Graph Convolutional Network (GCN) method is widely used in deep learning when dealing with locational-related information. Therefore, this paper used DGNN model, which used GCN method to predict human action from skeleton data. 3D NTU public datasets were processed in 2D and then trained with 2D DGNN models. The model showed 97.4% accuracy on tests, targeting five risky behaviors. We plan to conduct transfer learning on the pre-trained model with skeleton data obtained by filming selected behaviors with CCTV and then test it in the actual industrial workplace.
Analgesic effects of eucalyptus essential oil in mice
Lee, Ganggeun,Park, Junbum,Kim, Min Sun,Seol, Geun Hee,Min, Sun Seek The Korean Pain Society 2019 The Korean Journal of Pain Vol.32 No.2
Background: The use of aroma oils dates back to at least 3000 B.C., where it was applied to mummify corpses and treat the wounds of soldiers. Since the 1920s, the term "aromatherapy" has been used for fragrance therapy with essential oils. The purpose of this study was to determine whether the essential oil of Eucalyptus (EOE) affects pain pathways in various pain conditions and motor coordination. Methods: Mice were subjected to inhalation or intraperitoneal injection of EOE, and its analgesic effects were assessed by conducting formalin, thermal plantar, and acetic acid tests; the effects of EOE on motor coordination were evaluated using a rotarod test. To determine the analgesic mechanism, 5'-guanidinonaltrindole (${\kappa}$-opioid antagonist, 0.3 mg/kg), naltrindole (${\delta}$-opioid antagonist, 5 mg/kg), glibenclamide (${\delta}$-opioid antagonist, 2 mg/kg), and naloxone (${\mu}$-opioid antagonist, 4, 8, 12 mg/kg) were injected intraperitoneally. Results: EOE showed an analgesic effect against visceral pain caused by acetic acid (EOE, 45 mg/kg); however, no analgesic effect was observed against thermal nociceptive pain. Moreover, it was demonstrated that EOE did not have an effect on motor coordination. In addition, an anti-inflammatory effect was observed during the formalin test. Conclusions: EOE, which is associated with the ${\mu}$-opioid pain pathway, showed potential effects against somatic, inflammatory, and visceral pain and could be a potential therapeutic agent for pain.
김준범(JunBum Kim),홍성우(Seong-Woo Hong),이승재(SeungJae Lee) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
SMT 방식이 적용된 PCB에서는 패턴매칭, 객체 검출과 같은 방법으로 불량 검출을 했다. FPCB의 경우 PCB와 다르게 유연한 성질을 가지고 있어 기존의 방식을 사용할 경우 불량을 검출하는 것이 매우 어렵다. 이러한 이유로 자동화 대신 사람이 직접 검사를 하고 있어 낮은 생산성을 가지고 있고, 높은 비용을 필요로 한다. 본 논문에서는 위 문제들을 해결하기 위해 딥러닝을 이용한 해결책을 제시한다. 우리는 산업현장에서 요구하는 조건에 맞는 서로 다른 3가지 딥러닝 모델(U-Net, Siamese Neural Networks, AnoGAN))을 사용하여 실험을 진행하였고 각각의 결과에 대해 논의하고자 한다. On PCB, which is using SMT process, use such as pattern matching and object detection for defect inspection. In the case of FPCB, unlike PCB, it is hard to use the old ways. In addition, the current defect inspection cannot be automated, it has a low productivity and requires a high cost because a person directly inspects it. On this paper, we propose solution to solve these problems by using deep learning. We experiment 3 different model(U-Net, Siamese Neural Networks, AnoGAN), which is fit on request of industrial site, and discuss about the result.