MRI 신호획득과 영상재구성에서의 인공지능 적용

Junghwa Kang,Yoonho Nam 대한영상의학회 2022 대한영상의학회지 Vol.83 No.6

최근 인공지능기술은 자기공명영상(이하 MRI)의 폭넓은 분야에서 임상적 활용가치를 보여주고 있다. 특히, MRI에서 영상획득과정의 효율성 및 복원된 영상의 품질을 향상시키기 위한 목적으로 인공지능모델의 개발이 활발하다. 임상에서 활용되는 다양한 MRI 프로토콜에서 인공지능은 병렬영상기법과 같은 기존 가속화 방법 대비 추가적인 영상획득시간을 가능하게 해줄 수 것으로 기대된다. 또한, 펄스시퀀스 디자인, 영상의 인공물 감소, 자동화된 품질평가와 같은 영역에서도 인공지능모델은 도움을 줄 수 있는 연구 결과들이 소개되고 있다. 또한, 영상분석 과정에서 중요한 장비 및 프로토콜의 영향을 줄여줄 수 있는 방법으로도 인공지능 기반의 접근이 이루어지고 있다. 본 종설에서는 MRI 영상의 획득 과정에서 최근 인공지능기술들이 적용되고 있는 분야 및 해당 분야에서의 인공지능기술의 개발 및 적용과 관련된 현안들을 소개하고자 한다.

Convolutional Neural Network-Based Automatic Segmentation of Substantia Nigra on Nigrosome and Neuromelanin Sensitive MR Images

Kang, Junghwa,Kim, Hyeonha,Kim, Eunjin,Kim, Eunbi,Lee, Hyebin,Shin, Na-young,Nam, Yoonho Korean Society of Magnetic Resonance in Medicine 2021 Investigative Magnetic Resonance Imaging Vol.25 No.3

Recently, neuromelanin and nigrosome imaging techniques have been developed to evaluate the substantia nigra in Parkinson's disease. Previous studies have shown potential benefits of quantitative analysis of neuromelanin and nigrosome images in the substantia nigra, although visual assessments have been performed to evaluate structures in most studies. In this study, we investigate the potential of using deep learning based automatic region segmentation techniques for quantitative analysis of the substantia nigra. The deep convolutional neural network was trained to automatically segment substantia nigra regions on 3D nigrosome and neuromelanin sensitive MR images obtained from 30 subjects. With a 5-fold cross-validation, the mean calculated dice similarity coefficient between manual and deep learning was 0.70 ± 0.11. Although calculated dice similarity coefficients were relatively low due to empirically drawn margins, selected slices were overlapped for more than two slices of all subjects. Our results demonstrate that deep convolutional neural network-based method could provide reliable localization of substantia nigra regions on neuromelanin and nigrosome sensitive MR images.

Connectivity Maximization Algorithm for Indoor Users in Disaster Environment using Adaptive UAV Localization

Junghwa Kang(강정화),Jae-Hyun Kim(김재현) 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.2

Unmanned aerial vehicles (UAVs) can serve as aerial base stations when cellular networks are disrupted. Most studies on UAV communications in indoor disaster situations aimed to provide throughput while using less power consumption. However, it is necessary to establish a user connectivity and a path loss by considering the movements of UAVs and the target building situation. This paper proposes a new adaptive UAV localization algorithm to maximize the user connectivity subject to the throughput requirements in indoor disaster environment using searching time. In searching time, the connectivity of indoor users can be improved, whereas there are wasted movements. Therefore, we design an adaptive UAV localization using Lyapunov optimization to maximize connectivity with fewer UAV movements. Proposed localization algorithm guarantees throughput and connectivity, in contrast to the randomly located UAV scenario.

실내 재난 환경에서 강화학습 기반 UAV 최적 경로 학습 연구

강정화(Junghwa Kang),김경록(Kyeongrok Kim),김재현(Jae-Hyun Kim) 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.2

Unmanned aerial vehicle (UAV)는 셀룰러 네트워크를 이용하기 어려운 실내 재난 상황에서 공중 기지국 역할을 대신한다. 하지만 긴급 상황에서는 사용자들의 통신 요구사항과 지리적 환경 등이 급격하게 변하므로 통신 성능을 보장하기 어렵다. 따라서 본 논문에서는 non-orthogonal multiple access (NOMA)를 이용하여 네트워크의 성능을 보장하고 강화학습을 통해 UAV의 최적 이동 경로에 대해 학습한다. 성능분석 결과, NOMA를 사용한 경우, 기존의 프로토콜에 비해 총 소모 시간이 감소하였고, 할인 계수에 따라 총 소모 시간과 총 수신 데이터양 사이의 trade-off를 확인하였다.

이동 IP 환경에서의 효율적인 멀티캐스트 지원 방안

강정화(Junghwa Kang),임철수(Cheolsu Lim) 한국멀티미디어학회 2000 한국멀티미디어학회 학술발표논문집 Vol.2000 No.2

실내 재난 환경에서 강화학습과 가변적인 MCS 레벨 선택 기반 UAV 최적 경로 연구

강정화(Junghwa Kang),김재현(Jae-Hyun Kim) 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.6

O-GlcNAc Modification of MAVS Regulates Antiviral Signaling by Modulating Its Activity

Junghwa Seo,Tae Hyun Kweon,Jingu Kang,Seonjin Son,Jinju Song,Won Ho Yang,Jin Won Cho 한국당과학회 2021 한국당과학회 학술대회 Vol.2021 No.01

Post-translational modifications, including O-GlcNAcylation, play fundamental roles in modulating cellular events, including transcription, signal transduction, and immune signaling. Several molecular targets of O-GlcNAcylation associated with pathogen-induced innate immune responses have been identified; however, the direct regulatory mechanisms linking O-GlcNAcylation with antiviral RIG-I-like receptor signaling are not fully understood. In this study, we found that cellular levels of O-GlcNAcylation decline in response to infection with Sendai virus. We identified a heavily O-GlcNAcylated serine-rich region between amino acids 249–257 of the mitochondrial antiviral signaling protein (MAVS); modification at this site disrupts MAVS aggregation and prevents MAVS-mediated activation and signaling. O-GlcNAcylation of the serine-rich region of MAVS also suppresses its interaction with TRAF3; this prevents IRF3 activation and production of interferon-β. Taken together, these results suggest that O-GlcNAcylation of MAVS may be a master regulatory event that promotes host defense against RNA viruses.

Intravenous Fluid Selection for Unruptured Intracranial Aneurysm Clipping : Balanced Crystalloid versus Normal Saline

Kang, Jian,Song, Young Joo,Jeon, Sujeong,Lee, Junghwa,Lee, Eunsook,Lee, Ju-Yeun,Lee, Euni,Bang, Jae Seung,Lee, Si Un,Han, Moon-Ku,Oh, Chang Wan,Kim, Tackeun The Korean Neurosurgical Society 2021 Journal of Korean neurosurgical society Vol.64 No.4

Objective : While balanced crystalloid (BC) could be a relevant fluid regimen with buffer system compared with normal saline (NS), there have been no studies on the optimal fluid for surgery of an unruptured intracranial aneurysm (UIA). This study aimed to compare the effects of fluid regimens between NS and BC on the metabolic and clinical outcomes of patients who underwent surgery for UIA. Methods : This study was designed as a propensity score matched retrospective comparative study and included adult patients who underwent UIA clipping. Patient groups were categorized as NS and BC groups based on the types of pre-operative fluid and the amount of fluid administered during surgery. The primary outcomes were defined as electrolyte imbalance and acidosis immediately after surgery. The secondary outcomes were the length of stay in the intensive care unit (ICU) and duration from the end of the operation to extubation. Results : A total of 586 patients were enrolled in this study, with each of 293 patients assigned to the NS and BC groups, respectively. Immediately after surgery, serum chloride levels were significantly higher in the NS group. Compared to the NS group, the BC group had lower incidence rates of acidemia (6.5% vs. 11.6%, p=0.043) and metabolic acidosis (0.7% vs. 4.4%, p=0.007). As compared to NS group, BC group had significantly shorter duration from the end of the operation to extubation (250±824 vs. 122±372 minutes, p=0.016) and length of stay in ICU (1.37±1.11 vs. 1.12±0.61 days, p=0.001). Throughout multivariable analysis, use of BC was found to be significant factor for favorable post-operative results. Conclusion : This study showed that the patients who received BC during UIA clipping had lower incidence of metabolic acidosis, earlier extubation and shorter ICU stay compared to those who received NS. Therefore, using BC as a peri-operative fluid can be recommended for patients who undergo surgery for UIA.

Macrophage Activation Syndrome Presented in a Case of Neonatal Lupus

Kang, Chang Min,Choi, Jinwha,Lee, JungHwa The Korean Society of Neonatology 2021 Neonatal medicine Vol.28 No.3

Macrophage activation syndrome (MAS) is a potentially life-threatening complication in many autoimmune diseases. Early recognition and intervention are essential for a favorable outcome. Neonatal lupus, an acquired autoimmune disease in neonates caused by the transplacental passage of maternal autoantibodies, is rare and usually self-limited. Herein, we report a case of MAS in a patient with neonatal lupus, which improved with intravenous immunoglobulin.

Identifying CO₂ advection on a hill slope using information flow

Kang, Minseok,Ruddell, Benjamin L.,Cho, Chunho,Chun, Junghwa,Kim, Joon Elsevier 2017 Agricultural and forest meteorology Vol.232 No.-

<P><B>Abstract</B></P> <P>In hilly terrain affected by drainage flow, the horizontal advection of CO<SUB>2</SUB> makes it difficult to accurately observe the net ecosystem exchange of CO<SUB>2</SUB> by the eddy covariance technique. Downslope drainage can result in an overestimation of respiration at the bottom of a hill slope and an underestimation at the top, resulting in discrepancies among different flux corrections using filters based on the friction velocity, light response curve, and timing of advection. Vertical profiles of the CO<SUB>2</SUB> concentration from the ground to above the canopy were measured along with above-canopy EC flux measurements at the top and bottom of a hill slope at the Gwangneung KoFlux sites from 2008 to 2010. To infer the timing, direction, temporal scale, and structure of CO<SUB>2</SUB> advection from uphill to downhill, we constructed an information flow dynamical process network (DPN) based on the observed multi-level CO<SUB>2</SUB> concentrations. A site-specific quality control filter was developed to eliminate data strongly affected by CO<SUB>2</SUB> advection, which identifies the observations when strong downslope information flow exists in the DPN. This site-specific filter considerably reduced the discrepancies among different traditional flux corrections. This research provides a method for the general characterization of advection using information flow, and application of the method as a site-specific filter for eddy covariance observations in hilly and complex terrain.</P> <P><B>Highlights</B></P> <P> <UL> <LI> Drainage flow affects the eddy covariance CO<SUB>2</SUB> fluxes on the forest hill slope of the Gwangneung KoFlux sites. </LI> <LI> The CO<SUB>2</SUB> drainage is identified using an information flow dynamical process network (DPN). </LI> <LI> Based on the DPN, a site-specific filter was developed to eliminate the drainage-affected data. </LI> <LI> An information flow filter reduces discrepancies among the different nighttime flux corrections. </LI> </UL> </P>