• The Journal of Korean Institute of Next Generation Computing
• /
• v.14 no.5
• /
• pp.31-39
• /
• 2018

Recently, there has been active studies to provide a user-friendly interface in a virtual reality environment by recognizing user hand gestures based on deep learning. However, most studies use separate sensors to obtain hand information or go through pre-process for efficient learning. It also fails to take into account changes in the external environment, such as changes in lighting or some of its hands being obscured. This paper proposes a hand gesture recognition method based on deep learning that is strong in external environments without the need for pre-process of RGB images obtained from general webcam. In this paper we improve the VGGNet and the GoogLeNet structures and compared the performance of each structure. The VGGNet and the GoogLeNet structures presented in this paper showed a recognition rate of 93.88% and 93.75%, respectively, based on data containing dim, partially obscured, or partially out-of-sight hand images. In terms of memory and speed, the GoogLeNet used about 3 times less memory than the VGGNet, and its processing speed was 10 times better. The results of this paper can be processed in real-time and used as a hand gesture interface in various areas such as games, education, and medical services in a virtual reality environment.

• Journal of Environmental Health Sciences
• /
• v.47 no.6
• /
• pp.521-529
• /
• 2021

Background: The coronavirus disease (COVID-19) has caused changes in human activity, and these changes may possibly increase or decrease exposure to fine dust (PM_2.5). Therefore, it is necessary to evaluate the exposure to PM_2.5 in relation to the outbreak of COVID-19. Objectives: The purpose of this study was to compare and evaluate the exposure to PM_2.5 concentrations by the variation of dynamic populations before and after the outbreak of COVID-19. Methods: This study evaluated exposure to PM_2.5 concentrations by changes in the dynamic population distribution in Guro-gu, Seoul, before and after the outbreak of COVID-19 between Jan and Feb, 2020. Gurogu was divided into 2,204 scale standard grids of 100 m×100 m. Hourly PM_2.5 concentrations were modeled by the inverse distance weight method using 24 sensor-based air monitoring instruments. Hourly dynamic population distribution was evaluated according to gender and age using mobile phone network data and time-activity patterns. Results: Compared to before, the population exposure to PM_2.5 decreased after the outbreak of COVID-19. The concentration of PM_2.5 after the outbreak of COVID-19 decreased by about 41% on average. The variation of dynamic population before and after the outbreak of COVID-19 decreased by about 18% on average. Conclusions: Comparing before and after the outbreak of COVID-19, the population exposures to PM_2.5 decreased by about 40%. This can be explained to suggest that changes in people's activity patterns due to the outbreak of COVID-19 resulted in a decrease in exposure to PM_2.5.

• The Korean Journal of Applied Statistics
• /
• v.35 no.4
• /
• pp.457-468
• /
• 2022

In aged society, it is important to prevent older people from being disability needing long-term care. The purpose of this study is to develop a prediction model to discover high-risk groups who are likely to be beneficiaries of Long-Term Care Insurance. This study is a retrospective study using database of National Health Insurance Service (NHIS) collected in the past of the study subjects. The study subjects are 7,724,101, the population over 65 years of age registered for medical insurance. To develop the prediction model, we used logistic regression, decision tree, random forest, and multi-layer perceptron neural network. Finally, random forest was selected as the prediction model based on the performances of models obtained through internal and external validation. Random forest could predict about 90% of the older people in need of long-term care using DB without any information from the assessment of eligibility for long-term care. The findings might be useful in evidencebased health management for prevention services and can contribute to preemptively discovering those who need preventive services in older people.

• The Journal of the Korea Contents Association
• /
• v.21 no.12
• /
• pp.57-65
• /
• 2021

A hyper-connected intelligence information society is emerging that creates new value by converging IoT, AI, and Bigdata, which are new technologies of the fourth industrial revolution, in all industrial fields. Everything is connected to the network and data is exploding, and artificial intelligence can learn on its own and even intellectual judgment functions are possible. In particular, the Internet of Things provides a new communication environment that can be connected to anything, anytime, anywhere, enabling super-connections where everything is connected. Artificial intelligence technology is implemented so that computers can execute human perceptions, learning, reasoning, and natural language processing. Artificial intelligence is developing advanced technologies such as machine learning, deep learning, natural language processing, voice recognition, and visual recognition, and includes software, machine learning, and cloud technologies specialized in various applications such as safety, medical, defense, finance, and welfare. Through this, it is utilized in various fields throughout the industry to provide human convenience and new values. However, on the contrary, it is time to respond as intelligent and sophisticated cyber threats are increasing and accompanied by potential adverse functions such as securing the technical safety of new technologies. In this paper, we propose a new data modeling method to enable IoT integrated security control by utilizing artificial intelligence technology as a way to solve these adverse functions.

• Journal of the Korean Society of Radiology
• /
• v.16 no.4
• /
• pp.411-417
• /
• 2022

This study aimed to learn and evaluate the effectiveness of VGGNet in the detection of pulmonary emphysema using low-dose chest computed tomography images. In total, 8000 images with normal findings and 3189 images showing pulmonary emphysema were used. Furthermore, 60%, 24%, and 16% of the normal and emphysema data were randomly assigned to training, validation, and test datasets, respectively, in model learning. VGG16 and VGG19 were used for learning, and the accuracy, loss, confusion matrix, precision, recall, specificity, and F1-score were evaluated. The accuracy and loss for pulmonary emphysema detection of the low-dose chest CT test dataset were 92.35% and 0.21% for VGG16 and 95.88% and 0.09% for VGG19, respectively. The precision, recall, and specificity were 91.60%, 98.36%, and 77.08% for VGG16 and 96.55%, 97.39%, and 92.72% for VGG19, respectively. The F1-scores were 94.86% and 96.97% for VGG16 and VGG19, respectively. Through the above evaluation index, VGG19 is judged to be more useful in detecting pulmonary emphysema. The findings of this study would be useful as basic data for the research on pulmonary emphysema detection models using VGGNet and artificial neural networks.

• Journal of the Korean Burn Society
• /
• v.22 no.1
• /
• pp.1-9
• /
• 2019

Purpose: The aim of this study was to investigate the characteristics of Acute Kidney Injury Network (AKIN)-defined nephrotoxicity in patients undergoing intravenous colistimethate sodium (CMS) therapy for major burns. Methods: This retrospective study included burn patients who received more than 48 h of intravenous CMS between September 2009 and December 2015. Data collection was performed using the institution's electronic medical record system. Patients assigned to the developed nephrotoxic group experienced aggravation of current AKIN stage during CMS treatment; those assigned to the non-nephrotoxic group experienced no change in current or exhibited improved AKIN stage during CMS therapy. Results: A total of 306 patients were included in this study. All patients were grouped according to AKIN stage: AKIN 0 (n=152); AKIN 1 (n=6); AKIN 2 (n=9); AKIN 3 (n=139). The baseline creatinine (Cr) level was 0.73 mg/dL. The incidence of nephrotoxicity was 50.3% according to AKIN stage; overall mortality was 45.8%. The non-nephrotoxic group consisted of 127 (74.7%) patients and 43 (25.3%) were in the developed nephrotoxic group. In patients requiring continuous renal replacement therapy (CRRT), baseline Cr level was 0.83 mg/dL, pre-CMS Cr level was 1.17 mg/dL, and post-CMS Cr level was 1.34 mg/dL. Conclusion: CMS can be administered without signs of nephrotoxicity for a certain period (approximately 1 week), it can be used relatively safely for 2 weeks. Application of CMS is a reasonable option for treating infections caused by multi-drug resistant gram-negative bacteria in patients with major burns. The caution should be exercised nevertheless.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.10
• /
• pp.67-76
• /
• 2023

Although deep learning models are making innovative achievements in the field of computer vision, the problem of vulnerability to adversarial examples continues to be raised. Adversarial examples are attack methods that inject fine noise into images to induce misclassification, which can pose a serious threat to the application of deep learning models in the real world. In this paper, we propose a model that detects adversarial examples using differences in predictive values between edge-learned classification models and underlying classification models. The simple process of extracting the edges of the objects and reflecting them in learning can increase the robustness of the classification model, and economical and efficient detection is possible by detecting adversarial examples through differences in predictions between models. In our experiments, the general model showed accuracy of {49.9%, 29.84%, 18.46%, 4.95%, 3.36%} for adversarial examples (eps={0.02, 0.05, 0.1, 0.2, 0.3}), whereas the Canny edge model showed accuracy of {82.58%, 65.96%, 46.71%, 24.94%, 13.41%} and other edge models showed a similar level of accuracy also, indicating that the edge model was more robust against adversarial examples. In addition, adversarial example detection using differences in predictions between models revealed detection rates of {85.47%, 84.64%, 91.44%, 95.47%, and 87.61%} for each epsilon-specific adversarial example. It is expected that this study will contribute to improving the reliability of deep learning models in related research and application industries such as medical, autonomous driving, security, and national defense.

• IMMUNE NETWORK
• /
• v.24 no.2
• /
• pp.3.1-3.23
• /
• 2024

Cigarette smoke extract (CSE)-treated mouse airway epithelial cells (MAECs)-derived exosomes accelerate the progression of chronic obstructive pulmonary disease (COPD) by upregulating triggering receptor expressed on myeloid cells 1 (TREM-1); however, the specific mechanism remains unclear. We aimed to explore the potential mechanisms of CSE-treated MAECs-derived exosomes on M1 macrophage polarization and pyroptosis in COPD. In vitro, exosomes were extracted from CSE-treated MAECs, followed by co-culture with macrophages. In vivo, mice exposed to cigarette smoke (CS) to induce COPD, followed by injection or/and intranasal instillation with oe-TREM-1 lentivirus. Lung function and pathological changes were evaluated. CD68⁺ cell number and the levels of iNOS, TNF-α, IL-1β (M1 macrophage marker), and pyroptosis-related proteins (NOD-like receptor family pyrin domain containing 3, apoptosis-associated speck-like protein containing a caspase-1 recruitment domain, caspase-1, cleaved-caspase-1, gasdermin D [GSDMD], and GSDMD-N) were examined. The expression of maternally expressed gene 3 (MEG3), spleen focus forming virus proviral integration oncogene (SPI1), methyltransferase 3 (METTL3), and TREM-1 was detected and the binding relationships among them were verified. MEG3 increased N6-methyladenosine methylation of TREM-1 by recruiting SPI1 to activate METTL3. Overexpression of TREM-1 or METTL3 negated the alleviative effects of MEG3 inhibition on M1 polarization and pyroptosis. In mice exposed to CS, EXO^-CSE further aggravated lung injury, M1 polarization, and pyroptosis, which were reversed by MEG3 inhibition. TREM-1 overexpression negated the palliative effects of MEG3 inhibition on COPD mouse lung injury. Collectively, CSE-treated MAECs-derived exosomal long non-coding RNA MEG3 may expedite M1 macrophage polarization and pyroptosis in COPD via the SPI1/METTL3/TREM-1 axis.

• IMMUNE NETWORK
• /
• v.23 no.4
• /
• pp.33.1-33.13
• /
• 2023

Vaccination against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been acknowledged as an effective mean of preventing infection and hospitalization. However, the emergence of highly transmissible SARS-CoV-2 variants of concern (VOCs) has led to substantial increase in infections among children and adolescents. Vaccine-induced immunity and longevity have not been well defined in this population. Therefore, we aimed to analyze humoral and cellular immune responses against ancestral and SARS-CoV-2 variants after two shots of the BNT162b2 vaccine in healthy adolescents. Although vaccination induced a robust increase of spike-specific binding Abs and neutralizing Abs against the ancestral and SARS-CoV-2 variants, the neutralizing activity against the Omicron variant was significantly low. On the contrary, vaccine-induced memory CD4⁺ T cells exhibited substantial responses against both ancestral and Omicron spike proteins. Notably, CD4⁺ T cell responses against both ancestral and Omicron strains were preserved at 3 months after two shots of the BNT162b2 vaccine without waning. Polyfunctionality of vaccine-induced memory T cells was also preserved in response to Omicron spike protein. The present findings characterize the protective immunity of vaccination for adolescents in the era of continuous emergence of variants/subvariants.

• IMMUNE NETWORK
• /
• v.22 no.2
• /
• pp.16.1-16.20
• /
• 2022

The gastrointestinal tract is the first organ directly affected by fasting. However, little is known about how fasting influences the intestinal immune system. Intestinal dendritic cells (DCs) capture antigens, migrate to secondary lymphoid organs, and provoke adaptive immune responses. We evaluated the changes of intestinal DCs in mice with short-term fasting and their effects on protective immunity against Listeria monocytogenes (LM). Fasting induced an increased number of CD103⁺CD11b^- DCs in both small intestinal lamina propria (SILP) and mesenteric lymph nodes (mLN). The SILP CD103⁺CD11b^- DCs showed proliferation and migration, coincident with increased levels of GM-CSF and C-C chemokine receptor type 7, respectively. At 24 h post-infection with LM, there was a significant reduction in the bacterial burden in the spleen, liver, and mLN of the short-term-fasted mice compared to those fed ad libitum. Also, short-term-fasted mice showed increased survival after LM infection compared with ad libitum-fed mice. It could be that significantly high TGF-β2 and Aldh1a2 expression in CD103⁺CD11b^- DCs in mice infected with LM might affect to increase of Foxp3⁺ regulatory T cells. Changes of major subset of DCs from CD103⁺ to CD103^- may induce the increase of IFN-γ-producing cells with forming Th1-biased environment. Therefore, the short-term fasting affects protection against LM infection by changing major subset of intestinal DCs from tolerogenic to Th1 immunogenic.