Prediction of Malignant Lymph Node in EBUS-TBNA for Lung Cancer Patients Using Deep Learning

( Yeonjeong Heo ),( Choon Geun Lee ),( Sang Won Yoon ),( Jaeyoung Cho ),( Nakwon Kwak ),( Sun Mi Choi ),( Jinwoo Lee ),( Chang-hoon Lee ),( Sang-min Lee ),( Chul-gyu Yoo ),( Young Sik Park ) 대한결핵 및 호흡기학회 2020 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.128 No.-

Background During endobronchial ultrasound transbronchial needle aspiration (EBUS-TBNA), visual inspection is very important for prediction of malignant lymph node. In this study, we trained a deep convolutional neural network of archived ultrasound images of lymph nodes from lung cancer patients. Methods We collected ultrasound lymph nodes images of EBUS-TBNA of lung cancer patients from May 2010 to Feb 2020, which have been obtained during routine practice. All images were randomly divided as 4.9:2.1:3 for training, validation, and test datasets with the same proportion of malignant lymph nodes. We developed classification model using pretrained convolutional neural network algorithm (VGG16), and finally the performance of the model was evaluated using the test dataset. To compare the performance of the model, visual estimation of the same test dataset was done by three bronchoscopists. Results The total of 2952 ultrasound images of EBUS-TBNA were collected from 1249 lung cancer patients. Among these lymph nodes, 26.7% were malignant. The classification model was developed and trained using 2067 lymph nodes, and tested using 885 lymph nodes. The classification accuracy of the deep learning model was 79.3%. The average classification accuracy of visual estimation by three bronchoscopists was 68.7%. Conclusions We obtain the classification model with 79.3% accuracy using deep learning, which is comparable to expert visual estimation.

Correlation between Physical Activity and Lung Function in Dusty Areas: Results from the Chronic Obstructive Pulmonary Disease in Dusty Areas (CODA) Cohort

Han, Yuri,Heo, Yeonjeong,Hong, Yoonki,Kwon, Sung Ok,Kim, Woo Jin The Korean Academy of Tuberculosis and Respiratory 2019 Tuberculosis and Respiratory Diseases Vol.82 No.4

Background: Although physical activity is known to be beneficial to lung function, few studies have been conducted to investigate the correlation between physical activity and lung function in dusty areas. Therefore, the purpose of this study is to investigate the correlation between physical activity and lung function in a Korean cohort including normal and COPD-diagnosed participants. Methods: Data obtained from the COPD in dusty areas (CODA) cohort was analyzed for the following factors: lung function, symptoms, and information about physical activity. Information on physical activity was valuated using questionnaires, and participants were categorized into two groups: active and inactive. The evaluation of the mean lung function, modified Medical Research Council dyspnea grade scores, and COPD assessment test scores was done based on the participant physical activity using a general linear model after adjusting for age, sex, smoking status, pack-years, height, and weight. In addition, a stratification analysis was performed based on the smoking status and COPD. Results: Physical activity had a correlation with high forced expiratory volume in 1 second ($FEV_1$) among CODA cohort (p=0.03). While the active group exhibited significantly higher $FEV_1$ compared to one exhibited by the inactive group among past smokers (p=0.02), no such correlation existed among current smokers. There was no significant difference observed in lung function after it was stratified by COPD. Conclusion: This study established a positive correlation between regular physical activity in dusty areas and lung function in participants.

Correlation between Physical Activity and Lung Function in Dusty Areas: Results from the Chronic Obstructive Pulmonary Disease in Dusty Areas (CODA) Cohort

( Yuri Han ),( Yeonjeong Heo ),( Yoonki Hong ),( Sung Ok Kwon ),( Woo Jin Kim ) 대한결핵 및 호흡기학회 2019 Tuberculosis and Respiratory Diseases Vol.82 No.4

Background: Although physical activity is known to be beneficial to lung function, few studies have been conducted to investigate the correlation between physical activity and lung function in dusty areas. Therefore, the purpose of this study is to investigate the correlation between physical activity and lung function in a Korean cohort including normal and COPD-diagnosed participants. Methods: Data obtained from the COPD in dusty areas (CODA) cohort was analyzed for the following factors: lung function, symptoms, and information about physical activity. Information on physical activity was valuated using questionnaires, and participants were categorized into two groups: active and inactive. The evaluation of the mean lung function, modified Medical Research Council dyspnea grade scores, and COPD assessment test scores was done based on the participant physical activity using a general linear model after adjusting for age, sex, smoking status, pack-years, height, and weight. In addition, a stratification analysis was performed based on the smoking status and COPD. Results: Physical activity had a correlation with high forced expiratory volume in 1 second (FEV₁) among CODA cohort (p=0.03). While the active group exhibited significantly higher FEV₁ compared to one exhibited by the inactive group among past smokers (p=0.02), no such correlation existed among current smokers. There was no significant difference observed in lung function after it was stratified by COPD. Conclusion: This study established a positive correlation between regular physical activity in dusty areas and lung function in participants.

중학생의 컴퓨팅 사고력 군집 유형과 특성 분석

허희옥 ( Heeok Heo ),박연정 ( Yeonjeong Park ) 한국교육공학회 2019 교육공학연구 Vol.35 No.3

제4차 산업혁명에 따른 사회 및 경제 구조의 변화에 대비하는 인재 양성과 교육에 컴퓨팅 사고력을 기반으로 하는 고차적인 문제 해결력, 창의·융합 능력의 증진에 관한 관심이 전 세계적으로 집중되고 있다. 우리나라에서도 2015 개정 교육과정에 따라 초·중학교에서 소프트웨어(SW)교육 역량을 증진하기 위한 교육을 시행하고 있다. 이에 따라 SW교육의 효과 증진을 위한 다양한 연구와 실천이 이어지고 있다. 본 연구에는 SW교육의 핵심 목표인 컴퓨팅 사고력에 초점을 두고 있다. 컴퓨팅 사고력은 문제를 이해하고 해결하기 위하여 컴퓨팅 기술을 활용하는 능력을 말하며, 분석, 설계, 실현, 평가 능력으로 구성될 수 있다. 본 연구의 목적은 학생들의 컴퓨팅 사고력 유형을 분석하고 각 유형의 특징을 살펴보면서 효과적인 SW교육을 위한 시사점을 얻는 데 있다. 이를 위하여, 중학생 2,949명을 대상으로 컴퓨팅 사고력 기반의 문제해결 과정을 조사하고 그 결과를 토대로 학생들의 특성을 군집분석 방법으로 파악하였다. K-means 군집분석을 실시한 결과, 1차 모델로 5가지 유형을 도출하고 유형별로 중학생들의 흥미와 효능감 간의 유의한 차이가 있음을 발견하였다. 한편, 흥미와 효능감을 포함한 2차 군집모델을 중심으로 4가지 유형으로 구성된 군집과 이에 따른 학생의 성별, 지역 규모, 거주지 권역, 선도/일반학교 여부 등에서 유의한 차이를 확인하였다. 도출된 군집 유형은 학생들의 컴퓨팅 사고력 수준을 파악하는 것뿐만 아니라, 하위 능력별 강점과 약점 역량에 따른 교육 방향을 설정하는 데에 활용될 수 있다. The forth industrial revolution leads the social and economic changes which require human to enhance complex problem-solving, creativity, and convergence based on the computational thinking. The Korean national curriculum revised in 2015 highlights the software education for the elementary and middle school students. Since then various research and practices have conducted to improve the quality of software education as focusing on computational thinking. Therefore, the purpose of this study was to analyze the level and patterns of computational thinking of middle school students. Computational thinking is defined by the process of approaching a complex problem in a systematic manner and solving it in the way the computer science experts carry out. It consists of the subsets including the analysis of problem, design of algorithm, execution of solution, and evaluation. This study assessed computational thinking competency of 2,949 middle school students and analyzed their levels and patterns of the competency. The first clustering model including five patterns on computational thinking was derived by the K-means methods and the differences according to students’ interest and efficacy on information education were revealed significantly. Consequently, the second clustering model based on the six variables including four subsets of computational thinking and the two affective variables (interest and efficacy) produced four clusters with the balanced silhouette. Also, students’ demographic characteristics such as gender, the size and division of the residence, the school characteristics were significantly different according to the four clusters. Finally, some implications of the results were discussed for effective software education.

Mortality Rate Analysis Based on the Vital Signs During Intensive Care Unit Admission Using Random Forest Regression Model

( Oh Beom Kwon ),( Sang Jee Park ),( Tae Jun Ha ),( Yeonjeong Heo ),( Da Hye Moon ),( Seon-sook Han ),( Ki Won Moon ),( Woo Jin Kim ),( Seung-joon Lee ),( Jeongwon Heo ) 대한결핵 및 호흡기학회 2023 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.136 No.0