http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Development of machine learning-based early warning system
( Bo-gun Kho ),( In-jae Oh ),( Min-seok Kim ),( Tae-ok Kim ),( Cheol-kyu Park ),( Young-chul Kim ),( Soon Whan Kang ),( Jae Hoon Jung ),( Sung Hoon Baek ) 대한결핵 및 호흡기학회 2019 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.127 No.-
A rapid response system is a patient safety system that prevents unexpected mortality by providing immediate interventions. But there are also difficulties in costs and human resources for implementation. Machine learning is an application of artificial intelligence that provides systems the ability to automatically learn and improve from experience without being explicitly programmed. We try to develop an early warning system to predict intubation and cardiopulmonary resuscitation using machine learning. The training set was composed of all patients admitted to Chonnam National University Hwasun Hospital for two years (event group 277 vs. non-event group 3576 cases), and the test set was made from all patients admitted to Chonnam National University Hospital for four months (event group 61 vs. non-event group 2334 cases). The algorithm is developed using blood pressure, heart rate, body temperature, respiratory rate, and oxygen saturation. We compared the ensemble model with recurrent neural network (RNN), random forest, logistic regression, and modified early warning score (MEWS) using receiver operating characteristic area under curve (ROC AUC) and precision recall AUC (PR AUC) analyses. The ROC AUC value of the ensemble model was 0.9219 and the PR AUC was 0.0732. This values showed higher tendency than those of other models (RNN: 0.8806 and 0.0732, random forest: 0.9125 and 0.0566, logistic regression: 0.8734 and 0.0513, MEWS: 0.7253 and 0.0468). And the score of the event group rose above the threshold 36 hours before events in the ensemble model. Development of machine learning-based early warning system is feasible because the ensemble model showed high ROC AUC value than traditional models. However, the low PR AUC value suggests that more efforts are needed to use in practice.
고보건 ( Bo Gun Kho ),고명주 ( Myoung Ju Koh ),김우정 ( Woo Jeung Kim ),김희욱 ( Hee Wook Kim ),허철웅 ( Cheal Wung Huh ),정혜문 ( Hye Moon Chung ),김형중 ( Hyung Jung Kim ) 대한결핵 및 호흡기학회 2012 Tuberculosis and Respiratory Diseases Vol.72 No.2
Pulmonary sequestration is a rare congenital anomaly of the lung in which it is separately supplied from the aorta or one of its branches. Bilateral pulmonary sequestration is very rare, particularly in adults. In bilateral pulmonary sequestration, resection of both sides is usually recommended if both sides are infected and symptomatic. We report the case of a 37-year-old female patient with bilateral intralobar pulmonary sequestration treated by staged bilateral lower lobectomy.
Concordance with artificial intelligence and multidisciplinary tumor board for lung cancer
( Min-Seok Kim ),( Bo-Gun Kho ),( Chul-Kyu Park ),( In-Jae Oh ),( Young-Chul Kim ),( Seok Kim ),( Ju-Sik Yun ),( Sang-Yun Song ),( Kook-Joo Na ),( Jae-Uk Jeong ),( Mee Sun Yoon ),( Sung-Ja Ahn ),( Sae 대한결핵 및 호흡기학회 2019 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.127 No.0
Background: IBM Watson for Oncology (WFO) is a cognitive computing system which helps physicians quickly identify key information from extensive medical records and explore best treatment option. We evaluated the level of agreement between WFO and multidisciplinary team (MDT) for lung cancer. Methods: From January to December 2018, newly diagnosed lung cancer cases in Chonnam National University Hwasun Hospital were retrospectively examined using WFO version 18.4 according to four treatment categories; surgery, radiotherapy, chemoradiotherapy, and palliative care - including chemotherapy, targeted therapy, immunotherapy. Treatment recommendations were considered concordant if the MDT recommendations were designated ‘recommended’ by WFO. The concordance between MDT and WFO was assessed by Cohen’s kappa value. Results: Total 405 (male 340, female 65) cases with different histology (adenocarcinoma 157, squamous cell carcinoma 132, small cell carcinoma 94, the others 22 cases) were enrolled in this study. The overall concordance rate between MDT and WFO was 92.4% (k=0.881, p<0.001). The strength of agreement was very good in stage 4 NS CLC (100%, k=1.000) and extensive disease SCLC (100%, k=1.000). The concordance was good in stage 1 NSCLC (92.4%, k=0.855) and moderate in stage 3 NSCLC (80.8%, k=0.622). But that was relatively low in stage 2 NSCLC (83.3%, k=0.556) and limited disease SCLC (84.6%, k=0.435). The discordant cases occurred in surgery (7/57, 12.3%), radiotherapy (2/12, 16.7%), and chemoradiotherapy (15/129, 11.6%). There was no discordance case in metastatic disease. Conclusion: Treatment recommendations made by WFO and MDT were different according to tumor stage, and WFO could be clinically useful in metastatic stage lung cancer. However patient-doctor relationship and shared decision making might be more important than artificial intelligence in non-metastatic stage.
Park, Ha-Young,Oh, In-Jae,Kho, Bo Gun,Kim, Tae-Ok,Shin, Hong-Joon,Park, Cheol Kyu,Kwon, Yong-Soo,Kim, Yu-Il,Lim, Sung-Chul,Kim, Young-Chul,Choi, Yoo-Duk The Korean Academy of Tuberculosis and Respiratory 2019 Tuberculosis and Respiratory Diseases Vol.82 No.3
Background: Programmed death-ligand 1 (PD-L1), a transmembrane protein, binds to the programmed death-1 (PD-1) receptor, and anti-PD-1 therapy enables immune responses against tumors. This study aimed to assess clinical characteristics of PD-L1 expression using immunohistochemistry among Korean patients with lung cancer. Methods: We retrospectively reviewed the data of patients with pathologically proven lung cancer from a single institution. PD-L1 expression determined by Tumor Proportion Score (TPS) was detected using 22C3 pharmDx (Agilent Technologies) and SP263 (Ventana Medical Systems) assays. Results: From July 2016 to July 2017, 267 patients were enrolled. The main histologic type was adenocarcinoma (69.3%). Most participants were smokers (67.4%) and had clinical stage IV disease (60.7%). In total, 116 (42%) and 58 (21%) patients had TPS ${\geq}1%$ and ${\geq}50%$, respectively. The patients were significantly older in TPS ${\geq}1%$ group than in TPS <1% group ($64.83{\pm}9.38years$ vs. $61.73{\pm}10.78years$, p=0.014), not in TPS ${\geq}50%$ cutoff value ($64.69{\pm}9.39$ vs. $62.36{\pm}10.51$, p=0.178). Regarding histologic grade, higher proportions of poorly differentiated tumor were observed in the TPS ${\geq}1%$ (40.8% vs. 25.8%, p=0.020) and TPS ${\geq}50%$ groups (53.2% vs. 27.2%, p=0.004). Among 34 patients examined with 22C3 and SP263 assays, 27 had positive results in both assays, with a cutoff of TPS ${\geq}1%$ (r=0.826; 95% confidence interval, 0.736-0.916). Conclusion: PD-L1 expression, defined as TPS ${\geq}1%$, was related to older age and poorly differentiated histology. There was a similar distribution of PD-L1 expression in both 22C3 and SP263 results.
P-49 Major adverse cardiovascular events and hyperuricemia during treatment of tuberculosis
( Hong-joon Shin ),( Ha-young Park ),( Bo-gun Kho ),( Jin-sun Chang ),( Tae-ok Kim ),( Cheol-kyu Park ),( Jung-hwan Lim ),( In-jae Oh ),( Yu-il Kim ),( Sung-chul Lim ),( Young-chul Kim ),( Yong-soo Kw 대한결핵 및 호흡기학회 2017 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.124 No.-
Background: Hyperuricemia is common drug effect during treatment of tuberculosis consisting pyrazinamide (PZA). However, it is unclear the relationship between PZA-induced hyperuricemia and major adverse cardiovascular events (MACE). Aim: We evaluated the relationship between MACE and hyperuricemia in patients treated with PZA-based antituberculous drugs. Methods: We retrospectively reviewed the medical charts of patients who treated with tuberculosis at Chonnam National University Hospital between January 2010 and June 2017. Results: Total 1,674 patients were treated with tuberculosis. Among them, 229 patients were excluded due to lack of available serum uric acid levels because of early follow-up loss or local transfer. Mean age was 57.9 and 887 patients (61.4%) were male. Most patients (75.6%) who treated with tuberculosis were diagnosed pulmonary tuberculosis. The initial standard treatment regimens of isoniazid, rifampin, ethambutol, and PZA were introduced to most patients (92.9%). Among them, 61.9% patients at 2 weeks and 59.5% at 2 months form starting treatment had hyperuricemia. MACE were occured only 17 patients (1.2%) during treatment. There was no relationship between hyperuricemia and MACE in the univariate analysis (odds ratio, 0.72; 95% confidence interval, 0.18─2.90; P=0.648). Conclusion: Most patients who received PZA-based antituberculous drugs had hyperuricemia, however, it was not associated with development of MACE.
( Hong-joon Shin ),( Min-seok Kim ),( Bo Gun Kho ),( Ha Young Park ),( Tae-ok Kim ),( Cheol-kyu Park ),( Yong-soo Kwon ),( In-jae Oh ),( Yu-il Kim ),( Sung-chul Lim ),( Young- Chul Kim ) 대한결핵 및 호흡기학회 2019 대한결핵 및 호흡기학회 추계학술대회 초록집 Vol.127 No.-
Background: Pneumocystis jirovecii pneumonia (PCP) is a fatal respiratory infection frequently associated with immunocompromised (IC) conditions. Although PCP has been reported in non-immunocompromised (non-IC) patients, however, few studies have been conducted. This study was aimed to compare the clinical characteristics and prognosis of PCP in IC and non-IC patients. Methods: We retrospectively analyzed patients who were suspected of having PCP with polymerase chain reaction (PCR) test positive for Pneumocystis jirovecii from January 2013 to May 2019. IC group was classified into human immunodeficiency virus (HIV), hematologic, solid organ tumor, rheumatologic and immunosuppressive agent group. Results: A total 192 PCP cases including 176 IC cases and 16 non-IC cases were analyzed. Patients were older in the non-IC group compared with the IC group (72.5 vs. 62.0, P=0.002). Hematologic malignancy was the most common (47.2%), followed by HIV (14.8%) in the IC group. The interval between test for PCP-PCR and PCP treatment was shorter in the IC group compared with non-IC group (0 [0-3] vs. 4.0 [2.2-7.7] days, P=0.001). In-hospital mortality was not significantly different between IC and non-IC groups (43.2% vs. 62.5%, P=0.189). Age (odds ratio [OR] 1.05; 95% confidence interval [CI] 1.02-1.09; P=0.002) and PaO<sub>2</sub>/FiO<sub>2</sub> (OR 0.99; 95% CI 0.98-1.00; P=0.039) were the prognostic factors for in-hospital mortality in the multivariate logistic regression analysis. There was no significant difference between IC and non-IC group in 6-month survival. However, HIV group had better 6-month survival compared with non- IC group in the subgroup analysis (Hazard ratio 0.16; 95% CI 0.05-0.53; P=0.003]. Conclusion: Patients with PCP in non-IC group were older than IC group, and had similar prognosis as other IC group except HIV group.
( Ha-young Park ),( In-jae Oh ),( Bo Gun Kho ),( Tae-ok Kim ),( Hong-joon Shin ),( Cheol Kyu Park ),( Yong-soo Kwon ),( Yu-il Kim ),( Sung-chul Lim ),( Young-chul Kim ),( Yoo-duk Choi ) 대한결핵 및 호흡기학회 2019 Tuberculosis and Respiratory Diseases Vol.82 No.3
Background: Programmed death-ligand 1 (PD-L1), a transmembrane protein, binds to the programmed death-1 (PD- 1) receptor, and anti-PD-1 therapy enables immune responses against tumors. This study aimed to assess clinical characteristics of PD-L1 expression using immunohistochemistry among Korean patients with lung cancer. Methods: We retrospectively reviewed the data of patients with pathologically proven lung cancer from a single institution. PD-L1 expression determined by Tumor Proportion Score (TPS) was detected using 22C3 pharmDx (Agilent Technologies) and SP263 (Ventana Medical Systems) assays. Results: From July 2016 to July 2017, 267 patients were enrolled. The main histologic type was adenocarcinoma (69.3%). Most participants were smokers (67.4%) and had clinical stage IV disease (60.7%). In total, 116 (42%) and 58 (21%) patients had TPS ≥1% and ≥50%, respectively. The patients were significantly older in TPS ≥1% group than in TPS <1% group (64.83 ± 9.38 years vs. 61.73 ± 10.78 years, p=0.014), not in TPS ≥50% cutoff value (64.69 ± 9.39 vs. 62.36 ± 10.51, p= 0.178). Regarding histologic grade, higher proportions of poorly differentiated tumor were observed in the TPS ≥1% (40.8% vs. 25.8%, p=0.020) and TPS ≥50% groups (53.2% vs. 27.2%, p=0.004). Among 34 patients examined with 22C3 and SP263 assays, 27 had positive results in both assays, with a cutoff of TPS ≥1% (r=0.826; 95% confidence interval, 0.736-0.916). Conclusion: PD-L1 expression, defined as TPS ≥1%, was related to older age and poorly differentiated histology. There was a similar distribution of PD-L1 expression in both 22C3 and SP263 results.
우엔쫑니아 ( Trong-nghia Nguyen ),보탄헝 ( Thanh-hung Vo ),고보건 ( Bo-gun Kho ),이귀상 ( Guee-sang Lee ),양형정 ( Hyung-jeong Yang ),김수형 ( Soo-hyung Kim ) 한국정보처리학회 2021 한국정보처리학회 학술대회논문집 Vol.28 No.2
In-hospital cardiac arrest is a significant problem for medical systems. Although the traditional early warning systems have been widely applied, they still contain many drawbacks, such as the high false warning rate and low sensitivity. This paper proposed a strategy that involves a deep learning approach based on a novel interpretable deep tabular data learning architecture, named TabNet, for the Rapid Response System. This study has been processed and validated on a dataset collected from two hospitals of Chonnam National University, Korea, in over 10 years. The learning metrics used for the experiment are the area under the receiver operating characteristic curve score (AUROC) and the area under the precision-recall curve score (AUPRC). The experiment on a large real-time dataset shows that our method improves compared to other machine learning-based approaches.