Development of Polyp Detection Technology by Analyzing Deep-learning

은성종 차세대컨버전스정보서비스학회 2020 차세대컨버전스정보서비스기술논문지 Vol.9 No.2

This paper proposes a technique of guiding a clinician by recognizing colon polyp in patients in real time at the time of endoscopic examination. Automatic recognition of colon polyp by endoscopic image has applied Res-net’s advanced technique for calculating weight, existing deep learning method to enhance accuracy of recognizing colon polyp. Drawn result values provides information on shape of colon polyp guiding whether there is colon polyp and location of colon polyp in real time. Dataset used in developing analysis technique was collected from 5,000 cases from normal group and 5,000 cases from patients with colon polyp. For verification, accuracy was drawn by calculating confusion matrix with 10-fold cross validation. Accuracy of proposed technique was verified by drawing AUC of 0.96. Dataset was tested with data from control group and patient group in the ratio of 50%. We plan to develop deep learning technique that calculates the ratio of risk of occurrence of colorectal cancer. It is important to secure various test data such as images of colon polyp by type and whether there is adenocarcinoma. We plan to develop service that can predict ratio of risk of occurrence of colorectal cancer by applying relevant techniques automatically.

복부 MRI 영상에서의 자동 간 검출 알고리즘

은성종 차세대컨버전스정보서비스학회 2015 디지털예술공학멀티미디어논문지 Vol.2 No.2

Liver Segmentation can support not only checking the disease, but also information for deciding the method of surgery in abdominal MRI scans. In this paper, we propose an automatic method for liver segmentation. In this proposed method, we segment the liver region for using contour seed point by region growing method. As a result of this proposed method, we can check the efficiency at 85.7% performance. In future research, we will implement the 3D region growing in volume rendering for later. 복부 MR영상에서의 간 영역의 인식은 수술을 위한 확인뿐만 아니라 정보제공에 있어 중요한 역할을 한다. 따라서 본 논문은 이러한 간 영역의 검출을 자동으로 처리하는 알고리즘을 제안하고자 한다. 제안방법은 경계선의 정보를 씨앗점으로 한 영역성장법 기반의 간 영역 검출 방법을 수행한다. 제안방법 의 성능평가 결과 평균 85.7%의 정확도를 도출하였다. 향후, 효율적인 진단보조를 위해 3차원 기반의 영역성장법을 연구할 예정이다.

Colon Polyps Risk Correlation Analysis Technology based on Clinical Data

은성종 차세대컨버전스정보서비스학회 2020 차세대컨버전스정보서비스기술논문지 Vol.9 No.3

This paper proposes ways to draw elements having a high correlation to a risk of occurrence of colon polyp by analyzing life habit data and medical treatment data collected from physical examination center for the purpose of preventing colorectal cancer. A control group and a group with colon polyp are compared based on cross analysis and ANOVA to draw risk factors. Drawn risk factors include drinking habit, the average amount of drinking at one time per person, age of starting smoking, past duration of smoking, current duration of smoking, the average amount of smoking a day per person, whether family members were treated, and BMI. Validity of analysis of correlation can be verified by checking cases which P-value is less than 0.05. This paper is to propose service which provides life habit information record in physical examination center, cases of risk factors of colon polyp and colorectal cancer by using relevant analysis technique.

Development of an Artificial Intelligence-Based Support Technology for Urethral and Ureteral Stricture Surgery

은성종,박종목,김계환 대한배뇨장애요실금학회 2022 International Neurourology Journal Vol.26 No.1

Purpose: This paper proposes a technological system that uses artificial intelligence to recognize and guide the operator to the exact stenosis area during endoscopic surgery in patients with urethral or ureteral strictures. The aim of this technological solution was to increase surgical efficiency. Methods: The proposed system utilizes the ResNet-50 algorithm, an artificial intelligence technology, and analyzes images entering the endoscope during surgery to detect the stenosis location accurately and provide intraoperative clinical assistance. The ResNet-50 algorithm was chosen to facilitate accurate detection of the stenosis site. Results: The high recognition accuracy of the system was confirmed by an average final sensitivity value of 0.96. Since sensitivity is a measure of the probability of a true-positive test, this finding confirms that the system provided accurate guidance to the stenosis area when used for support in actual surgery. Conclusions: The proposed method supports surgery for patients with urethral or ureteral strictures by applying the ResNet-50 algorithm. The system analyzes images entering the endoscope during surgery and accurately detects stenosis, thereby assisting in surgery. In future research, we intend to provide both conservative and flexible boundaries of the strictures.

유사도 분석 기반의 자동 뇌졸중 진단 알고리즘 개발

은성종,권용만 차세대컨버전스정보서비스학회 2019 차세대컨버전스정보서비스기술논문지 Vol.8 No.1

This paper proposes algorithm processing calculation of ASPECT score automatically for diagnosis of stroke. In this paper, we propose a CT image based automatic ASPECT score determination algorithm to diagnose stroke, one of central nervous system diseases. ASPECT score to be judged is classified grades and increases according to severity of the disease. Classification is divided into 11 grades. The higher the severity of disease, the lower the score of ASPECT score. The proposed method compares the correlation information of the left and right regions based on the detected gray matter regions. We propose the method to judge difference of ASPECT score according to difference of correlation information to be compared. This paper proposes an effective method featuring accuracy and simplicity. 본 논문에서는 뇌졸중 진단을 위해 ASPECT 비율을 자동으로 계산하는 방법을 제안한다. 본 논문에서는 중추 신경계 중 하나인 뇌졸중을 진단하기 위해, CT영상 기반의 자동 분석 알고리즘을 제안한다. ASPECT 스코어의 결정은 질병의 중증도에 따라 분류 등급이 나누어진다. 등급의 분류는 총 11가지의 등급으로 나누며, 질환 중증도가 높을수록 낮은 ASPECT 스코어 점수로 산정되어 진다. 본 논문의 제안 방법은 검출된 회백질 영역을 기반으로, 뇌 좌우 영역의 상관 성보를 비교한다. 비교하고자 하는 상관 정보의 차이에 따라 ASPECT 스코어의 차이를 판단하는 방법을 제안한다. 제안 방법의 평가 결과 높은 정확성을 확인하였고, 기존의 방법에 비해 보다 단순화된 특징을 가지고 있어 효율성을 확인하였다.

설진 영상에서의 자동 홍반 검출 알고리즘

은성종 차세대컨버전스정보서비스학회 2016 디지털예술공학멀티미디어논문지 Vol.3 No.2

한방에서는 사람의 혀의 상태를 인체 생리학적, 임상학적인 건강 상태를 알수 있는 중요 진단 지표로사용하고 있다. 본 논문은 이러한 혀의 진단(설진)을 위해 자동 홍반 검출 알고리즘을 제안한다. 제안방법은 한방 진단의 효율적인 진단보조로 활용되고자 제안되었다. 성능평가는 한방 임상전문의의 기준과 비교하여 분석한 결과 평균 89.5%의 정확도를 도출하였다. 향후 특수한 경우의 홍반 검출이 아닌일반적인 자동 검출 방법의 연구를 진행할 예정이다. In oriental medicine, the status of a tongue is an important indicator in diagnosing the condition of one's health, like the physiological and the clinicopathological changes of internal organs in a body. In this paper, we propose an algorithm that demonstrates the extraction the erythema of the tongue region for tongue diagnosis. According to this, it can support the oriental medicine diagnosis through a proposed method. For experiment, we evaluate the performance in 89.5% accuracy by comparing doctors’ criteria with our result. In future research, we will implement the generalization in automatic erythema extraction algorithm.

안전한 해양 다이빙을 위한 맞춤형 다이빙 컴퓨터 개발

은성종,정은영,박동균,황보택근 한국차세대컴퓨팅학회 2017 한국차세대컴퓨팅학회 논문지 Vol.13 No.6

본 논문은 다이버의 안전요소를 높이기 위해 슈트에 부착된 텍스타일형 센서와 연동되는 맞춤형 다이빙 컴퓨터를 개발하는 것을 목적으로 하고 있다. 본 다이빙 컴퓨터의 개발은 크게 3가지 단계로 구분된다. 첫 번째로 슈트에 부착된 텍스타일형 센서와 연동되어 다이버의 상태를 체크할 수 있는 기능 개발, 두 번째로 다이빙 컴퓨터 개발을 기존폐쇄적인 OS기반에서 모바일 OS기반으로 개발, 세 번째로 방수 기능에 중점을 둔 실제 다이빙 컴퓨터 하드웨어의개발로 구성한다. 결과물의 성능 평가는 사용성 평가를 통해 84.7%의 종합 만족도와 기능성 평가를 통해 안전성및 효율성을 검증하였다. 향후 연구로 기존 제품들과의 차별성 확보를 위해 안전성 요소를 부각하여 지속적인 연구를 수행할 예정이다. In this paper, we develop the customized diving computer that is linked with a textile sensor attached to a diver suit to enhance the safety of the diver. The development of the diving computer is roughly divided into three stages. First, we develop the function that can check the state of the diver in conjunction with the textile type sensor attached to the diver suit first. Second, we develop the diving computer based on open source mobile platform rather than the existing closed OS platform. Third, we develop the actual hardware of diving computer, with emphasis on waterproof function. Our evaluation results verified the safety and efficiency by achieving the total satisfaction rate of 84.7% and through ease of use and functional evaluation. In future works, in order to ensure the differentiation from existing products, we will conduct ongoing research with highlighting safety factors. ※

급성 뇌경색 환자에서 Deconvolution perfusion CT의 유용성 : Diffusion MRI와 비교

은성종,김영근,Eun, Sung-Jong,Kim, Young-Keun 대한방사선기술학회 2003 방사선기술과학 Vol.26 No.3

급성 허혈성 뇌경색 환자에서 Deconvolution 관류 전산화 단층촬영(computed tomography, CT)의 뇌혈류용적(cerebral blood volume, CBV), 뇌혈류(cerebral blood flow, CBF) 그리고 평균 조영제 통과 시간(mean transit time, MTT) 지도를 확산 강조(Diffusion weighted) MRI과 비교하여 그 유용성을 알아보고자 하였다. 뇌관류 CT와 확산강조 자기공명영상(diffusion weighted MR imaging, DWMRI)을 시행한 급성뇌경색 환자를 대상으로 Deconvolution 기법으로 획득한 CBV, CBF, MTT 지도에서 병변의 면적과 DWMRI의 병변의 면적을 비교 측정하여 일치도를 알아보았다. 또한, 병변 부위와 정상측 대칭부위의 MTT시간을 측정하여 차이를 알아보았다. 본 연구의 모든 예에서 CBV, CBF, 그리고 MTT 세 지도 모두 관류 결손을 인지할 수 있었고, 관류 결손이 인지되는 부위에서 MTT 시간의 현저한 지연이 있었다. 뿐만 아니라, MTT 지도의 결손부위 면적은 DWMRI 보다 크게 나타나 허혈성 패넘브라를 추측할 수 있었다. 결론적으로 Deconvolution 관류 CT의 지도를 이용하면 뇌경색의 조기진단뿐만 아니라 허혈 중심부, 그리고 허혈성 페넘브라를 예측할 수 있어 DWMRI에 비해 관류결손 부위의 혈류 역학적 상태의 평가나 효과적인 치료계획을 세우는데 보다 유용할 것으로 생각된다. This study was performed to evaluate the usefulness of Deconvolution perfusion CT in patients with acute cerebral infarction. Nine patients with acute cerebral infarction underwent conventional CT and cerebral perfusion CT within 23 hours of the onset of symptoms. The perfusion CT scan for each patient was obtained at the levels of basal ganglia and 1cm caudal to the basal ganglia. By special imaging software, perfusion images including cerebral blood volume(CBV), cerebral blood flow(CBF), and mean transit time(MTT) maps were created. The created lesions were evaluated on each perfusion maps by 3 radiolocical technician. MTT delay time was measured in the perfusion defect lesion and symmetric contralateral normal cerebral hemisphere. Lesion sire were measured on each perfusion map and compared with the value obtained by diffusion weighted MR imaging(DWMRI). All perfusion CT maps showed the perfusion defect lesion in all patients. There were remarkable CT delay in perfusion defect lesion. In comparison of lesion size between each perfusion map and DWMRI, the lesion on CBF map was the most closely correlated with the lesion on DWMRI(7/9). The size of perfusion defect lesion on MTT map was larger than that of lesion on DWMRI, suggesting that m map can evaluate the ischemic penumbra. Deconvolution Perfusion CT maps make it possible to evaluate not only ischemic core and ischemic penumbra but also hemodynamic status in perfusion defect area. These results demonstrate that perfusion CT can be useful to the diagnosis and treatment in the patients with acute cerebral ischemic infarction.

Effective Brain Contour Segmentation based on Active Contour Model

은성종,정은영,박동균 차세대컨버전스정보서비스학회 2017 차세대컨버전스정보서비스기술논문지 Vol.6 No.2

Object recognition is usually processed based on region segmentation algorithm. This paper suggests effective region segmentation method based on R2-map within the magnetic resonance (MR) theory. When we do pre-processing, proposed method was composed of R2-map process. And then we do the Gray-white matter segmentation by Active Contour Model(ACM) in post-processing. In this study, the experiment had been conducted using images including the brain region and by getting up contrast enhancement image of R2-map for segmentation to extract region (white matter) segmentation even when the border line was not clear. As a result, an average area difference of 5.8%, which was higher than the accuracy of conventional region segmentation algorithm, was obtained.

X선 촬영시 테이블 주변 촬영도움자의 피폭선량 측정

은성종,김성길,민병운 한국방사선학회 2011 한국방사선학회 논문지 Vol.5 No.5

본 연구는 방사선 일반 촬영시 촬영 도움자가 피폭받을 수 있는 테이블 주변의 산란선에 의한 피폭선량을 알아보고 자 하였다. 두개부, 흉부, 복부, 요추부, 고관절, 그리고 슬관절 촬영시 테이블 중앙으로부터 옆으로 45cm, 75cm떨어 진 지점에서 바닥으로부터 70cm, 80cm, 130cm, 150cm 높이에서 산란선에 의한 피폭선량을 측정하였다. 측정결과 요추 측방향, 복부 정면, 그리고 고관절 촬영이 80cm 높이에서 66.21μSv, 34.22μSv, 와 32.35μSv로 가장 높은 피폭을 보였고, 흉부, 복부, 고관절, 두개부, 슬관절 순서로 선량이 감소하였다. 촬영 테이블 높이는 70cm 보다 80cm에서 대 체로 높은 선량을 보였으며, 80cm 이상에서는 높이가 높아질수록 선량은 감소하였다. 측정된 선량들은 촬영 도움자가 납 방호복을 착용할 경우 일반인 연간 피폭선량한도인 1mSv와는 큰 차이가 있어 우려할 선량은 아닌 것으로 나타났 다. The purpose of this study is to know the exposure dose nearby table on the general radiography(skull AP, chest PA, abdomen AP, lumbar lateral, hip joint, knee joint) and to find the reducing it. We measured beside the 45cm and 75cm table center, and 70cm, 80cm, 130cm, 150cm height from the bottom. That were measured highly from the radiography of lumbar lateral, abdomen AP and hip joint as followed 66.21μSv, 34.22μSv, and 32.35μSv at the 80cm height beside 45cm from the center of table. Measured doses were reduced in order of chest PA, abdomen, hipjoint, skull, knee joint projection. It appears exposure dose of nearby radiographic table was extremely low amount compared with limit of dose(1mSv). If it protected by Pb apron, exposure dose of assisting person will not be exceed a limit of dose. Conclusively, Wearing apron is very important to avoid radiation from the general radiography