• Clinical Endoscopy
• /
• 제55권3호
• /
• pp.390-400
• /
• 2022

Background/Aims: Previous artificial intelligence (AI) models attempting to segment gastric intestinal metaplasia (GIM) areas have failed to be deployed in real-time endoscopy due to their slow inference speeds. Here, we propose a new GIM segmentation AI model with inference speeds faster than 25 frames per second that maintains a high level of accuracy. Methods: Investigators from Chulalongkorn University obtained 802 histological-proven GIM images for AI model training. Four strategies were proposed to improve the model accuracy. First, transfer learning was employed to the public colon datasets. Second, an image preprocessing technique contrast-limited adaptive histogram equalization was employed to produce clearer GIM areas. Third, data augmentation was applied for a more robust model. Lastly, the bilateral segmentation network model was applied to segment GIM areas in real time. The results were analyzed using different validity values. Results: From the internal test, our AI model achieved an inference speed of 31.53 frames per second. GIM detection showed sensitivity, specificity, positive predictive, negative predictive, accuracy, and mean intersection over union in GIM segmentation values of 93%, 80%, 82%, 92%, 87%, and 57%, respectively. Conclusions: The bilateral segmentation network combined with transfer learning, contrast-limited adaptive histogram equalization, and data augmentation can provide high sensitivity and good accuracy for GIM detection and segmentation.

• 대한의용생체공학회:의공학회지
• /
• 제41권2호
• /
• pp.101-106
• /
• 2020

Although benign gastric ulcers do not develop into gastric cancer, they are similar to early gastric cancer and difficult to distinguish. This may lead to misconsider early gastric cancer as gastric ulcer while diagnosing. Since gastric cancer does not have any special symptoms until discovered, it is important to detect gastric ulcers by early gastroscopy to prevent the gastric cancer. Therefore, we developed a Convolution Neural Network (CNN) model that can be helpful for endoscopy. 3,015 images of gastroscopy of patients undergoing endoscopy at Gachon University Gil Hospital were used in this study. Using ResNet-50, three models were developed to classify normal and gastric ulcers, normal and gastric cancer, and gastric ulcer and gastric cancer. We applied the data augmentation technique to increase the number of training data and examined the effect on accuracy by varying the multiples. The accuracy of each model with the highest performance are as follows. The accuracy of normal and gastric ulcer classification model was 95.11% when the data were increased 15 times, the accuracy of normal and gastric cancer classification model was 98.28% when 15 times increased likewise, and 5 times increased data in gastric ulcer and gastric cancer classification model yielded 87.89%. We will collect additional specific shape of gastric ulcer and cancer data and will apply various image processing techniques for visual enhancement. Models that classify normal and lesion, which showed relatively high accuracy, will be re-learned through optimal parameter search.

• Journal of Korean Neurosurgical Society
• /
• 제58권6호
• /
• pp.513-517
• /
• 2015

Objective : Microscopic indocyanine green (ICG) angiography is useful for identifying the completeness of aneurysm clipping and the preservation of parent arteries and small perforators. Neuroendoscopy is helpful for visualizing structures beyond the straight line of the microscopic view. We evaluated our prototype of endoscopic ICG fluorescence angiography in swine, which we developed in order to combine the merits of microscopic ICG angiography and endoscopy. Methods : Our endoscopic ICG system consists of a camera, a light source, a display and software. This system can simultaneously display real-time visible and near infrared fluorescence imaging on the same monitor. A commercially available endoscope was used, which was 4 mm in diameter and had an angle of $30^{\circ}$. A male crossbred swine was used. Results : Under general anesthesia, a small craniotomy was performed and the brain surface of the swine was exposed. ICG was injected via the ear vein with a bolus dose of 0.3 mg/kg. Visible and ICG fluorescence images of cortical vessels were simultaneously observed on the display monitor at high resolution. The real-time merging of the visible and fluorescent images corresponded well. Conclusion : Simultaneous visible color and ICG fluorescent imaging of the cortical vessels in the swine brain was satisfactory. Technical improvement and clinical implication are expected.

• 대한의용생체공학회:의공학회지
• /
• 제19권5호
• /
• pp.531-538
• /
• 1998

본 논문은 내시경 수술시 입체영상을 제공하여 수술의 편리성과 정확성을 향상시키기 위한 고화질 3차원 내시경 영상의 처리와 디스플레이 방법을 연구하였다. 기존의 3타원 내시경 수술을 위한 디스플레이 방법은 안경 차용의 무게감과 플리커가 심한 전자셔터 방식으로 사용의 불편함이 많았다 된 연구에서는 기존 내시경 영상의 3D 디스플레이 방식인 전자셔터식에 대한 입체 영상의 특징을 분석하고, 이에 대한 문제점을 보완하기 위한 편광방식의 입체 영상처리와 재현 방법에 대하여 연구하였다. 또한 설계한 디스플레이 시스템의 성능 평가를 수행하여 기존 방식인 CRT방식에 의한 영상재현 방법에 비해 화질과 시야각 특성에서 우수함을 확인하였다.

• 한국컴퓨터정보학회논문지
• /
• 제26권3호
• /
• pp.19-28
• /
• 2021

위내시경 촬영은 조기에 위 병변을 진단하기 위해서 주로 사용한다. 하지만 위내시경을 했음에도 불구하고 위 내부를 자세히 관찰하지 못해서 10~20% 위 병변을 놓치는 경우가 생기는 것으로 보고되고 있다. 미국, 영국, 일본 등의 일부 국가와 세계내시경협회(Wold Endoscopy Organization)에서는 위내시경 시에 맹점 없는 관찰을 위해서 반드시 촬영해야 할 부위에 대한 촬영지침을 제안한 바 있다. 이에 본 논문에서는 수련의가 내시경을 하는 데 있어 위 내부를 자동으로 맹점 없이 관찰하는데 필요한 딥러닝 기술인 해부학적 분류모델을 제안한다. 제안한 모델은 위내시경 이미지에 적합한 전처리 모듈과 데이터 증강 기술들을 사용한다. 실험결과를 통해 최대 F1 점수 99.6% 분류 성능을 확인하였다. 또한, 실제 데이터를 통한 실험결과에서도 에러율이 4% 미만을 보였다. 이러한 성능을 바탕으로 설명 가능한 모델임을 보여 임상에서의 사용 가능성을 확인하였다.

• Clinical Endoscopy
• /
• 제55권1호
• /
• pp.113-121
• /
• 2022

Background/Aims: We have been developing artificial intelligence based polyp histology prediction (AIPHP) method to classify Narrow Band Imaging (NBI) magnifying colonoscopy images to predict the hyperplastic or neoplastic histology of polyps. Our aim was to analyze the accuracy of AIPHP and narrow-band imaging international colorectal endoscopic (NICE) classification based histology predictions and also to compare the results of the two methods. Methods: We studied 373 colorectal polyp samples taken by polypectomy from 279 patients. The documented NBI still images were analyzed by the AIPHP method and by the NICE classification parallel. The AIPHP software was created by machine learning method. The software measures five geometrical and color features on the endoscopic image. Results: The accuracy of AIPHP was 86.6% (323/373) in total of polyps. We compared the AIPHP accuracy results for diminutive and non-diminutive polyps (82.1% vs. 92.2%; p=0.0032). The accuracy of the hyperplastic histology prediction was significantly better by NICE compared to AIPHP method both in the diminutive polyps (n=207) (95.2% vs. 82.1%) (p<0.001) and also in all evaluated polyps (n=373) (97.1% vs. 86.6%) (p<0.001) Conclusions: Our artificial intelligence based polyp histology prediction software could predict histology with high accuracy only in the large size polyp subgroup.

• 대한의용생체공학회:의공학회지
• /
• 제18권2호
• /
• pp.121-126
• /
• 1997

Endoscopy has become a crucial diagnostic and therapeutic procedure in clinical areas. Over the past three years, we have developed a computerized system to record and store clinical data pertaining to endoscopic surgery of laparascopic cholecystectomy, pelviscopic endometriosis, and surgical arthroscopy. In this study, we developed a computer system, which is composed of a frame yabber, a sound board, a VCR control board, a LAN card and EDMS(endoscopic data management software. Also, computer system has controled peripheral instruments such as a color video printer, a video cassette recorder, and endoscopic input/output signals(image and doctor's comment). Digital endoscopic data management system is based on open architecture and a set of widely available industry standards, namely: windows 3.1 as a operating system, TCP/IP as a network protocol and a time sequence based database that handles both images and doctor's cotnments. For the purpose of data storage, we used MOD and CD-R. Digital endoscopic system was designed to be able to store, recreate, change, and compress signals and medical images.

• 한국실내디자인학회:학술대회논문집
• /
• 한국실내디자인학회 2002년도 춘계학술발표대회 논문집
• /
• pp.62-65
• /
• 2002

The purpose of this study was first, to confirm of the key hypothetical concepts through the research of previous literature analysis results, and second, to Identify responses of various prospective consumers on key hypothetical concepts and the validity of the plan for improvement before construction was employed through the small group workshop. The participants in workshop were working housewives with 1 or 2 children, residing in an apartment of a net area of 25.7py. The materials used in the process were questionnaire, existing apartment floor plan, developing floor plan for universal apartment, small scale model, endoscopy images, computer simulation images, and recording equipment like audio recorder, video tape recorder, and camera.

• Current Optics and Photonics
• /
• 제7권4호
• /
• pp.398-407
• /
• 2023

We explored a method to evaluate imaging performance for the optimal assembly of an endoscopic miniature lens and a sensor constituting an imaging module at the distal end of gastrointestinal endoscopy. For the assembly of the imaging module, the image sensor was precisely located at the focal plane when collimated light passed through the endoscopic lens. As another method, the distance between the lens and sensor was adjusted to obtain the highest focus index from images measured the star chart of the International Organization for Standardization (ISO) standard at various positions. We analyzed the slanted-edge modulation transfer function (MTF), corresponding depth of field, and number of line pairs for MTF 50% and 20% at each working distance within the range of 5-100 mm for imaging modules assembled in different ways. Assembly conditions of the imaging module with better MTF performance were defined for each working distance range of 5-30 mm and 30-100 mm, respectively. In addition to the MTF performance, the focus index of each assembled module was also compared. In summary, we examined the performance of imaging modules assembled with different methods within the suggested working distance and tried to establish the optimal assembly protocol.

• 지능정보연구
• /
• 제15권2호
• /
• pp.33-48
• /
• 2009

캡슐내시경검사는 일반 내시경 검사에 비해 고통이 없고 합병증이 적다고 보고되어 있어 향후 지속적인 발전 가능성이 매우 큰 분야로 잘 알려져 있다. 그러나, 캡슐내시경을 이용한 진단은 캡슐이 동일한 위치에 상주할 경우 반복적으로 촬영된 유사한 영상을 오래도록 관찰 하여야 하므로, 진단자로 하여금 막대한 시간적 비용을 발생하게 한다. 따라서 보다 현실적이고 실용적인 캡슐내시경 검사를 위한 효율적인 탐색 및 진단 방법으로써 캡슐내시경영상에 대한 지능형 탐색방법이 요구된다. 본 논문에서 제안하는 지능형 판독보조 시스템은 영상차감을 통해 중복영상을 최소화한 후 프레임단위로 영상이 내포한 정보를 일차원도표(map)의 형태로 제공하고, 이러한 결과도표의 분석도구 및 방법을 제안함으로써 진단시간을 큰 폭으로 단축할 수 있는 방법을 제안하였다. 즉, 비교연산 한 정규화된 교차상관(Normalized Cross-Correlation) 방법을 통해 전처리 된 인접영상에 대한 유사도를 추출하고, 설정된 임계값이상의 영상들만을 탐색 범위로 지정하여 중복 촬영된 영상의 탐색을 최소화 한다. 이외에도 영상간 유사도, 엔트로피와 명암도를 통해 얻어진 이동도표, 특성도표와 명암도표를 분석하여 효율적으로 사용자가 탐색을 원하는 부위에 대한 탐색밀도를 높이는 등의 다양한 진단 매뉴얼을 제시한다.