[논문]위 내시경 이미지 품질에 따른 병변 검출 모델의 성능 비교 연구

이율희; 김영재; 김광기

doi:10.9718/jber.2023.44.2.118

위 내시경 이미지 품질에 따른 병변 검출 모델의 성능 비교 연구
A Performance Comparison Study of Lesion Detection Model according to Gastroscopy Image Quality 원문보기

Journal of biomedical engineering research : the official journal of the Korean Society of Medical & Biological Engineering, v.44 no.2, 2023년, pp.118 - 124

이율희 (가천대학교 간호대학 간호학과) , 김영재 (가천대학교 의과대학 의공학교실) , 김광기 (가천대학교 의과대학 의공학교실)

Abstract ▼ AI-Helper

Many recent studies have reported that the quality of input learning data was vital to the detection of regions of interest. However, due to a lack of research on the quality of learning data on lesion detetcting using gastroscopy, we aimed to quantify the impact of quality difference in endoscopic images to lesion detection models using Image Quality Assessment (IQA) algorithms. Through IQA methods such as BRISQUE (Blind/Referenceless Image Spatial Quality Evaluation), Laplacian Score, and PSNR (Peak Signal-To-Noise) algorithm on 430 sheets of high quality data (HQD) and 430 sheets of low quality data (PQD), we showed that there were significant differences between high and low quality images in lesion detecting through BRISQUE and Laplacian scores (p<0.05). The PSNR value showed 10.62±1.76 dB on average, illustrating the lower lesion detection performance of PQD than HQD. In addition, F1-Score of HQD showed higher detection performance at 77.42±3.36% while F1-Score of PQD showed 66.82±9.07%. Through this study, we hope to contribute to future gastroscopy lesion detection assistance systems that involve IQA algorithms by emphasizing the importance of using high quality data over lower quality data.

주제어

표/그림 (9)

그림 그림 1. 품질이 좋지 않은 데이터의 예시: (a) 정상 어두운 이미지, (b) 정상 빛 반사 이미지, (c) 비정상 흐린 이미지, (d) 비정상 흐린 빛 반사 이미지 Fig. 1. Examples of Poor Quality Data: (a) normal dark image, (b) normal light-reflected image, (c) abnormal blurred image, (d) abnormal blurred and light-reflected image
그림 그림 2. 품질이 좋은 데이터의 예시: (a) ,(b) 정상 이미지, (c) 암 이 미지, (d) 암 출혈 이미지 Fig. 2. Examples of High Quality Data: (a), (b) normal image, (c) cancer image, (d) cancer bleeding image
그림 그림 3. RetinaNet의 아키텍처: ResNet-FPN을 백본 네트워크로 사용하고 각 크기에 대한 기능 맵을 사용하여 각 계층에 대한 분류 및 상자 회귀 분석 수행 Fig. 3. Architecture of RetinaNet: Using ResNet-FPN as a Backbone Network, use Feature Maps for each size to perform classification and box regression for each layer
그림 그림 4. BRISQUE, Laplacian Score 알고리즘을 사용하여 품질 평가 값을 시각화한 그림 상자: (a) BRISQUE (b) Laplacian Score Fig. 4. Box plot visualizing Poor Quality Data, High Quality Data quality assessment values using BRISQUE, Laplacian Score algorithm: (a) BRISQUE (b) Laplacian Score
그림 그림 5. 품질이 좋은 데이터 병변 검출 결과 예시: 초록색 상자=참값 ROI, 빨간색 상자=예측된 ROI Fig. 5. Examples of High Quality Data lesion detection result: The green box is the Ground Truth ROI and the red box is the predicted ROI
그림 그림 6. 품질이 좋지 않은 데이터 병변 검출 결과 예시: 초록색 상자 =참값 ROI, 빨간색 상자=예측된 ROI Fig. 6. Examples of Poor Quality Data lesion detection result: The green box is the Ground Truth ROI and the red box is the predicted ROI
표 표 1. 알고리즘을 사용한 품질 평가 집단 통계량 및 T검정 결과 Table 1. Quality Assessment Statistics and T Test Results Using Algorithms
그림 그림 7. 품질에 따른 병변 검출 모델 FROC Curve: (a) 품질이 좋지 않은 데이터 (b) 품질이 좋은 데이터 Fig. 7. Lesion Detection Model FROC Curve Based on Data Quality: (a) Poor Quality Data (b) High Quality Data
표 표 2. 품질에 따른 RetinaNet의 성능 평가 비교 표 Table 2. Comparison table of RetinaNet's performance assessment by data quality

참고문헌 (26)

Ferlay J, Ervik M, Lam F, Colombet M, Mery L, Pieros M,？Bray F. Global cancer observatory: cancer today. Lyon, France:？International Agency for Research on Cancer. 2018;3(20):1-2.
https://www.who.int/news-room/fact-sheets/detail/cancer.？Accessed on 3 Feb 2022. cancer.
Choi ES, Jeon WH, Sohn HJ. Clinical study of early gastric？carcinoma. Korean J Gastroenterol. 1995;27:31-38.
Park IS, Lee YC, Kim WH, Noh SH, Lee KS, Kim H. Clinicopathologic characteristics of early gastric cancer in Korea.？Yonsei Medical Journal. 2000;41(5):607-614.

상세보기
Song HT, Kim CD, Ryu HS, Hyun JH. A clinical study of？early gastric cancer. Korean J Gastroenterol. 1994;26:789-799.
National Cancer Center. Cancer Facts and Figures 2014. 2014;18-27.
Lee SA, Cho HC. A New Image Augmentation Method for？Improving Computer-aided Diagnosis System Performance.？The transactions of The Korean Institute of Electrical Engineers. 2021;70(1):102-107.
Belle A, Kon MA, Najarian K. Biomedical informatics for？computer-aided decision support systems: a survey. The Scientific World Journal. 2013;2013:1-8.
Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, Wang Y. Artificial intelligence in healthcare: past, present and future.？Stroke and Vascular Neurology. 2017;2(4):230-243.

상세보기
Miller RA. Medical diagnostic decision support systems-past, present, and future: a threaded bibliography and brief？commentary. Journal of the American Medical Informatics？Association. 1994;1(1):8-27.
Kim DH, Cho HC. Deep Learning based Computer-aided？Diagnosis System for Gastric Lesion using Endoscope. The？transactions of The Korean Institute of Electrical Engineers.？2018;67(7):928-933.
Lee SA, Cho HC, Cho HC. Computer-aided Diagnosis System？for Abnormalities Classification in Gastric Endoscopy Images？using Machine Learning. The Transactions of the Korean？Institute of Electrical Engineers. 2020;69(1):107-113.
Park SH, Cho YB. An Effect of Genetic Algorithm for Creating a？Dataset. Journal of Digital Contents Society. 2019;20(1):127-133.

상세보기
Lee JM, Ham GW, Bae KH, Park HK. Detection Ability of？Occlusion Object in Deep Learning Algorithm depending on？Image Qualities. Journal of the Korean Association of Geographic Information Studies. 2019;22(3):82-98.
Yoon MS, Oh JH, Kang BG, Im JU, Kwon GT, Han DG.？Effects of Image Extensions in Chest X-ray Deep Learning-Based Pneumothorax Detection Studies. Green Collection of？the Korean Society of Emergency Medicine's Academic Conference. 2020;2020(2):75.
Kim YJ, Kim KG. Development of an Optimized Deep？Learning Model for Medical Imaging. Journal of the Korean？Society of Radiology. 2020;81(6):1274-1289.
Cho YK, Lee Y, Lee K. Quality Evaluation of Ultrasonographic Equipment Using an ATS-539 Multipurpose Phantom in Veterinary Medicine. Journal of Veterinary Clinics.？2022;39(3):114-120.

원문보기 상세보기
Mittal A, Moorthy AK, Bovik AC. No-reference Image Quality？Assessment in the Spatial Domain. IEEE Transactions on？Image Processing. 2012;21(12):4695-4708.
Bansal R, Raj G, Choudhury T. Blur image detection using？Laplacian operator and Open-CV. in 2016 International Conference System Modeling & Advancement in Research Trends？(SMART). 2016;63-67.
He X, Cai D, Niyogi P. Laplacian score for feature selection,？in proceeding of Advances in Neural Information Processing？Systems. 2006;18:507-514.
Rustam Z, Primasari I, Widya, D. Classification of cancer？data based on support vectors machines with feature selection using genetic algorithm and laplacian score. In AIP Conference Proceedings. 2018;2023(1):020234.
Kang KS, Lee, J-H. PSNR appraisal of MRI image. Journal？of the Korean Society of Radiology. 2009;3(4):13-21.
Jung HK. Fundamentals of multimedia chapter 8, Lossy？compression Algorithms. Hallym University. 2008.
Lin TY, Goyal P, Girshick R, He K, Dollr P. Focal loss for？dense object detection. in Proceedings of the IEEE international conference on computer vision. 2017;2980-2988.
Lee JH. Radiological feature estimation and risk assessment？of mammographic mass and density by computer-aided analysis. Department of Medical Science The Graduate School,？Yonsei University. 2004;3-50.
Cheng HD, Cui M. Mass lesion detection with a fuzzy neural？network. Pattern Recognition. 2004;37(6):1189-1200.

내보내기 구분	파일저장 인쇄 메일전송
구성항목	기본정보 상세정보 관리번호, 논문명, 저널/프로시딩명, 저자 , 발행년, 권, 호, 시작페이지, 끝페이지, 발행기관 관리번호, 논문명, 대등논문명, 저자 , 저널/프로시딩명, 발행기관, 발행년, 발행언어, 권, 호, 시작페이지, 끝페이지, ISBN, ISSN, 주제분야, 키워드, 초록(한글), 초록(영문), 저자(소속기관)
저장형식	Text(ASCII format) Excel format RefWorks Direct Export RIS format (for Reference Manager, ProCite, EndNote), Scholar's Aids, Mendeley
메일정보	받는사람 (필수) @ 보내는사람 (선택) @ 제목 내용 KISTI 검색결과 이메일 서비스
안내	총 건의 자료가 검색되었습니다. 다운받으실 자료의 인덱스를 입력하세요. (1-10,000) 검색결과의 순서대로 최대 10,000건 까지 다운로드가 가능합니다. 데이타가 많을 경우 속도가 느려질 수 있습니다.(최대 2~3분 소요) 다운로드 파일은 UTF-8 형태로 저장됩니다. 파일의 내용이 제대로 보이지 않을실 때는 웹브라우저 상단의 보기 -> 인코딩 -> 자동선택 여부를 확인하십시오. ~ Text(ASCII format) Excel format

연합인증