인문학
사회과학
자연과학
공학
의약학
농수해양학
예술체육학
복합학
DBpia AI
AI 에이전트
AI 아이디어
AI 리더
AI 뷰어
Citeasy
크롬 서지관리 다운로드
지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
논문 기본 정보
- 자료유형
- 학술대회자료
- 저자정보
- 발행연도
- 2025.11
- 수록면
- 1,446 - 1,450 (5page)
이용수
초록· 키워드
Computed tomography (CT) is a medical imaging procedure that uses X-rays from different angles to create clear cross-sectional images of the body. Although CT is especially useful, it exposes patients to radiation, which can increase the risk of health problems like cancer. To reduce this risk, low-dose CT (LDCT) is used, but this makes the images noisier and less clear, which can make diagnosis more difficult. Traditional methods to clean up noisy images usually require both normal-dose and low-dose image pairs, but obtaining such data is difficult and raises ethical concerns. In this study, we propose a self-supervised Neighbor2Neighbor denoising method that uses only single low-dose images for training. We use ResUNet as the base model and build three improved versions by adding Efficient Channel Attention (ECA) and Pyramid Pooling Module (PPM). We evaluated the models on whole-body CT images of piglets acquired with only 10% of the usual radiation dose [1]. We measured image quality using PSNR and SSIM compared to normal dose images. The results show that all our models perform better than the originalUNet and ResUNet, with less noise and clearer images.
#Low-Dose Computed Tomography
#Convolutional Neural Network
#Neighbor2Neighbor
#ResUNet
#Efficient Channel Attention
#Pyramid Pooling Module
상세정보 수정요청해당 페이지 내 제목·저자·목차·페이지정보가 잘못된 경우 알려주세요!
목차
- Abstract
- 1. INTRODUCTION
- 2. METHOD
- 3. EXPERIMENT
- 4. DISCUSSION AND CONCLUTION
- REFERENCES
참고문헌
참고문헌 신청최근 본 자료
