자기 지도 학습훈련 기반의 Noise2Void 네트워크를 이용한 PET 영상의 잡음 제거 평가: 팬텀 실험
The Evaluation of Denoising PET Image Using Self Supervised Noise2Void Learning Training: A Phantom Study
2
조회
0
다운로드
1
인용
0
참고문헌
키워드
초록
Positron emission tomography (PET) images is affected by acquisition time, short acquisition times results in low gamma counts leading to degradation of image quality by statistical noise. Noise2Void(N2V) is self supervised denoising model that is convolutional neural network (CNN) based deep learning. The purpose of this study is to evaluate denoising performance of N2V for PET image with a short acquisition time. The phantom was scanned as a list mode for 10 min using Biograph mCT40 of PET/CT (Siemens Healthcare, Erlangen, Germany). We compared PET images using NEMA image-quality phantom for standard acquisition time (10 min), short acquisition time (2min) and simulated PET image (S2 min). To evaluate performance of N2V, the peak signal to noise ratio (PSNR), normalized root mean square error (NRMSE), structural similarity index (SSIM) and radio-activity recovery coefficient (RC) were used. The PSNR, NRMSE and SSIM for 2 min and S2 min PET images compared to 10min PET image were 30.983, 33.936, 9.954, 7.609 and 0.916, 0.934 respectively. The RC for spheres with S2 min PET image also met European Association of Nuclear Medicine Research Ltd. (EARL) FDG PET accreditation program. We confirmed generated S2 min PET image from N2V deep learning showed improvement results compared to 2 min PET image and The PET images on visual analysis were also comparable between 10 min and S2 min PET images. In conclusion, noisy PET image by means of short acquisition time using N2V denoising network model can be improved image quality without underestimation of radioactivity.
Abstract
Positron emission tomography (PET) images is affected by acquisition time, short acquisition times results in low gamma counts leading to degradation of image quality by statistical noise. Noise2Void(N2V) is self supervised denoising model that is convolutional neural network (CNN) based deep learning. The purpose of this study is to evaluate denoising performance of N2V for PET image with a short acquisition time. The phantom was scanned as a list mode for 10 min using Biograph mCT40 of PET/CT (Siemens Healthcare, Erlangen, Germany). We compared PET images using NEMA image-quality phantom for standard acquisition time (10 min), short acquisition time (2min) and simulated PET image (S2 min). To evaluate performance of N2V, the peak signal to noise ratio (PSNR), normalized root mean square error (NRMSE), structural similarity index (SSIM) and radio-activity recovery coefficient (RC) were used. The PSNR, NRMSE and SSIM for 2 min and S2 min PET images compared to 10min PET image were 30.983, 33.936, 9.954, 7.609 and 0.916, 0.934 respectively. The RC for spheres with S2 min PET image also met European Association of Nuclear Medicine Research Ltd. (EARL) FDG PET accreditation program. We confirmed generated S2 min PET image from N2V deep learning showed improvement results compared to 2 min PET image and The PET images on visual analysis were also comparable between 10 min and S2 min PET images. In conclusion, noisy PET image by means of short acquisition time using N2V denoising network model can be improved image quality without underestimation of radioactivity.
참고문헌 (24)
- LeCun Y, "Gradientbased learning applied to document recognition", Proceeding of the IEEE, 86(11), pp.2278~2324 (1998)
- Buades A, "A Non-local algorithm for image denoising", Proceedings Conference on Computer Vision and Pattern Recognition, pp.60~65 (2005)
- 정의환, "핵의학 감마카메라 정도관리의 딥러닝 적용", 방사선기술과학, 43(6), pp.461~467 (2020), 대한방사선과학회
- Chen J, "An ensemble of convolution neural networks for image classification based on LSTM", Proceedings of International Conference on Green Informatics(ICGI), pp.217~222 (2017)
- Tajbakhsh N, "Convolution neural networks for medical image analysis: Full training or fine tuning?", IEEE Trans Med Imaging, 35, pp.1299~1312 (2016)
- Anwar SM, "Medical image analysis using convolutional neural networks : A review", Journal of Medical System, 42, pp.226~ (2018)
- Zhang K, "Beyond a Gaussian denoiser : Residual learning of deep cnn for image denoising", IEEE Transactions on Image Processing, 26(7), pp.3142~3155 (2017)
- Burger HC, "Image denoising:Can plain neural networks compete with BM3D?", Computer Vision and Pattern Recgnotion(CVPR). IEEE Conference on IEEE (2012)
- Chen Y, "Trainable nonlinear reaction diffusion : A flexible framework for fast and effective image restoration", IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6), pp.1256~1272 (2017)
- Jain V, "Natural image denoising with convolutional networks", Advances in Neural Information Processing Systems, pp.769~776 (2009)
- 허광해, "영상에서 패치기반 CNN 모형을 이용한 잡음제거", 한국데이터정보과학회지, 30(2), pp.349~363 (2019), 한국데이터정보과학회[DOI]
- 조영현, "PET", MR 영상에서의 팬텀을 활용한 노이즈 감소를 위한 변형된 중간값 위너필터의 적용 효율성 연구, 2021, pp.44(3), 225~229[DOI]
- Krull A, "Noise2void-learning denoising from single noisy images", Proceedings of the IEEE Conference on Computer Vision and Pattern Rcognition(CVPR) (2019)
- Leonard J, "Improvement of the backpropagation algorithm for training neural networks", Computer & Chemical engineering, 14(3), pp.337~341 (1990)
- Anwar SM, "Real image denoising with feature attention", Processing of the IEEE International Conference on Computer Vision(ICCV) (2019)
- Lehtinen J, "Noise2Nois : Learning image restoration without clean data", International Conference on Machine Learning(ICML), 80, pp.2965~2974 (2018)
- Aide N, "EANM", EARL harmonization strategies in PET quantification : From daily practice to multicenter oncological studies, 2017, pp.44, 17~31
- Ronneberger O, "U-net:Convolutional networks for biomedical image segmentation", MIC-C, pp.234~241 (2015), Springer
- Zhao H, "Loss function for Image restoration with neural networks", IEEE Transactions on Computational Imaging, 3(1), pp.47~57 (2017)
- Hofheinz F, "Suitability of bilateral filtering for edge-preserving noise reduction in PET", European Journal of Nuclear Medial Molecular Imaging Research, 1, pp.23~ (2011)
- Chen KT, "Ultra low dose 18F-flor-betaben amyloid PET imaging using deep learning with multi contrast MRI inputs", Radiology, 290(3), pp.649~656 (2019)
- Lu W, "An investigation of quantitative accuracy for deep learning based denosing in oncological PET", Physics in Medicine Biology, (64), pp.165019~ (2019)
- Jeong YJ, "Restoration of amyloid PET images obtained with short time data using a generative adversarial networks framework", Scientific Reports, 11, pp.4825~ (2021)
- 김영재, "의료 영상에 최적화된 딥러닝 모델의 개발", 대한영상의학회지, 81(6), pp.1274~1289 (2020), 대한영상의학회[DOI]