(Peer-Reviewed) Interpretable low-dose CT enhancement via multi-Gaussian cluster variance reduction
Xiaofeng Zhang 张晓峰 ¹, Yilan Zhu 朱伊蓝 ¹, Yongsheng Huang 黄永盛 ¹, Jielong Yang 杨杰龙 ², Zhili Wang 王志立 ³, Kai Zhang 张凯 ⁴, Si Chen 陈思 ⁵, Linbo Liu 刘琳波 ⁶, Xin Ge 葛昕 ¹
¹ School of Science, Shenzhen Campus of Sun Yat-sen University, Shenzhen 518107, China
中国 深圳 中山大学深圳校区 理学院
² College of Information Science and Engineering, Northeastern University, Shenyang 110819, China
中国 沈阳 东北大学信息科学与工程学院
³ School of Physics, Hefei University of Technology, Hefei 230009, China
中国 合肥 合肥工业大学物理学院
⁴ Beijing Synchrotron Radiation Facility, Institute of High Energy Physics, Beijing 100049, China
中国 北京 中国科学院高能物理研究所 北京同步辐射装置
⁵ Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore 117597, Singapore
⁶ Guangzhou National Laboratory, No. 9 XingDaoHuanBei Road, Guangzhou International Bio Island, Guangzhou 510005, China
中国 广州 广州国家实验室
Opto-Electronic Science
, 2026-03-25
Abstract
Computed tomography (CT) is indispensable in both clinical medicine and biological research, yet reducing radiation exposure while maintaining image quality remains a big challenge. To address this, we propose multi-Gaussian Cluster Variance Reduction (mGCVR), a method that enables low-dose CT images to approximate the quality of high-dose scans. mGCVR models the heterogeneous tissue CT intensity distribution using multiple Gaussian components, and performs denoising by shrinking the variance within each component.
In biological imaging experiments, mGCVR consistently improves image quality across the entire field of view. Compared with classical denoising algorithms, mGCVR produces images that more closely resemble high-dose clinical CT images and achieves superior performance in quantitative metrics. These results validate the effectiveness of mGCVR and highlight its potential for broad use in both medical imaging and scientific applications.
Holotomography-driven learning unlocks in-silico staining of single cells in flow cytometry by avoiding fluorescence co-registration
Daniele Pirone, Giusy Giugliano, Michela Schiavo, Annalaura Montella, Martina Mugnano, Vincenza Cerbone, Maddalena Raia, Giulia Scalia Ivana Kurelac, Diego Luis Medina, Lisa Miccio Mario Capasso, Achille Iolascon, Pasquale Memmolo, Pietro Ferraro
Opto-Electronic Science
2026-02-25
A hybrid integrated high-precision tunable semiconductor laser
Yiran Zhu, Botao Fu, Zhiwei Fang, Qiyue Hu, Jianping Yu, Yunpeng Song, Yu Ma, Min Wang, Kunpeng Jia, Zhenda Xie, Ya Cheng
Opto-Electronic Advances
2026-02-12
Millisecond-level electrically switchable metalens for adaptive rotational depth mapping and diffraction-limited imaging
Yeseul Kim, Jihae Lee, Won-Sik Kim, Hyeonsu Heo, Dongmin Jeon, Beomha Yang, Xiaotong Li, Harit Keawmuang, Shiqi Hu, Young-Ki Kim, Trevon Badloe, Junsuk Rho
Opto-Electronic Advances
2026-02-12
