Recent Advances of Deep Learning-based Approaches in Computed Tomography

Deep learning based methods have been applied to the image reconstruction field, leading a new field named “deep reconstruction”. This has become the mainstream in the imaging field. Unlike the compressed sensing based methods that totally rely on an accurate imaging models, deep reconstruction is empowered by big data with which a deep network can be trained for superior image quality. In other words, the prior knowledge can be automatically learned/extracted from big data. Unlike natural images, clinical data is limited for training deep reconstruction networks. To overcome this challenge, in this talk we will report two of our recent works on weakly supervised learning and generative models. The first is our low-level vision masked autoencoders for low-dose CT denoising (IEEE Journal of Biomedical and Health Informatics, 28(11):6815-6827), and the second is the physics-informed score-based diffusion model for limited-angle reconstruction of cardiac CT (IEEE Transactions on Medical Imaging, accepted, https://ieeexplore.ieee.org/document/10747553).

Hengyong Yu is a Full Professor and Director of the Imaging and Informatics Lab, Department of Electrical and Computer Engineering, University of Massachusetts Lowell, Lowell, MA, 01854. He received his Bachelor’s degrees in information science & technology (1998) and computational mathematics (1998) respectively, and his PhD degree in information & communication engineering (2003) from Xi’an Jiaotong University. His interests include computed tomography and medical image processing. He has authored/coauthored >220 peer-reviewed journal papers with an H-index of 50 according to Google Scholar. He was the founding Editor-in-Chief of JSM Biomedical Imaging Data Papers, serves as an Editorial Board member or associate editor for IEEE Transactions on Medical Imaging, Medical Physics, IEEE Access, Signal Processing, etc. He is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE), American Association of Physicists in Medicine (AAPM), American Institute of Medial and Biological Engineering (AIMBE), AAIA (Asia-Pacific Artificial Intelligence Association), and International Artificial Intelligence Industry Alliance (AAIA), and a member of Biomedical Engineering Society (BMES), American Association for the Advancement of Science (AAAS), and the international society for optics and photonics (SPIE). In 2005, he was honored for an outstanding doctoral dissertation by Xi’an Jiaotong University, and received the first prize for a best natural science paper from the Association of Science & Technology of Zhejiang Province. In January 2012, he received an NSF CAREER award for development of CS-based interior tomography. In September 2022, he received the IEEE R1 Technological Innovation Award (Academic) for “pioneering contributions and international leadership in tomographic imaging, especially interior tomography and machine learning-based tomographic imaging”.