Research Profile of Ming Jiang

I lead a research group (~15 graduate students, post-docs, and faculty members) focused on mathematical innovations in biomedical imaging and image processing, with x-ray computed tomography, optical tomography and multi-modality biomedical imaging as the major applications. Our group also collaborates with colleagues at other institutions and with industry to develop the theory, algorithms, software and hardware for biomedical applications. I have been the PI/MPI/Co-PI of large-scale projects funded by the National Science Foundation of China, Ministry of Science and Technology of China, Ministry of Education of China, and Sino-German Center. The total grant support is 14.998 million in RMB as PI/MPI and 11.36 million in RMB as Co-PI. My publications profile can be found at ResearchID or Google Scholars.

Selected Contributions

My most significant contributions are as follows.

[Iterative Image Reconstruction Algorithms] With the increasing complexity of imaging modalities, iterative reconstruction methods become more and more useful because a closed form solution is hardly available. With collaborators, we proved the convergence of SART (simultaneous algebraic reconstructive technique), which is a most popular iterative image reconstruction algorithm, and established its dependence on initial value. This work was extended by us to a unified framework for block-iterative Landweber algorithms. My most representative publications in this work are

·         Ming Jiang, Ge Wang, Convergence of the simultaneous algebraic reconstruction technique (SART), IEEE Transactions on Imaging Processing, 2003.

·         Ming Jiang, Ge Wang, Convergence studies on iterative algorithms for image reconstruction, IEEE Transactions on Medical Imaging, 2003.

This work is still considered to be an important work and a good thing by leading experts after 10 years.

[Bioluminescence Tomography] Bioluminescence tomography (BLT) is an optical tomography technique developed since 2004 to image in vivo 3D distributions of bioluminescent probes for preclinical molecular imaging of small animals. With collaborators, we published the first journal paper on the theoretical aspects of BLT. There are ~20 groups in this area, and many bioluminescence imagers for studies on animal models of almost all human diseases. My most representative publication in this work is

·         Ge Wang, Yi Li, Ming Jiang, Uniqueness theorems in bioluminescence tomography, Medical Physics, 2004,

which is reviewed in Nature Biotechnology, and Inverse Problems by leading experts.

[Regularization Techniques] Imaging and image processing problems are typical ill-posed inverse problems, for which the regularization technique with priors or other regularization approaches within the general Bayesian inference are necessary. With collaborators, we developed the theory and algorithms of higher order total variations, which is a non-trivial extension of the widely used total variations regularization, and has been applied successfully to interior tomography of x-ray CT and SPECT.   Our recent work is the regularization properties of the Mumford-Shah functional, which can used to simultaneously reconstruct image and its segmentation.  The edge prior can help improve the reconstructed image quality, which is missed in other conventional priors. My most representative publication in this work is

·         Ming Jiang, Peter Maaß, Thomas Page, Regularizing properties of the Mumford-Shah functional for imaging applications, Inverse Problems, 2014.

Our work are selected to the Highlights Collections of 2010, 2012 and 2014 of Inverse Problems.

Research Projects

With a solid background in mathematics and professional practice in real-world imaging applications, it is my penchant to contribute to the imaging techniques and image processing problems from their mathematical perspectives. In this interdisciplinary world, I am honored to work with colleagues from physics, computing technology, biology and clinics to develop implementations with software and hardware. There are three directions we are currently pursuing. 

[Theoretical Thrust] The mathematical aspects of imaging techniques and image processing problems is always our interests. Our recent focus are as follows.

·         Phase-contrast imaging and tomography. With collaborators, we have established the mathematical theory for 2D-grating based phase-contrast imaging with x-ray and neutron. Our recent refinement is the partial coherence theory for phase-contrast imaging. These results can be applied to optimize the system design for maximal fringe visibility or resolution. We are working on the tomographic problems of this phase-contrast imaging technique with x-ray and neutron sources by collaborations with Profs. Seung Wook Lee (Pusan University) and Alfred K Louis (Saarland University).

·         Multi-modality imaging. Images from different modalities can be quite different in their contents, but possess similar features. Our hypothesis is that the feature information from one modality can improve and steer the reconstruction of another modality, and vice versa, in an iterative manner, with an appropriately defined feature and its similarity measure. Edge is a fundamental feature of image. The regularization by Mumford-Shah functional provides the reconstruction of edge in addition to image. We are recently funded by the Sino-German Center to conduct a joint project entitled “Feature based bi-modal image reconstruction”, in collaboration with Profs. Alfred K Louis (Saarland University), Peter Maaß (University of Breman) and Xiaoqun Zhang (Shanghai Jiao Tong University). Our aim is to establish the simultaneous reconstruction theory and algorithms for x-ary CT/DOT and x-ray CT/BLT in this project. 

·         Iterative image reconstruction algorithms. With collaborators, we have proved the necessary and sufficient convergent conditions, and the convergence conditions within finite iterations for the simultaneous Landweber method. We have established the iterative algorithms for the regularization by the higher order total variations and Mumford-Shah functional.  Our recent refinement is an asynchronous parallel iterative reconstruction algorithm for the regularization of Mumford-Shah functional.

[Implementation Development] We are working on the energy-efficient hardware implementation with FPGA for our asynchronous parallel iterative reconstruction algorithm for the regularization of Mumford-Shah functional. Our target application is the low dose x-ray CT in clinics and cone-tile electron transmission tomography.  Our recent implementation by the high-level synthesis tool from UCLA demonstrates the same performance as GPU with a much lower energy consumption. We are refining the implementation to deliver it to clinical evaluation this year. This work is supported by the National Science of Foundation of China as one key international collaboration project with Profs. Jason Cong, Nicholas Brecha, Alex Bui, William Hsu, Luminita Vese, from UCLA. Another implementation task in this project is for the super-resolution reconstruction from low-resolution video to future ultra-high definition video by our regularization techniques and algorithms with colleagues from the National Engineering Laboratory for Video Technology of China.

[Practical Evaluation] With our collaborators, we have the access to x-ray CT scanners, neutron imaging systems, and electron microscopes to get real data to evaluate the performance of our implementations.  We have own system for x-ray and optical experiments, which is based on Kodak Image Station In-Vivo FX system (Carestream Health, Inc., Rochester, NY). The original system has an X-ray module with cone-beam configuration only for radiographic imaging but lacks the functionality of tomography.  We have improved the system by mounting a 2-axis motorized scanning system and other optical components and developed the control software so that it can perform cone-beam x-ray CT with a two -circle-plus line geometry and optical tomography including DOT and BLT. Now it serves for us for teaching and preliminary evaluation for x-ray CT and optical tomography. The following diagram is the current system after mounting a 2-axis motorized scanning system in the chamber and other optical components (left).  X-ray cone-beam tomography is conducted by rotating an object along the first vertical circle, followed by a line, and then another vertical circle with a two-circle-plus line geometry (right). DOT is conducted by shedding a laser beam from a laser diode outside the chamber onto the object while the object is rotated by the motorized system (right). Both measurements for XCT and DOT are by the CCD in the original system. The scanning parameters are adjustable by our control program. Please refer to the following paper for more information.

·         Yanbin Lu, Jiansheng Yang, John W Emerson, Heng Mao, Tie Zhou, Yuanzheng Si, Ming Jiang, Cone-beam reconstruction for the two-circles-plus-one-line trajectory. Physics in Medicine and Biology, 2012.

Text Box:


Click for a high resolution diagram