Dr Bogdan Roman

I'm a Senior Researcher in the Centre for Mathematical Imaging in Healthcare at the Pure Maths department, a Research Fellow at the Computer Laboratory department, and Director of Studies for Computer Science at Queens' College. I also chair the Computer Science Admissions Test (CSAT), which I initially designed.

Office: F2.05



  • Our proposal "IMPROVED SPATIAL RESOLUTION AND TARGETED SAMPLING IN MRI" by Bogdan Roman, Martin J Graves, Anders Hansen, David J Lomas was selected by the University after an internal competition as the University's nomination for the prestigious Rosetrees Interdisciplinary Prize 2016 (250,000 GBP). This is work done in collaboration with the Radiology department and Cambridge University Hospitals on MRI, using General Electric 1.5T and 3T scanners. It showcases visible resolution improvement and shorter acquisition time simultaneously in a clinical setting. This can impact both clinical and research applications, allowing for improved morphological and functional imaging with the potential for earlier and improved diagnosis and outcomes. The work extends the practical validation done by Siemens in 2015 of our compressed sensing research.

    The Rosetrees Trust was established in 1987 to fund life-changing medical research. The Rosetrees Interdisciplinary Prize is an annual award for innovative and impactful interdisciplinary research on biomedical problems.

  • Siemens validated in practice, using a modified MRI machine, the asymptotic sparsity, asymptotic incoherence and high resolution concepts introduced by our work (see Breaking the coherence barrier: A new theory for compressed sensing and also On asymptotic structure in compressed sensing). Their ISMRM paper can be found here. Their results and conclusion states:

    “The image resolution has been greatly improved [...]. Current results practically demonstrated that it is possible to break the coherence barrier by increasing the spatial resolution in MR acquisitions. This likewise implies that the full potential of the compressed sensing is unleashed only if asymptotic sparsity and asymptotic incoherence is achieved. ”

Research, Teaching

Co-lecture the Part III course on Sampling and Compressed Sensing. See my CL homepage for my teaching/supervising for compsci/engineering.

Interested in compressed sensing, signal processing, sampling theory, inverse problems, computational mathematics. See my CL homepage for my research on wireless comms.

Admissions, Exercises

I welcome alternative solutions, extensions or new questions for any of the above.


  • (Very) Fast C++ MEX Hadamard (Walsh-Hadamard) transform with support for sequency and natural ordering, multi-core/multi-cpu and also complex values. Orders of magnitude faster than Matlab's fwht() function.
    Download v1.3, Oct 2014. Includes 64bit binaries for Linux, MacOSX and Windows, and Matlab wrappers for 1D and 2D transforms, unitary and non-unitary.
  • Compiling MEX files directly from Matlab seamlessly under Windows using GCC (MinGW64). This should normally work with most C/C++ MEX files that work under Linux. You also won't need to carry or distribute the std runtime .DLLs from MingW/Cygwin.
    Download this mexopts.bat file and place it in %USERPROFILE%\AppData\Roaming\Mathworks\MATLAB\R2014a\. See more instructions and details inside the downloaded mexopts.bat file (see also this StackExchange answer).
  • MacOH. Automated tool for stress testing Mac machines to reveal throttling or overheating. It downloads all needed tools, starts benchmarks, monitors CPU temperature and frequency, and plots them versus time (sample output). First created as a personal test tool. It can do CPU and GPU tests (Prime95, x264, GpuTest, user defined).


PhD Thesis:
B. Roman, Scalable Cross-Layer Wireless Medium Access Control, University of Cambridge, Computer Laboratory, 2011