The winning paper describes an iterative algorithm for image reconstruction in circular cone-beam CT (CBCT), the most popular configuration for X-ray volumetric imaging. Circular CBCT is currently the industry standard approach for a range of CT applications, including dedicated breast CT, micro-CT, image-guided radiotherapy and C-arm systems employed to guide surgery.

The most widely used algorithm for obtaining the reconstructed volume from a circular cone-beam scan is the FDK derivative of the filtered back-projection algorithm. However, this algorithm is limited by the need for dense projection sampling and a large scanning arc. To reduce radiation exposure to patients, researchers are keen to develop image reconstruction algorithms that operate with fewer projections, less of a scanning arc or lower X-ray intensity per view.

"Our article demonstrated an algorithm with the potential to perform accurate image reconstruction with far fewer CBCT projections than are needed when FDK is employed," explained Sidky and Pan. "Potentially, a factor of ten fewer views needs to be acquired."

Iterative approach

The algorithm, referred to as ASD-POCS (adaptive-steepest-descent-projection onto convex sets), uses iterative image reconstruction (IIR) and was part of a larger effort to incorporate IIR into CT. This target was initially hampered by the fact that CT acquires far more projection data than other tomographic modalities. IIR algorithms are also more computationally demanding than FDK algorithms.

Sidky and Pan explained that the ASD-POCS algorithm benefited from three factors: existing effort and experience in iterative image reconstruction; the development of fast, low-cost computational resources such as graphics processing units; and a demonstration of undersampled image reconstruction on an idealized MRI model, which spawned the field of compressive sensing.

Compressive sensing attempts to relate image sampling conditions to some identified sparsity in the subject being imaged. If the subject can be 'sparsified', it may be possible to accurately reconstruct the image with a large reduction in sampling. "This idea resonated with those of us in CT research interested in projection view reduction," the authors explained.

They noted that ASD-POCS represented the culmination of an effort to adapt sparse-sampling image reconstruction to circular CBCT. "This led to a rich line of image reconstruction research for CT sampling reduction based on subject sparsity, and hence the large volume of follow up research and citations," they explained.

Since publication of the research paper, which demonstrated ASD-POCS on simulated CBCT data, the authors have worked to adapt it to various applications using actual CT scanner data. "We have been emphasizing applications with reduced projection sampling, but other researchers have adapted the algorithm to aid in accurate reconstruction for time-resolved imaging of the heart and lungs," explained Sidky and Pan.

Rotblat medal

The PMB citations prize is marked with the presentation of the Rotblat medal, named in honour of Professor Sir Joseph Rotblat (PMB's second and longest-serving editor). Sidky and Pan say that they feel both honoured and humbled to receive this prize. "We did not expect that the number of citations would be so high, because image reconstruction is only one of many important components of a complete tomographic imaging system," they told medicalphysicsweb.

"In that respect, we would like to thank many collaborators who have helped us demonstrate this algorithm on actual CT scanner data. The results obtained by image reconstruction algorithms can only be as good as data quality from the CT systems. Our research would not be possible without our collaborators' high-quality CT systems and data."

• The winner of the 2012 Physics in Medicine & Biology citations prize is: Image reconstruction in circular cone-beam computed tomography by constrained, total-variation minimization Phys. Med. Biol. 53 4777.