Sparse-sampling approach to low-dose CT

As a viable option to low-dose CT imaging, we have proposed a novel method called many-view under-sampling (MVUS) that can provide a sparse data sampling without switching the tube power (Opt. Eng. 2012, US Patent Application No. 14449587). A multi-slit collimator is placed in between a tube and a patient, and it efficiently reduces the radiation dose to the patient. To increase the uniformity of sampling density, we proposed to reciprocate the collimator during a scan. To optimize a sparse sampling scheme under a given constraint of radiation dose, we recruited two metrics that stem from the compressive sensing theory: sampling density and data correlation (Med. Phys. 2013). With a given sparsity of the imaged object, enough sampling has to be secured with a minimal correlation. We have analyzed these metrics in various sparse-sampling schemes, and found their usefulness in optimizing a sparse-sampling scheme (Optics Express 2014). An experimental study has shown promising results, and additional benefits such as scatter reduction and a single-scan dual-energy imaging have also been demonstrated (IEEE Trans. Nucl. Sci., 2016; 2017, IEEE Trans. Med. Imag. 2017). We are currently working on implementation of the technique in clinical CT systems such as in-room CT and C-arm CT.