LAUSR

Spectral images from photon counting spectral detectors are being explored for material decomposition applications such as for obtaining quantitative maps of tissue types and contrast agents. While these detectors allow acquisition of multi-energy data in a single exposure, separating the total photon counts into multiple energy bins can lead to issues of count starvation and increased quantum noise in resultant maps. Furthermore, the complex decomposition problem is often solved in a single inversion step making it dicult to separate materials with close properties. We propose a multi-step decomposition method which allows solving the problem in multiple steps using the same spectral data collected in a single exposure. During each step, quantitative accuracy of a single material is under focus and one can exibly optimize the bins chosen in that step. The result thus obtained becomes part of the input data for the next step in the multi-step process. This makes the problem less ill-conditioned and allows better quantitation of more challenging materials within the object. In comparison to a conventional single-step method, we show excellent quantitative accuracy for decomposing up to six materials involving a mix of soft tissue types and contrast agents in micro-CT sized digital phantoms.