Resolution effects in a dense linear source array X-ray micro-tomograph

Abstract

We study reconstruction methods and resolution effects in a proposed multi-source X-ray micro-computed tomography (CT) system. The proposed device is based on dense arrays of microfabricated field-emission X-ray sources which are individually addressable. The proposed system has two 10 cm linear arrays of X-ray sources, each with 50 sources spaced at 2 mm. The two source arrays form two contiguous sides of a square, and two 10 cm area detectors form the other two sides of the square. We estimate that this results in an effective field of view of a 3.2 cm diameter cylinder, suitable for a mouse. No motion of source or subject is needed; simply flashing the individual sources creates angular sampling for tomography, though it is a limited-angle problem. We demonstrate implementations of two iterative reconstruction techniques, block-iterative transmission (BIT) and ordered-subsets convex (OSC). Using data simulated from a realistic mouse phantom, we study the reconstruction quantitative accuracy, point spread functions and noise responses at several locations in the field of view. We demonstrate that the reconstructions are quantitatively accurate and mostly free from disturbing artifacts, while the point spread functions in this geometry are anisotropic and spatially-varying. The OSC algorithm converges faster, showing better resolution yet worse noise performance than BIT after five iterations. We conclude that the proposed geometry is viable for fast, high-resolution and motion-free imaging