Algorithms to process data from the MARS molecular imaging system

Abstract

The MARS molecular imaging project aims to research, develop, and commercialize a spectral computed tomography (CT) system. My thesis describes the work I performed to reconstruct volumes from the MARS prototypes. The challenge was to develop the algorithms while maintaining image processing software that met the immediate needs of the MARS team.

Using the Medipix detector, the current prototype is capable of simultaneously scanning up to 8 energy bins. Every additional energy bin improves the potential for material discrimination at the molecular level. Data acquired from the MARS prototypes are a collection of exposures over various geometric transformations of the source, detector, and subject. To process these, I developed two applications, mPPC (MARS Preprocessing Chain) and mART (MARS Algebraic Reconstruction). The application mPPC prepares data for reconstruction while also improving the image quality. In particular, various issues that result from the Medipix detector are addressed in the preprocessing software. The application mART reconstructs the preprocessed data into volumes. It adopts a variation of SART to simultaneously reconstruct all the energy bins. The results are a good balance between quality and performance. To link the components of the data processing chain, all the software adopts the stable and popular DICOM standard. The DICOM standard provides formats to package the data while also providing protocols for both storing (and backup) and transferring the data. To summarize, the outcomes of my thesis are two applications which, together, perform all the necessary steps to reconstruct high quality volumes from the MARS system. With the addition of the DICOM standard to store and transfer the data, the result is a data processing chain which fulfils the needs of the MARS team.