Daniel Niederlöhner

Quantitative Material Reconstruction in CT with Spectroscopic X-ray Pixel Detectors -- a Simulation Study

Modern photon counting X-ray detectors offer energy resolving capabilities for every single photon interacting in the sensor layer. This provides a new dimension of information which could and should be used to improve image quality. As the energy dependance of X-ray attenuation is a characteristic for the material, one can gain information on the composition of the object by analysis of the attenuation as a function of energy. One possibility is the material reconstruction, where the areal densities of chosen basis materials are obtained in projective geometry. For instance, this method could be used not only to distinguish and identify contrast agent from other highly absorbing regions, but also to gain quantitative information on the specific material. We have successfully applied this technique to computed tomography with Monte-Carlo simulated data and investigated the properties of this approach. Material selective images are presented and image quality properties are evaluated. The Signal Difference to Noise Ratio (SDNR) in the material-reconstructed and in the conventional CT images are compared.

Using the Medipix2 detector for photon counting computed tomography

The Medipix2 detector is a photon counting hybrid pixel detector. It consists of 256 by 256 pixels (55/spl times/55 /spl mu/m/sup 2/ each) with a semiconductor sensor layer. Two thresholds enable X-ray imaging with an energy window. Integration of this detector into our setup requires very precise alignment of all devices: Precision of up to 0.05/spl deg/ and 10 /spl mu/m was realised with special objects. For the verification of the system resolution we constructed phantoms with high contrast at high spatial frequency. The resolution limit of our setup is about 50 /spl mu/m which is dominated by the relatively large X-ray spot size. Further on we investigated the impact of flatfield correction on ring artifacts and additionally applied a special filtering algorithm for suppression of those. Computed tomography acquisition series with different organic objects have been carried out to prove the system performance for realistic imaging tasks: Using a peanut as well as a newborn mouse produced high quality volume data sets. In addition, we used the quad version of the Medipix2 detector (2/spl times/2 ASICS) for CT of larger objects with diameters up to 26 mm. With Monte-Carlo simulations we investigated the interaction of energy weighting with the filtered back projection algorithm if energy resolved data is available: Applying the energy weighting technique before filtering and back projection or afterwards; or with intermixed order for different energy intervals and different kernels. We propose weighting before filtering and recommend smoothing kernels for lower photon energies and high resolution kernels for energy intervals with higher statistics.

Threshold characterisation of the Medipix1 chip

Abstract The Medipix1 chip is a hybrid pixel detector working in photon counting mode and has the ability to put a discriminating threshold on each photon signal. To increase the image quality the threshold voltage can be electronically fine tuned with a 3 bit threshold adjust for each pixel. This electronic adjustment can only equalise the inhomogeneities of the electronics, not those of the physical frontend (conversion material and bump bonds). The remaining noise of the threshold has an obvious impact on the image quality. To optimise the uniformity over the whole chip we did threshold scans with a cadmium X-ray source analysing every pixel. With those scans we were able to create a bit mask including the physical frontend. Comparing the results of the different methods we can judge the quality of the electronically generated mask and draw conclusions about the properties of the physical frontend. Furthermore, simulations have been carried out for comparison with the measured data.