Poste Offre de thèse : Caractérisation de la tomodensitométrieà faible dose par comptage de photons / Characterization of low dose photon counting X-ray CT

Offre de thèse : Caractérisation de la tomodensitométrie à faible dose par comptage de photons / Characterization of low dose photon counting X-ray CT These DeadLine: 01/12/2009 morel@cppm.in2p3.fr http://imxgam.in2p3.fr Project description : Given its fairly low cost, X-ray micro-Computed Tomography (CT) represents a cost-effective alternative to Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) for many biomedical applications in fundamental research. However, progresses are still expected, particularly for soft tissue imaging where detectors with low noise and large dynamics are needed. Furthermore, it is necessary to reduce as much as possible the dose absorbed by the animal, especially when longitudinal studies are envisaged. Hybrid pixel detectors developed for High Energy Physics1 are pixelised photon counting detectors that have the following new features as compared with charge integration detectors: • Background noise suppression by photon energy triggering; • Very large X-ray flux dynamics, no background, ability to record very low X-ray flux from 0.01 to 106 photons/pixel/sec; • Virtually infinite integrated X-ray flux (> 1032), no background and continuous data readout during photon counting; • Energy selection, one threshold per pixel, but possibility to select several energies simultaneously by forming clusters of pixels. Pixels are formed by a semiconductor sensor that is segmented in small square pixels of 50 to 300 μm, on which integrated circuits containing dedicated electronics are bump bounded. The sensor is a high purity semi-conductor (Si or CdTe) with implanted diodes on one side and a resistive contact on the other side. Direct conversion of the Xray photons in the sensor prevents the spread of the energy deposited by the X-rays. During photon conversion in a scintillation crystal, the scintillation light is scattered in the neighbouring pixels. Bundles of optical fibres or collimators are used to solve partially this drawback. Direct conversion in a sensor where electric field is perpendicular to the surface of the pixels avoid this problem. The choice of the sensor allows for using a material with a density and a thickness appropriate for the energy range of the X-rays, thus maximizing their absorption for a detection efficiency of about 100% that results in a dose reduction (by at least a factor 10) and a gain in acquisition duration. Last but not least, the electronic shutter allows for gating a given period of time with a precision of the order of the microsecond in order to select a period of interest (e.g. synchronised with cardiac and/or respiratory movements) and to follow the evolution of a phenomenon. Thus, the use of hybrid pixel detectors in X-ray CT seems to allow for improving soft tissue contrast and for reducing the dose. Effectively, the integrated intensity dynamics, i.e. the number of gray levels, can be as large as we want. Actually, it depends only on the counting time (the noise is negligible and the counters are read out during acquisition). There are no charge migrations into neighbouring pixels, nor phosphorescence, because the X-ray photons are directly converted in electric charges in the sensor (no intermediate conversion in low energy photons like with scintillation). The energy selection can be easily performed by the independent setting of every pixel threshold. One can even target a photoelectric absorption ray by acquiring simultaneously energy ranges below and above the absorption drop of a contrast agent (like iodine) and subtracting corresponding statistics (in real time): the image of the labelled organ will appear with a different “colour”. Finally, image acquisition is extremely fast, because pixels are counting independently one another (up to one million photon counts per second). Parallelism is thus maximum. Repetition time between two images, essentially due to data readout, is faster than 2 ms without losses. These features should then allow for significant progress in low dose X-ray CT, which is a prerequisite to the observation of small organ deformations in transgenic mice for instance. In order to study the added value of hybrid pixel detectors in this field, the imXgam team at CPPM has developed a prototype demonstrator of an X-ray micro-CT scanner for small animals named PIXSCAN (Fig. 2).2 A first version of the demonstrator was equipped with the XPAD2 detector developed at CPPM with pixels of 330 x 330 μm2. To make it simpler, it was decided to have the Xray source and the detector at rest and let the mouse rotate. Figure 3 shows the result of an angiography using an iodine contrast agent. A second version of the PIXSCAN demonstrator is presently under construction and will be equipped with the XPAD3 detector 3 formed by pixels of 130 x 130 μm2. The objective of the thesis is to carry out a complete study of the performance of a hybrid pixel micro-CT scanner as a function of the dose absorbed by the object. A comprehensive comparison with a charge integration X-ray detector (of CCD and/or CMOS pixel type) using exactly the same experimental setup (same source, identical geometry, same scan duration, same dose monitors) will allow for assessing the use of hybrid pixels in the field of micro-CT and for extrapolating their use for clinical X-ray CT. This necessitates a deep characterization of the demonstrator, the optimisation of its performance, the study of contrast enhancement methods and iterative image reconstruction techniques adapted for low statistics. The ability to reduce the dose thanks the use of a bi-dimensional detector operating at low X-ray flux with an efficiency of about 100% will determine the future of hybrid pixel technology in the field of precilinical and clinical imaging. The work will be based on direct measurements and on simulated models. The candidate will develop skills in instrumentation, image treatment, image reconstruction and Monte Carlo simulation. Prerequisite knowledge and skills: Physics or biomedical engineering, skills in image processing / tomographic reconstruction, knowledge of object oriented programming wished.