Metrology of Multiview energy dispersive X-ray CT instruments

X-ray Computed Tomography (CT) is an imaging technique expanding from medical applications into other industries, such as manufacturing and aviation security. This PhD thesis focuses on performing measurements and evaluating measurement quantities using X-ray CT instruments. It is the first extensive study evaluating dimensional measurement using an X-ray CT instrument that utilises energy-dispersive Photon-Counting Detectors (PCDs). The project considers three research questions regarding the variations of the measurements, adjustment of the instrument’s geometry, and variations in the energy domain. The work was performed in collaboration with the company Exruptive, which develops instruments for scanning cabin luggage in aviation security. Exruptive’s scanner uses a new large fixed-gantry design of Energy-Dispersive X-ray CT (EDXCT) to scan and analyse a full-sized luggage trolley. EDXCT instruments with a fixed-gantry design and PCDs have not been implemented in an industrial setting or studied in detail prior to this work.
A new Forward Ray Tracing Instrument (FRTI) was presented for modelling X-ray CT instruments. The FRTI takes the scanner’s geometry, a representation of the sample volume, and the instrument’s operation settings as input and simulates radiographs to be combined into sinograms for 3D reconstruction. This work investigates several of the effects of input parameters on the reconstructed volume and examines methods to evaluate measurement quantities. Simulations of standards relevant to aviation security were used to determine that 107 rays were needed to achieve a sufficient contrast. Measurement of the simulation time showed that simulations of a CT instrument with 107 rays within a reasonable time of a few days would require GPU paralisation. The FRTI was used to simulate CT data for dimensional measurement and to test new designs and configurations.
Dimensional measurements were performed of the three material measures: the CT Tube, the CT Ball Plate, and the new CT Ball Bar. Preliminary measurements of the CT Tube showed a measurement error within ±0.5 mm. This was later repeated in a more expensive study of both the CT Tube and the CT Ball Plate with measurement errors of the two material measures within ±0.5 mm and ±0.75 mm, respectively. A study was performed using the CT Ball Bar placed, acquired, and reconstructed using various configurations and settings acquired in both Exruptive’s test and industrial scanners. The study used an analysis method inspired by a design of experiment approach to investigate the interactions between the different factors. The study showed the strongest factors were the position and orientation of the CT Ball Bar material measure. The test scanner’s Interface Structural Resolution (ISR) was investigated using an hourglass model. First, the hourglass model was tested using a metrological X-ray micro CT instrument at the University of Padova (Italy). This showed a dependence on the orientation and positioning of the hourglass material measure. Afterwards, the IST was evaluated for the test scanner. This analysis showed difficulty in evaluating the numerical value of the ISR without a subvoxel surface determination method. These studies aided in answering the first research question.
The fixed-gantry design with small PCDs requires a large number of detector modules to cover the full field of view, resulting in a complex gantry with a large number of degrees of freedom. A conceptual and mathematical framework was developed for designing a tool to adjust the scanner geometry. The new geometry adjustment tool was designed to work for in-line scanners and measure the geometry by casting a series of sphere shadows onto the array of detector modules. The design was optimised to reduce the effect of geometrical offsets and measurement noise. Due to time constraints, the tool was not built, but technical drawings and a measurement strategy were proposed for future work. This work supported the second question.
Finally, the variation in the energy domain was investigated. The energy variation of the emitted X-ray beam and the reconstructed voxels were investigated. The results show the beam’s relative intensity only decreasing slightly (<5%) over an ±10° opening angle, at a maximum of 15% at the peak energy. Studying the reconstructed voxels showed a relatively low-intensity variation in the smaller region excluding the edges of the volume in the energy range E ≃ [40 keV; 120 keV]. Dimensional measurements of the CT Ball Bar showed a variation in the measurement error with an amplitude comparable to other strong influence factors of around 0.4 mm. Implementing a dynamic threshold segmentation method reduced the variation to around ±0.1 mm. This answered the third research question by showing variations in the energy domain, but also how to minimise these variations.