Abstract
This thesis reports on the characterisation of two modules of the MARS spectral CT system: the x-ray source and the x-ray camera. Understanding each of them directly improves material identification and quantification. A parameterised semi-analytical x-ray source model was developed. This work led to five published conference proceedings. Imaging performance of the CdTe-Medipix3RX detector was characterised. This research work gave rise to two provisional patents: an algorithm for unstable pixel identification; and a method for using unstable pixel clusters reliably; and in addition two published conference proceedings. As a part of the MARS development team, my research work has been and will continue to be used within the group for improvement in the MARS spectral capabilities.
The primary feature sought from the parameterised semi-analytical x-ray source model is off-axis spectral variation. This aspect is overlooked in existing x-ray source models but is helpful in spectral CT. In addition, the models which do provide off-axis information do not match well with the specification of the x-ray tubes used in MARS spectral systems. This is an important aspect to consider in spectral CT since spectral detectors are sensitive to the polychromatic structure of the x-ray beams. Incorporating such variation in spectral image processing can improve the image quality. Our model provides off-axis information within ±17◦ along vertical direction (θ) and ±5◦ along horizontal direction (φ) of the central axis in the diagnostic imaging energy range (30-120 kVp). Comparisons of our model with existing models at central axis show good agreement (within 4%). In addition, the off-axis comparison of the model with experimental data collected with the MARS scanner was consistent within 3%.
Temporal stability of the detector with different bias voltage settings was analysed by investigating variations in count with time. These investigations revealed that significant instability occurs within the first 5-10 minutes, after the application of the bias voltage. This showed that warm-up is required for a few minutes for stable operation of the detector. This instability was correlated with a rise in ASIC temperature. The effect of this instability decreases with increased bias voltage, which favors the use of higher bias voltages. The investigation of pixel behaviour facilitates the classification of pixels into malfunctioning, non-working and well-behaved pixels. Further investigations on the malfunctioning pixels revealed the slow drifting pixels which are hard to detect. To identify these malfunctioning pixels, a pixel masking technique is developed. A significant reduction in streaks and ring artefacts is observed with this technique. Further, it was identified that some clusters of malfunctioning pixels have correlated behaviour. This shows that when such clusters are treated collectively, they behave the same as a well-behaved pixel. This allows their use in image processing with minimal characterisation.
In summary, an accurate beam modelling of an x-ray source and better understanding of the detector enable us to improve the spectral outcome of the MARS spectral system.