This thesis describes the techniques for modeling and control of 3D X-ray cardiovascular systems in terms of Image Quality and patient dose, aiming at optimizing the diagnostic quality. When aiming at maximum Image Quality (IQ), a cascaded system constituted from inter-dependent imaging components, is not easily optimized and may have a sub-optimal operation. Multi-dimensional optimization in the vast design area, including pre-sampling (physics) components and post-sampling (processing) parts, is difficult to implement. To come to this overall optimized IQ, with a low patient dose, we address the complete process in three parts. We commence with optimizing an important influencing part of the detector, called the scintillator, which shows a deterioration by ghost images through the imaging history. In the first part of the thesis, we propose a new solution to reduce ghost images in the scintillator layer, by means of external illumination and accurate modeling of the process. After and before the detector, a system based on a chain of image processing components appears with mutual interdependencies. One possible way to establish overall optimization of this system, is to make performance models of all system components and then use these models for IQ improvement. These models include inherent IQ, environmental- and system-induced deterioration effects and (patient) dose control. This approach eventually enables the design of an optimized 3D system. This modeling and optimization forms the middle part of this thesis. The third part of the thesis is dedicated to an optimized dynamic dose control, which is the ultimate parameter for the control system that is used for defining a detector dose, that is virtually immune for adverse irradiation conditions originating from direct radiation. The three areas are briefly summarized below. The reduction of ghost images in medical X-ray imaging is investigated by UV-light exposure of a CsI:Tl scintillator. We expect that by using UV light, so-called traps are filled, thereby reducing the memory effects caused by X-ray irradiation. Ideally, the image sensed by the detector is a function of the X-ray pattern projected on the detector’s surface at a current point in time. This means that the detector is ideally not affected by previously captured images (memory-less capturing). However, the X-ray patterns previously projected on the detector, disturb the properties of the different components therein. This disturbance is attributed to charge-carrier traps in the scintillator layer, as well as in the detector photodiode array. In the produced images, this disturbance is seen as spurious "ghost images" of previously recorded images, which are superimposed on the actual image. This effect is not desired, as it hampers the diagnostic value of low-contrast images. The research is devoted to the explanation and modeling of the phenomenon of ghost-reduction by using UV-light on a CsI:Tl scintillator layer, both at an atomic level and at a macroscopic level. In the second chapter, an extensive literature study is reported in the field of scintillation and radiation effects. In a succeeding chapter, a model is proposed, explaining the effect on a microscopic scale, followed by a model linking the effects to Image Quality. The modeling work also involves validation experiments on the X-ray imaging devices as well as on complete state-of-the-art X-ray systems. The modeling pertains to X-ray, optics, electronics, optoelectronics, chemical analysis, image analysis and image processing. Experiments have been complemented by extensive measurements on scintillator samples, referring to specific materials properties. The prototypes have been analyzed in terms of Image Quality, such as Modulation-, Contrast- and Noisetransfer and related to materials properties. With the support of Philips Healthcare Best, we have realized optimized prototypes of a detector subsystem incorporated at a system level. A model on UV-irradiation effects has been developed, through which the optimization of a fully-functional detection system has been reached. We have found that an optimum wavelength and an optimum irradiation exists. For these conditions, the results of the applied method are equally good as for using existing methods. As a bonus, the short-term afterglow is reduced considerably, thereby improving the X-ray IQ. In comparison with published results on co-doping with Sm or Eu, wherein sensitivity losses up to 30% are encountered, our method increases the sensitivity up to 10%. The method has the inherent advantage of an increased lifetime of the components and does not require patient removal, so that the examination room is more efficiently used. The second part of this thesis involves the modeling and the verification of an X-ray 3D cone-beam system based on multiple components, in terms of IQ and Patient Dose. Again, a comprehensive literature survey supplies a basis for modeling, which is extended with a discussion on fundamentals of IQ descriptors. The research includes an extensive verification of the IQ at the system level and necessitates the development of appropriate measurement means. IQ is influenced by both internal and external factors. We discuss environmental- and system-induced vibration effects on sharpness, both in 2D subtraction angiography as well as in 3D reconstruction imaging. For vibration-induced deterioration effects on IQ, a verified model is embedded in the 3D simulation module. The method results in a fast system design for IQ, while keeping patient doses within predefined levels. Optimized designs for IQ are realized at the system level by addressing all the components in the imaging chain. The IQ system performance is expressed in Sharpness, Contrast and Noise, balanced with a well-defined patient dose and uses appropriate figures of merit for optimization, such as the Modulation Transfer Function and Contrast-to-Noise Ratio. The experiments have been conducted together with the industry, which has fueled an application for validation. To this end, an optimized protocol for Nitinol stent imaging has been introduced at the system level, which has resulted in a substantial increase of the IQ, at a patient dose index well below European recommendations. The protocol enables to visualize the stent’s struts by virtually continuous strokes and allows for an accurate post-intervention inspection. The third part of the thesis deals with dynamic dose control within the framework of the total system optimization on IQ. The final IQ is reinforced by the design of a novel dose-control signal processing. Conventional dose control systems use a control signal derived from an image, which relies on arithmetic averaging. This method is extremely sensitive to highlights entering a measuring field used for deriving an input control signal. It is shown that an evaluation based on taking a geometric average yields a significant improvement of the low-contrast detection limit for de-centered objects. The new design is virtually insensitive to adverse detector-irradiation conditions related to highlights originating in direct radiation or highly transparent body parts. The thesis ends with conclusions and an outlook. It is discussed that an IQ-based FOM can be based on Shannons expression for information capacity on one hand and a Pareto-based analysis of the design space, set up by the parameters defined in this thesis on IQ and Patient Dose (PD) on the other hand. When detailed knowledge of the parameter space is established, it could be used to explore the design freedom for a cost-effective X-ray system. With respect to reduction of memory effects in the CsI:Tl scintillator, we pose that our solution using UV illumination may also be applicable to general CT scanning devices.
|Qualification||Doctor of Philosophy|
|Award date||15 Nov 2012|
|Place of Publication||Eindhoven|
|Publication status||Published - 2012|