Radiation propagation modelling and image reconstruction for X-ray time-of-flight computed tomography

点击翻译按钮获取中文摘要

X-ray computed tomography (CT) is a very important 3D imaging modality finding widespread use in modern medicine. From its inception in the early 1970's to the 1990's, major developments in CT aimed at increasing the contrast in images to ease diagnostic. Without significant improvements in spatial resolution at that time and the more recent software developments in image enhancement techniques, greater contrast was largely obtained by increasing the X-ray dose injected into the patient. At the turn of the millennium, studies have pointed out the dose's adverse effects, where post-scan cancers were observed, and genetic modifications were suspected. Several approaches, both hardware and software, have then been developed to reduce the dose while preserving contrast. CT contrast is notably affected by the noise originating from photons that are deviated from a straight-line path, so-called scattered photons. Current CT image reconstruction algorithms assume that photons travel in straight lines, and thus detected scattered photons contribute noise in the reconstructed image. Our group has set up an ambitious research program for reducing the dose in CT, while increasing image quality and quantitative accuracy. Our approach relies on measuring the time of flight (ToF) of X photons to determine if they traveled in straight line (ballistic photons), or if they scattered. The principle is that a scattered photon takes more time to reach a detector than if it travels in straight line. This makes it possible to discriminate ballistic from scattered photons based on their ToFs. We are currently the world leaders in this field. In simulations and preliminary experiments, we have demonstrated that exploiting ToFs allows reducing the dose by a factor of 4. We have undertaken a joint endeavor whose ultimate goal is to build a complete ToF-CT scanner. CT image reconstruction taking advantage of ToFs has not yet been addressed. The present Discovery Grant application focuses on developing image reconstruction algorithms that do not yet exist owing to the novelty of the field, and which are essential to build a fully operational scanner. In addition to dose reduction predicted by our preliminary results, this application addresses another fundamental problem in CT. Apart from contributing noise, an adverse effect of scattering is to make impossible with current approaches the true quantitative measurement of absorption in absolute units, which is highly desirable for medical diagnostics. This lack of quantitative accuracy leads, for the same patient, to measurement non-uniformities when results from different scanners are compared (especially from different vendors). These non-uniformities are currently partially mitigated by way of approximate calibrations, with possible consequences on diagnostic accuracy. Additionally to dose reduction, the proposed work will provide an avenue to improve the quantitative accuracy of CT, leading to more reliable diagnosis.