Task-oriented optimization of high-performance positron emission tomography

点击翻译按钮获取中文摘要

Positron Emission Tomography (PET) allows the inside of objects to be investigated, be it the human brain for medical research or the inside of a mixing vessel for industrial processes. The objective of this research program is to advance the computational algorithms and methods used for imaging these processes. At present, the way the raw data are processed, reconstructed and analyzed does not pay full attention to the reason why the PET scan is being done. This leads to poor image quality and unnecessarily high noise. This program of research will radically change the way the processing is performed, to always keep in mind: i) the reason for doing the scan, ii) how to draw the best out of the limited noisy data, and iii) exploitation of other sources of information (e.g. images from other modalities). By accounting for these factors much improved image quality will be possible, permitting the full potential of any given PET scanner to be realized. The top priority for brain imaging will be more accurate measurements of brain function, whereas for industrial process imaging it will be enhanced resolution imaging of mixing or reaction dynamics. Ideally all algorithms will be fully integrated into a single unit to ensure optimal use of raw PET data and accurate noise modeling at every stage, achieving the highest performance for the task in hand with the available data. A further pivotal component to this research program will be robust, advanced validation and quality assurance for novel PET scanners, which will be vigorously pursued in order to underpin all of the image quality improvements.