Research On System Correction Methods For All-Digital PET Based On Multi-Voltage Threshold Sampling

Positron emission tomography (PET) is a state-of-the-art, non-invasive, nuclear imag-ing technique. It could detect the distribution of the radioactive tracer uptake in vivo, and then provide the metabolic activities, biochemical reactions, biological activities and per-fusion, dynamically and quantitatively. Nowadays, PET plays a significant role in the di-agnosis and treatment of oncology, cardiovascular disease and nervous system, as well as in the fields of biomedical research, pre-clinical applications, pathology research and drug development. As applications promote PET to be applied in more fields, higher sensitivity and resolution performances are seriously required in PET.Traditional PET usually has to confront several bottlenecks, such as low sensitivity, difficulty of system correction and inaccurate quantitative, due to its special electronic cir-cuits, fixed detector geometry and inflexible system architecture. In this dissertation, we have designed and implemented an all-digital PET data acquisition system (DAQ) based on multi-voltage threshold (MVT) sampling, which consists of basic detection modules (BDM-s) and software based coincidence detection system. Specially, the MVT sampling method was firstly employed to address the issue arising from directly digitizing scintillation pulses at the early stage in the signal chain, and construct all-digital PET scanner. Moreover, this method also provides a new opportunity to solve the problem of system correction in PET. The experimental results show that the detector module can provide an overall energy reso-lution of14.1%@511keV and a block-level coincidence timing resolution of1.5ns. Also in the micro-Derenzo imaging study, this detector module can clearly distinguish the1.0mm hollow channels, which indicates that this module has an intrinsic spatial resolution of1.0mm. The proposed DAQ system provides extensively flexibility, reliability and scalability to perform self-calibration and parameter adjustment, which allows detector modules to be arranged into different geometries to meet different imaging needs.In the DAQ system, baseline is feasible to shift due to the effect of leak current, pulse tail, thermal drifts of electronics and power-line disturbances, which would degrade signal-to-noise ratio (SNR) of PET raw data. This dissertation has proposed a novel pulse charac-terization based baseline restoration algorithm (PCBRA), to address the baseline issue for the MVT sampling. In this method, several time-threshold samples, digitally sampled from scintillation pulses with MVT sampling, are obtained to fit the waveform with a linear curve representing leading edge and an exponential curve representing falling edge, respective-ly. According to the mathematical model of the falling edge mentioned above, the baseline offset can be obtained and then subtracted from the waveform to retrieve the pulse without baseline shift. The experimental results show that this method can effectively retrieve the baseline offset, as well as the event energy and timing information at various baseline shifts. Moreover, this method also has encouraged correction performance in the4Mcps count rate situation.In the high count rate situation, the pileup effect would cause spatial mis-position, dis-tortion of energy spectrum and deterioration of timing resolution in the DAQ system. This dissertation has proposed a novel digital pulse pileup correction (DPPC) method, to solve the pileup problem for the MVT sampling. In this method, several time-threshold pairs, digital-ly sampled from scintillation pulses with MVT method, are obtained to determine whether the pulse is a multi-event pileup or a single-event by the counts of leading edge in scintil-lation duration. Each leading edge samples and its following tail samples are combined to implement the pulse reconstruction in a pileup. By using this method, pileup pulse could be separated into two or more reconstructed single-event pulses. And, the single events timing and energy information could also be derived from these reconstructed pulses. The exper-imental results show that this method can recover timing and energy information of each single-event in pileup with a count rate up to5Mcps, as well as significantly increase the coincidence events.In DAQ system, each detection unit has channel time delay due to the effect of physical characteristics difference between crystal-level units, inconsistent delay characteristics between PMTs and differences between time pick-off circuits, which would degrade system-level timing resolution. This dissertation has proposed a timing calibration method based on the timing resolution contribution matrix (TRCM), to solve the channel time delays for all-digital PET scanner. In this method, the ratios of the coincidence events in each line-of-response to the total coincidence events are obtained, respectively. Then, the channels that have significantly contribution to the whole timing resolution are selected to perform timing calibration. The experimental results show that this method can effectively obtain the time delays both for block-level units and crystal-level units, with significantly improving the timing resolution as well. Moreover, this method can effectively reduce the calibration complexity for multi-channel PET system, and achieve rapid timing calibration.