Positron emission tomography (PET) is a medical diagnosis technique that shows images of the distribution of radioisotopes inside the body, thus allowing the observation of the metabolic processes that occur there.

PET systems detect pairs of gamma rays, of 511KeV each, which are generated after the annihilation of a positron with an electron. From each detected gamma ray, it is possible to obtain its energy, the position in which the ray is detected and the exact time in which the event has been detected. The coincidence time can be defined as the difference between the detected times of two consecutive events. The time resolution of a system will be determined by the timing error that is generated in the measurement of the coincidence time. An improvement in the time resolution implies a noise reduction in the PET reconstructed images.

Traditionally, received event time extraction techniques have been implemented analogically. These techniques allow the achievement of a very good time resolution. However, the electronic that is used in analog techniques is adapted to a specific detector, is complex and with few possibilities of configuration. The search for coincidences in systems with many detectors that are based on analog techniques is a dificult task, because it requires an increment in the complexity that is unaffordable in many occasions. Nowadays, it exists a current trend to reduce the PET systems analog electronic, substituting it with digital electronic by means of an early digitalization of the detector signals. Consequently, digital algorithms for time stamp extraction from the detected events should be proposed and evaluated. Right now, digital techniques do not achieve the time resolutions of analog techniques.

The present study deals about digital time discriminators. In the research group, there is a setup for measurement with two gamma ray detectors. Their output signals are processed in an analog front-end. The signals obtained from the analog front-end are digitalized in a data acquisition system where they are digitally processed. The initial motivation of the present research was the search for an improvement in the time resolution of the system and the fact that the digital processing, still with a low application in PET systems, can provide a big improvement in that resolution.

During the PhD Thesis, different digital algorithms for time information extraction of the received pulses are proposed. In a first step, a state of the art is prepared. Based on it, different possible solutions to the problem are proposed. The proposed algorithms are based on digital processing blocks that can be combined between them. These algorithms extract the time information of the processed signals by means of digital and configurable versions adapted from their analog equivalents. One of these digital blocks that has been implemented is used to increase the sampling frequency by means a low pass filter interpolation. The block can interpolate the input signals by factors x2 and x3. Among the analyzed algorithms, the study proposes new discriminators based on a digital calculus of the charge of the received interpolated or not interpolated pulses, obtaining better results than the classical configurations that work directly with the acquired pulse without processing.

To conduct the study about the behavior of the evaluated discriminators, a simulation is developed to validate all the PET system behavior from the detectors, generating pulses as realistic as possible; the analog electronic, simulated by means of SPICE simulators; and, at digital level, studying the behavior of the proposed algorithms. The algorithms are simulated with different configurations, clock jitter conditions and noise.

At last, all the proposed algorithms are programmed in the FPGA of the system selecting the right architectures to work in real time as soon as events are detected. A number of experiments using the available measurement setup are defined and conducted to validate in real conditions the proposed discriminators. The results obtained from the measurements also validate the developed simulation testbech. That way, a testbench is available for future developments.

In conclusion, the present work shows how the time resolution of the PET system can be improved just using the right digital signal processing methods. 

It is expected that the next generation PET systems will be based on an intense digital signal processing. That is due to the great flexibility of the digital processing when compared with the analog one, the increase of the velocity of programmable digital devices and the increase of the sampling rate and the analog to digital converters (ADCs) integration. Consequently, the solutions that are investigated in works like the present one will provide a basic and necessary contribution for future developments.