At the University of California in Los Angeles, a group of positron emission tomography (PET) researchers selected Nutaq’s PicoDigitizer for their project because of its support for model-based development and its capacity for high-speed sampling and processing.

First, let’s review the basics of PET technology. The positron is the antiparticle associated with the electron. It has the same spin and mass as an electron but has a charge of –1 instead of +1. Medical imagery uses positron emission to monitor metabolic activities, a task that helps diagnose diseases like cancer.

Patients are injected with a radioactive isotope that decays emitting positrons (this type of decay is called positive beta decay). The positrons undergo collisions, causing them to lose kinetic energy up to a point where they then interact with the surrounding electrons. The combination of the two antiparticles causes their mutual annihilation and produces a pair of annihilation (gamma) photons, travelling in opposite directions. In existing PET scanners, the gamma photons are intercepted by scintillator crystals that stop them and use their energy to create lower energy photons (light). The light is then detected by silicon avalanche photodiodes (APD) or photomultiplier tubes (PMT).

The images of metabolic activities are obtained after processing the capture inputs, which need to be sampled at 0.5 ms. Previously, the processing of the high-throughput data was performed in the analog components. But, by incorporating the latest technology, the researchers were able to use rapid digital-to-analog converters to transform the data in digital signals for processing in parallel processors like field-programmable gate arrays (FPGAs).

FPGAs are well-suited for PET applications because they can process data very quickly. The algorithms in the FPGA serve multiple purposes, from image reconstruction to mapping the array of scintillators crystals in the detector.

The algorithm used for image reconstruction is based on the Radon transform, which relies on the scattering data to determine the original density, thus enabling image reconstruction. These calculations are made on the FPGA at very fast rates. The FPGA also makes other calculations to determine which crystal was hit by a gamma photon by using the energy information received by each photomultiplier. This is done prior to image reconstruction by using a technique developed by Charlie Burnham (1) and has been used with block detectors on all dedicated tomographs since 1985. Block detectors are optical multiplexing devices that combine “multiple small scintillator pixels on a single photomultiplier” (thus forming what we referred to earlier as an array).

Using the ancestor to Nutaq’s picoDigitizer, the VHS-ADC, researchers at UCLA proved that the model-based approach could be used to successfully program their algorithms in the FPGA. The VHS-ADC hardware was also successful in the production of low-cost commercial PETs with a small form factor. A key advantage of using Nutaq`s hardware was that it didn’t require an HDL programmer. Having only a single person be able to program the hardware would have very negative effects on the project team, as there would have been only one person able to understand the algorithm implementation. With a model-based design, each team member is able to understand and contribute to its implementation.