Because field-programmable gate array (FPGA) processors enable the parallelism of operations, they are ideally suited for applications with high-speed processing requirements, like medical imagery. A positron emission tomography (PET) scanner is an example of a system that has very complex processing schemes and requires the high-speed treatment of data from many sensors.
Figure 1: The phases of data acquisition (DAQ)
As illustrated by Figure 1, data acquisition has three main phases. A sensor is used to acquire the analog (real world) signals, then analog-to-digital (A/D) converters transform the analog signals into digital data. Finally, the processing unit applies an algorithm to the data. The processed data may then be sent to a computer, such as for image construction.
The advantage of FPGAs is that, in this type of processor, the data may be processed in parallel, as shown by Figure 2. Parallelism accelerates the processing and also enables multiple channels to be processed at the same time, rather than needing the data to be submitted to a complex chain of events that serialize the treatment. What’s more, parallel processing enables you process multiple parallel inputs, such as the multiple outputs of an analog-to-digital converter.
Figure 2: Accelerating data processing via parallelism
The FPGAs enable you to:
- Deal with large numbers of sensors (multiple A/C converter channels)
- Rapidly perform complex and sophisticated processing
- Reduce the complexity and cost of designing multi-channel DAQ systems
In PET imagery, for example, there are many sensors. Sometimes, thousands of analog sensors must be taken into account to reproduce an image. To make the technology affordable, researchers from institutes such as the Imaging Sciences Lab at the UCLA Crump Institute for Molecular Imaging, must find a way to limit the number of analog-to-digital converters. This may involve time-multiplexing, in which the data of many sensors is sent to a multiplexer that switches at each clock step from one sensor to another. The data is then sent to the FPGA which contains an algorithm for re-constructing the data to create the image.
FPGAs and low-cost DAQs reduce costs
Researchers of the UCLA Crump institute are using Nutaq’s MI125 FPGA mezzanine card (FMC) acquisition board, based on a Linear Technology LTM®9012 and a Perseus601x advanced mezzanine card (AMC), for PET scan applications. A portable low-cost PET scanner is being developed for small animal imagery. Nutaq’s hardware provides high-speed processing (125 MSPS) and the ability to have many input channels (Nutaq’s FMC cards can be stacked on top of each other, up to 32 channels per stack). The researchers then use multiplexer circuitry to combine the scanner’s many inputs from the sensors into the 32 channels of Nutaq’s MI125.
One of the objectives of the research in PET scanning is to lower the costs of imagery systems, which often range into the multi-millions, as described in a previous blog article on the cost of small-animal scanners.
FPGAs are well-suited and Nutaq’s hardware reduces development costs
By providing fast parallel processing on FPGAs and enabling low-cost data acquisition from multiple channels, Nutaq hardware is ideal for PET applications. The hardware has been shown to efficiently reduce development costs and thus lower the acquisition cost of imaging systems.