On-chip intelligence advances scientific imaging

16 Oct 2007

Active-pixel sensor technology has the potential to break the constraints of current scientific imaging, according to new work by a UK-based research collaboration.

Autoradiography

A new class of pixel-based imaging devices being developed by a UK consortium is now being trialed in scientific imaging applications ranging from particle physics and space science to medicine and biology. Active-pixel sensor (APS) technology exploits mainstream CMOS technology to incorporate programmable transistors into each pixel of the image array.

This in-built programmability enables the pixels to perform signal-processing functions such as analogue-digital conversion and time-digital conversion. Other features include the ability to home in on specific regions of interest on the chip, and to enhance the sensitivity via back-thinning of the chip.

"With APS, we can put intelligent signal-processing on the chip and then we can do things like improve the sensitivity or quality of the images," said Nigel Allinson, professor of image engineering at the University of Sheffield and leader of the Multidimensional Integrated Intelligent Imaging (MI-3) project. This collaborative project, which was set up three years ago and involves 11 research centres, will receive funding totaling £4.4 million ($9 million) over its four-year lifetime.

According to Allinson, the extra features of APS sensors are needed to meet the demands of scientific applications. As well as low noise and high speed, such devices must offer a large dynamic range, high linearity and a large active area. Another must-have is a wide spectral range covering high-energy gamma rays and ionizing particles through to IR radiation.

APS-based detectors offer the potential to meet all these requirements. The first APS to come out of the MI-3 project is a 520×520-pixel chip called Vanilla. The sensor contains 25 µm pixels with a total active area of 13×13 mm², operates at 100 frames/s and is two-side buttable to provide a larger sensor area.

Meeting of minds
Vanilla is currently being used by consortium members for a variety of applications. Speaking at the MI-3 project's open day, held in October at University College London (UCL), Allinson noted that "it's probably the medical side that has made the greatest progress".

Robert Speller, professor of medical physics at UCL, highlighted several medical imaging projects that are now exploiting the Vanilla chip:

• Researchers at the University of Surrey are investigating the Vanilla APS as a possible digital replacement for radiographic film, which suffers from limited dynamic range, poor linearity and low sensitivity, which in turn leads to long exposure times. The main challenge here is that the signal being collected can be extremely weak, which means the sensor needs a low noise level.

Speller presented an image of a tritium-labelled sample, recorded over 1.5 days with the Vanilla APS, which showed near-comparable quality to a corresponding image taken over 28 days with film. "The ability to collect this information rapidly is the big advantage of a CMOS sensor," he explained.

• At UCL, meanwhile, researchers are examining the use of tissue diffraction imaging for diagnosing breast cancers. This technique involves analysing the scattered X-rays, which are usually filtered out of a mammogram, for structure-specific information.

APS technology is ideal for this application as the sensor's high dynamic range enables it to record both the transmitted and scattered radiation on one chip. In addition, it may be possible to develop on-pixel intelligence to automatically analyse the sample composition.

"If you look at the scattered photons, you can recognize the different types of tissue," Speller explained. "We feel that you can take these data and adapt them for use in a conventional mammography system."

• A team at Brunel University and the Institute of Cancer Research (ICR) are looking to use APS technology for so-called portal imaging, in which the radiotherapy beam is used to acquire images that are then used to to verify a patient's treatment plan during intensity-modulated radiotherapy. The CCDs and flat-panel detectors currently employed for this task are not generally fast enough to enable real-time control, but an APS sensor could allow immediate feedback on any problems during treatment. "APS offers the possibility of being able to rapidly analyse the collected data and decide whether the beam is the beam you expect," said Speller.

Forward look
The consortium is now on the brink of releasing its large-area sensor, or LAS. LAS is a stitched sensor comprising a 1400×1400 array of 40 µm pixels, with a total active area of 56×56 mm². Also in the pipeline is eLeNA, a low-noise APS. "We are eagerly awaiting those new sensors that will come online soon so that we can take all of these applications further," said Speller.

•MI-3 comprises: the University of Sheffield; the Science and Technology Facilities Council; Brunel University; the University of Glasgow; the University of Liverpool (Liverpool Semiconductor Detector Centre and Laboratory for Environmental Gene Regulation); University College London; the University of Surrey; the University of York; the Institute for Cancer Research; and the MRC Laboratory of Molecular Biology, Cambridge.