Electronic Thesis and Dissertation Repository

Degree

Doctor of Philosophy

Program

Medical Biophysics

Supervisor

Cunningham, Ian A.

Affiliation

Robarts Research Institute

Abstract

X-ray imaging for early cancer detection, such as screening mammography, requires images with high signal-to-noise ratio (SNR) using low levels of radiation exposure. Conventional detectors consist of a matrix of sensor elements, producing images where each pixel corresponds to a single sensor element. This imposes a fundamental limitation on image contrast and SNR for imaging fine detail for a given exposure. The work presented here reconsiders x-ray image formation using a new x-ray detector design that synthesizes image pixels from a large number of very small sensor elements with the goal of optimizing contrast and SNR.

Our new detector design, called apodized-aperture pixel (AAP), makes use of recent technology developments to produce images from an “over-sampled” sensor signal while suppressing both signal and noise aliasing to improve the modulation transfer function (MTF) and detective quantum efficiency (DQE).

Signal and noise performance of the AAP approach is described theoretically using a cascaded-systems analysis. This approach preserves the MTF of the small sensor elements up to the image sampling cut-off frequency where the MTF is increased by up to 53%. Frequencies above the cut-off are suppressed, eliminating both signal and noise aliasing artifacts and corresponding to a high-frequency DQE increase by 2.5x. X-ray interactions in a scintillator introduce signal and noise correlations, including x-ray reabsorption and converter blur, resulting in reduced aliasing and decreased improvement in DQE. Best results with the AAP design were obtained using a high-resolution converter, such as selenium (Se), with little impact from reabsorption.

Implementation on a Se/CMOS micro-sensor prototype with 7.8\mum element size with image pixel size approximately 50\mum showed a flat DQE curve (ideal) up to 10cycles/mm. AAP images of resolution test patterns, mammography phantoms, and specimen imaging of micro-calcifications from biopsies showed the expected improvements in SNR and visibility of fine-detail.

It is concluded that synthesizing image pixels from small physical sensor elements can increase MTF and DQE, and eliminate aliasing artifacts, for a desired image pixel size. The resulting increase in SNR may benefit all forms of radiography, and in particular mammography, where accurate visualization of fine detail is important for early cancer detection.

Download

Share

COinS