Image quality in forensic CT imaging

Abstract

What factors affect the diagnostic image quality in forensic computed tomography (CT) imaging? In this thesis I briefly present the main steps from CT scanning to diagnostics. The thesis is built on four core papers. The first paper focuses on reconstructions of a quality control phantom measurement series. Objective image quality descriptors (contrast to noise ratio (CNR), mean CT numbers) are used for comparisons. Besides the reported objective measurements, some unexpected properties are identified, for instance, decreasing CNR when low level of iterative reconstruction is used. The second paper evaluates the applicability of an in vivo CT protocol in post mortem imaging. Radiographers and radiographer students are asked to score nine reconstruction alternatives from three post mortem cardiac cases. The results indicate that in this case the in vivo protocol is applicable in post mortem conditions. Images not only have to be measured, but also have to be presented. The third paper proposes a tone mapping approach which can compress the dynamic range of the CT image while preserves local contrast. Traditionally, a CT volume is read several times using different intensity windows. This contribution might reduce the number of required readings of the same volume, and could improve visualization of multiple pathologies in a single image. The fourth paper tackles a classic image processing problem, the de-quantization problem. The appearance of false contours might be both disturbing and misleading. If false contours are already present in an image, e.g. due to aggressive denoising, this approach might reduce or eliminate them. These four contributions are the core of this thesis, augmented with non peer reviewed contributions.