Use of model-based iterative reconstruction (MBIR) techniques in computed tomography (CT)

Over the past decades, the use of computed tomography (CT) has risen exponentially due to expanding need in medical diagnosis and treatment planning. The use of CT is not entirely without risk and involves theoretical risk from radiation. This has led to several advancement in dose reduction techniques. Most promising is the use of so called “iterative reconstruction” which is a post image acquisition step combining a software/hardware configuration that allows the images acquired to be processed more efficiently, eliminating noise which increases with low dose techniques. This thesis examines the use of a specific iterative reconstruction technique known as model-based iterative reconstruction (MBIR) and assesses image quality, diagnostic accuracy, and feasibility of implementation in routine clinical practice. With no prior guidelines for implementation, first we assessed this technique in pre-clinical setting by scanning a realistic model of a human (known as a phantom). This allows us to scan at various tube current/voltage settings as well as fixed/automatic adjustments to assess the most optimal parameters by assessing image quality. Second, we assessed if image quality can be improved using the same dose for routine patients. Third, we assessed image quality and diagnostic accuracy in 3 body parts by scanning patients with matched intra-individual comparison of standard and low dose techniques. This was done for chest, abdomen/pelvis, and colon. Fourth, we tested the use of low dose techniques in specific clinical situations, assessing diagnostic accuracy in the context of lung nodule follow-up and renal stone detection in patients presenting with renal colic.