PhD defense Fasco van Ommen

On Thursday 26 November 2020, Fasco van Ommen will defend his thesis entitled: “Dual-energy CT and CT perfusion for improved CT stroke imaging.”

Open Access: http://dspace.library.uu.nl/handle/1874/400117

Abstract

CT is often used as the modality of choice for the diagnosis of patients with suspicion of stroke. CT imaging allows visualization of occluded blood vessels and identification of location and volume of the infarction. Current CT technological innovations are aimed at increasing the diagnostic image quality, increasing the predictive value and quantitative accuracy of CT, and reducing the radiation dose. Modern CT techniques include CT perfusion (CTP) for infarct assessment and dual-energy CT (DECT) for improved tissue characterization. The goal of this thesis was to improve the diagnostic performance for infarct detection using novel DECT and improve CTP analysis with DECT and high-resolution CTP. In part one of this thesis, we showed that dual-layer detector CT, using a 120 kVp tube potential, can be routinely used in daily clinical practice to provide additional information without increasing radiation dose when compared with a conventional single-layer detector CT scanner. We also found that the image quality of conventional CT images acquired on a dual-layer CT scanner is similar to its counterpart on a conventional CT scanner for medium-sized phantoms, and slightly lower for (very) large phantoms at lower tube voltages. Next, we found that non-contrast DECT virtual monochromatic images (VMI) significantly improves the image quality of non-contrast brain CTs compared with conventional CT. In part two, we found that non-contrast head VMI at 80–90 keV more accurately detected and localized infarcts compared with conventional CT. Next, we found that virtual ischemia maps from non-contrast brain CT were more accurate than conventional CT in approximating infarct core volumes and the infarct location. This suggests that DECT more accurately differentiates between infarcted and healthy tissue than conventional CT. In addition, the added value of dual-energy VMI CTP scans to the quality of the perfusion maps was shown. We found that the image quality and visual quality of 50 keV CT perfusion maps is superior to that of conventional 80 and 120 kVp images. In part three, we found that increasing the acquisition interval may introduce a bias in the perfusion parameters, but this bias can be corrected by calibration of the perfusion maps and therefore still allow distinction between healthy and infarcted tissue. Infarct volumes can likewise be influenced by the acquisition interval, but visual inspection indicated minor differences in infarct volumes between acquisition intervals. For a commercial block-circulant singular value decomposition (bSVD) perfusion analysis package acquisition intervals up to 4 seconds could be achieved, and for a non-commercial bSVD and a non-linear regression model-based method intervals up to 5 seconds. We also shown the ability of thin-slice CTP to detect small-volume infarctions by noise reduction using two bilateral filters. We found that perfusion values are estimated more accurately and with higher contrast using guided bilateral filtering compared with time-intensity profile similarity (TIPS) bilateral filtering on thin-slice CTP. While the detection of small-volume infarctions remains difficult, infarcts could be detected with higher sensitivity and significantly higher diagnostic certainty and improved image quality using guided bilateral filtering than with the current state-of-the-art TIPS filter.