Functional CT in lung with a conventional scanner: simulations and sampling considerations.

Abstract

Due to rapid transit times, motion artefacts from breathing and the low signal intensity, functional computed tomography (f-CT) studies in lung tissue remain challenging with conventional CT scanners. The purpose of this study is to examine the accuracy of parameter estimates when performing deconvolution analysis with signals from lung tissue. The effects of partial volume averaging in lung tissue, differing transit times, variable vascular and capillary responses, expected noise levels, differing sampling rate and durations were simulated on a computer. Deconvolution using singular-value decomposition (SVD) analysis was performed for realistic lung signals using published and measured values of the arterial input and noise levels. The accuracy, bias and variance of the estimated residue functions and their associated parameter estimates were evaluated. We find that f-CT signals may be measured and analysed using SVD and other deconvolution approaches. Functional CT signals in the lung may be analysed provided that the rise and fall of the tissue and input curves are well sampled (regardless of sampling rate) and noise levels in the lung ROI tissue are approximately 20 HU or less, even for regions of interest that are mostly occupied by air. Estimates of the mean tissue transit time (MTT) are insensitive to air volume. Other decovolution methods such as fast Fourier transform methods provide more accurate estimates of PBF, whereas SVD approaches provide more accurate estimates of pulmonary blood volume and MTT. F-CT of the lung with a conventional scanner should be possible, when the extra dose is not a consideration.