Perfusion CT measurements in healthy cervical spinal cord: feasibility and repeatability of the study as well as interchangeability of the perfusion estimates using two commercially available software packages

Abstract

Our purpose was to examine the feasibility and reproducibility of perfusion CT studies in the cervical spinal cord and the
interchangeability of the values obtained by two post-processing methods. The perfusion CT studies of 40 patients with neck
tumours were post-processed using two software packages (Software-1: deconvolution-based analysis with adiabatic tissue homogeneity
approach and Software-2: maximum-slope-model with Patlak analysis). Eight patients were examined twice for assessing the reproducibility
of the technique. Two neuroradiologists separately post-processed the images with two arterial input functions (AIFs): (1)
the internal carotid artery (ICA) and (2) the vertebral artery (VA). Maps of blood flow (F) in ml/min/100g, blood volume
(V) in ml/100g, mean transit time (MTT) in seconds (s) and permeability (PS) in ml/min/100g were generated. The mean F,
V, MTT and PS (Software-1) with VA-AIF and ICA-AIF were 8.93, 1.12, 16.3, 1.88 and 8.57, 1.19, 16.85 and 1.94, respectively.
The reproducibility of the techniques was satisfactory, while the V and MTT values (in Software-1) and the F and V values
(in Software-2) were dependent on the site of the AIF (p≥0.03 and p=0.02, respectively). The interobserver agreement was very
good. The significant differences in measurements for a single patient (%) using Software-1/Software-2 were ±120%/110%, 90%/80%,
180% and 250%/130% for F, V, MTT and PS, respectively. Only F and PS values in the healthy tissue seemed to be interchangeable.
Our results were in essential agreement with those derived by invasive measurements in animals. The cervical spine perfusion
CT studies are feasible and reproducible. The present knowledge has to be validated with studies in spinal cord tumours in
order to decide the usefulness of the perfusion CT in this field.