Effect of training and experience on qualitative and quantitative CT perfusion data.
Academic Article
Overview
abstract
BACKGROUND AND PURPOSE: To evaluate interobserver reliability of obtaining CT perfusion (CTP) data for qualitative identification of perfusion abnormality and quantitative assessment through regions-of-interest (ROIs) placement. MATERIALS AND METHODS: Six observers participated in the study (neuroradiology attending physician, neurology attending physician, neuroradiology fellow, radiology resident physician, senior and junior CT technologists). After a brief training session, each observer evaluated 20 CTP datasets for qualitative identification of a right- or left-sided perfusion abnormality or symmetric perfusion. Observers also placed a single ROI of standard size to obtain quantitative data on the most severely hypoperfused region. An additional 10 ROIs were placed on the cortex to quantitatively evaluate global cortical perfusion. Mean quantitative cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time (MTT) values were analyzed. RESULTS: The kappa values for qualitative assessment of a perfusion abnormality ranged from 0.55 to 1.0. Coefficients of variation for quantitative assessment of ischemia/infarct region were 27.10% for CBF, 13.33% for CBV, and 4.66% for MTT. Coefficients of variation for quantitative assessment of global cortical perfusion were 11.88% for CBF, 13.66% for CBV, and 3.55% for MTT. The junior CT technologist and neuroradiology fellow showed significant differences compared with other observers for the ischemia/infarct region and global cortical perfusion, respectively. CONCLUSION: Overall, quantitative differences seen in this study would not necessarily affect quality of interpretation of ischemia/infarct region or global cortical perfusion. Therefore, obtaining qualitative and quantitative CTP data can reliably be performed in the clinical setting among observers with various levels of skill and experience when using a uniform and standard technique.
