Departments of Radiology1 and Psychiatry2, University of California, San Diego San Diego, CA
Computerized tomography (CT) is the mainstay imaging modality for evaluation of ventricular size after shunting1 or external ventricular drain (EVD) placement.2 Rapid and accurate evaluation of ventricular size in the presence of shunts and EVDs is integral to guiding patient management. CT CoPilot (ZepMed, LLC, San Diego, CA) was developed to aid in interpretation of standard head CTs, which typically suffer from variable patient positioning, by automatically aligning CT images to a standard orientation. CoPilot also generates subtraction images between co-registered sequential scans to allow rapid visualization of changes between serial scans (Figure 1). In this study, we evaluated the ability of CT CoPilot to increase reader efficiency in detecting changes in ventricular volume in patients with ventricular shunts or EVDs.
48 patients with a ventricular shunt or EVD were retrospectively evaluated. All patients underwent two consecutive non-contrast head CT scans reconstructed in standard planes and also automatically post-processed with CT CoPilot. Scans were interpreted by two radiology residents and one neuroradiology fellow for change in ventricular volume between consecutive scans (decreased, unchanged, increased). Readers also recorded the time needed to confidently assess the change in ventricular volume. Readers first evaluated each patient for change in ventricular volume using only standard CT reconstructions. After a period of 3 weeks, the readers then re-evaluated each patient using CoPilot aligned reconstructions and CoPilot subtraction images. A linear mixed-effects model was used to determine whether CoPilot improved the speed of interpretation compared to interpretations performed without CoPilot. Bland-Altman tests were performed to evaluate the agreement between the readers and the gold standard assessment performed by a board-certified neuroradiologist.
The results of linear mixed-effects model showed a significant decrease in interpretation time when using CoPilot (mean = 10.71 s, SD = 10.26 s) compared to interpretation without CoPilot (mean = 39.28 s, SD = 21.02 s), F(1, 8.7) = 156.3, p < .0001. In other words, the readers could determine change in ventricular volume 73% faster using CoPilot than without CoPilot (Figure 2). Reader agreement with the gold standard assessment was also improved using CoPilot, as demonstrated by decreased limits of agreement in Bland-Altman analyses (1.39 with CoPilot versus 1.73 without).
CT CoPilot significantly improves reader efficiency when assessing for change in ventricular volume between consecutive CT scans in patients with shunts and EVDs, reducing average interpretation times by 73% without sacrificing accuracy. These results suggest that CT CoPilot could be a useful tool in any setting requiring rapid interpretation of head CTs for evaluation of shunt/EVD functionality.