Diagnostic Accuracy of an Algorithm for Discriminating Presumed Solid and Gaseous Microembolic Signals During Transcranial Doppler Examinations

The aim of the work described here was to assess the diagnostic accuracy of a new algorithm (SGA-a) for time-domain analysis of transcranial Doppler audio signals to discriminate presumed solid and gaseous microembolic signals and artifacts (SGAs). SGA-a was validated by human experts in an artifact cohort of 20 patients subjected to a 1-h transcranial Doppler exam before cardiac surgery (cohort 1). Emboli were validated in a cohort of 10 patients after aortic valve replacement in a 4-h monitoring period (cohort 2). The SGA misclassification rate was estimated by testing SGA-a on artifact-free test files of solid and gaseous emboli. In cohort 1 (n = 24,429), artifacts were classified with an accuracy of 94.5%. In cohort 2 (n = 12,328), the accuracy in discriminating solid/gaseous emboli from artifacts was 85.6%. The 95% limits of agreement for, respectively, the numbers of presumed solids and gaseous emboli, artifacts and microembolic signals of undetermined origin were [−10, 10], [−14, 7] and [−9, 16], and the intra-class correction coefficients were 0.99, 0.99 and 0.99, respectively. The rate of misclassification of solid test files was 2%, and the rate of misclassification of gaseous test files was 12%. SGA-a can detect presumed solid and gaseous microembolic signals and differentiate them from artifacts. SGA-a could be of value when both solid and gaseous emboli may jeopardize brain function such as seen during cardiac valve and/or aortic arch replacement procedures.