Validation of an algorithm that separates gaseous micro-embolic signals and artifacts during transcranial Doppler persistent foramen ovale examinations

Abstract

Objective

Persistent foramen ovale (PFO) is a risk factor for young stroke. Agitated saline serum is used to deliver small gaseous emboli to the brain in transcranial Doppler ultrasound (TCD) for the detection and grading of PFO. In this study we validated a PFO algorithm that can differentiate between gaseous emboli and artifacts.

Methods

The validation cohort comprised 18 patients with positive PFO examinations. The PFO algorithm uses a binary tree that separates high-intensity transients signals (HITs) into gaseous emboli or artifacts based on intensity, zero-crossing, and velocity parameters.

Results

The cohort exhibited 385 macroscopic gaseous emboli meeting the >3 dB criterion. An additional 137 gaseous emboli were noticed below the 3-dB intensity cutoff value. The low-intensity gaseous emboli included both macroscopic and microscopic air bubbles observed in curtains. Nearly all emboli (98 %) above the 3-dB level showed overt frequency modulation. The overall accuracy of the PFO algorithm in discriminating macroscopic gaseous emboli and artifacts was 96.4 %, with a similar percentage of sensitivity and specificity (96.4 %). The inter-observer agreement of human experts was excellent (ic-CC 0.989 and 0.953).

Conclusions

Macroscopic gaseous emboli and artifacts during PFO exams can be accurately discriminated by the PFO algorithm. The PFO algorithm cannot be used as a standalone system as microscopic air bubbles might escape proper identification. This knowledge will be important in the design of future PFO algorithms which should make it possible to classify the PFO grade without the interference of humans.