AGREEMENT BETWEEN TRAINED AND UNTRAINED OBSERVERS IN GRADING INTRAVASCULAR BUBBLE SIGNALS IN ULTRASONIC IMAGES

In our experience, it is easier to identify gas bubbles in ultrasonic images than in aural Doppler signals. To verify this, we asked 27 obse rvers with no previous training to estimate the quantity of gas bubble s in video tapes containing sequences of ultrasound images recorded du ring decompression experiments. The amount of bubbles was graded accor ding to a non-linear grading system with six levels. The results obtai ned were compared to evaluations performed on-site by a trained observ er. Approximately 70% of the evaluations performed by the untrained ob servers agreed completely with the on-site gradings, and more than 95% agreed within 1 grade unit. The strength of agreement can be describe d by use of the weighted kappa statistic, and we have compared the agr eement in our study with agreement obtained in a previous study using Doppler signals for bubble detection. We find that in grading bubble s ignals in images, untrained observers perform equally as well as train ed observers grading bubbles in Doppler signals. We conclude that ultr asonic imaging offers a useful and cost-effective alternative to Doppl er systems for detection and quantification of intravascular gas bubbl es.