摘要
Ultrasound assessment of the respiratory-induced change in size of the inferior vena cava is a useful technique in the evaluation and management of critically ill patients. We have developed an automated technique based on the Kanade–Lucas–Tomasi feature tracker and pyramidal segmentation to continuously track the diameter of the inferior vena cava during ultrasound. To test the accuracy of this automated process, the inferior vena cava of 47 spontaneously breathing patients were measured by trained ultrasound physicians and compared against the results obtained via the automated tracking. Good agreement between the techniques was found, with intra-class correlation coefficients for maximum vessel diameter, minimum diameter and caval index of 0.897, 0.967 and 0.975, respectively. More than 95% of the difference between physicians and automated measurements agreed to within 10% of the inferior vena cava collapse. Furthermore a phenomenon of cardiac collapsibility index variability was observed and reported. The accuracy and precision of this algorithmic technique provide a foundation for future automated measures for critical care ultrasound.