210 Towards Reducing Operator Dependence in Ultrasound: Oriented 3D Volumetric Imaging of Second Trimester Fetuses and Reliability of Image Volume Interpretation by Expert and Novice Raters

Abstract

Point-of-care (POC) obstetric ultrasound (US) is an essential tool in the emergency setting, but its clinical utility is limited by the inherent operator dependence of US. Performing a thorough and accurate US examination requires highly trained and experienced operators. We developed a low cost system enabling acquisition of oriented 3D image volumes using any existing 2DUS machine. We hypothesized that a novice sonographer could use the device to acquire image volumes of second trimester fetuses that could then be interpreted for assessment of basic parameters and reliable biometric measurements by both novice and expert raters. A student with minimal US experience (novice) swept the 2DUS transducer with paired research device across the region of interest to acquire image volumes of second trimester fetuses from 32 subjects. Volumes were reconstructed using a pixel-based algorithm. Acquisition and reconstruction times were automatically recorded. Image volumes were viewed in 3DSlicer; the percentage of interpretable volumes was recorded. An expert made blinded, timed assessments of placental location (PL), fetal presentation (FP), and amniotic fluid volume (AFV) for each subject. Both novice and expert raters blindly, independently measured BPD, humerus, and femur; corresponding gestational age (GA) estimates were calculated. Inter-rater reliability of biometric measurements and associated GA between each rater and clinical US reports and between raters were assessed by intraclass correlation coefficient (ICC). Mean inter-rater differences were analyzed by 1-way ANOVA. Mean (SD) age, GA, and body mass index were 30.6 years (±6.2), 18.9 weeks (±2.6), and 26.0 (±4.4). Volume acquisition and reconstruction required mean 30.4s (±5.7) and 70.0s (±24.0). PL, FP, and AFV could be evaluated from volumes for all subjects. At least 1 structural measurement was possible for 31 subjects (97%). Mean time for expert assessment of PL, FP, and AFV was 16s (± 0.0). Agreement for all measures between all rater pairs was excellent (ICCs greater than 0.95). ICCs between all rater pairs for each biometric measure and corresponding GA estimate demonstrated either good or excellent reliability. ICCs were highest for GA estimates by BPD between both raters and reported values (expert ICC: 0.97 CI95% 0.94-0.99; novice ICC: 0.97 CI95% 0.94-0.99). Mean inter-rater differences were not significant. Comprehensive image volumes of second trimester fetuses were rapidly acquired by a novice sonographer and reconstructed for interactive visualization. PL, FP, and AFV could be quickly assessed by an expert for all subjects. Measurements of BPD, femur, and humerus were made for the vast majority of subjects. Both the measurements and GA estimates were highly reliable between both novice and expert raters and clinical reports. This low cost system could reduce the operator dependence of US, thereby enhancing its utility in any clinical scenario in which POC US assessment is required. It could also facilitate task-shifting among health care personnel to increase the efficiency of triage and diagnostic evaluation.