143. Reliability of semi-automated spinal measurement software

Abstract

<h3>BACKGROUND CONTEXT</h3> In the treatment of patients with adult spinal deformity, analysis of spinopelvic balance is essential in determining treatment options. This includes pelvic tilt (PT) and incidence (PI), lumbar lordosis (LL), and the sagittal vertical axis (SVA). Historically, the measurement of radiographs was done by hand with poor to moderate inter-observer reliability. More recently digital PACS tools provided improved accuracy and reliability, with the latest software semi-automating this process with a promise of increased efficiency. <h3>PURPOSE</h3> To our knowledge, no studies have examined the inter-observer reliability of semi-automated PACS software. This study aims to compare the reliability of this software versus PACS to determine if the improved efficiency occurs at the cost of reliability. <h3>Methods</h3> Full spine X-rays were retrospectively reviewed from 25 patients older than 18 years seen at a tertiary medical center between 2014 and 2017. Patients were included if they had greater than 5cm of sagittal imbalance, without prior surgical spinal fusion and/or instrumentation. Spinopelvic parameters were measured in two radiographic programs: one with basic, non-spine specific measurement tools (eUnity, Client Outlook, Waterloo, Canada); and a second with spine-specific measurement tools that semi-automate the measurement of these parameters (Sectra, Sectra AB, Linköping, Sweden). Balance parameters included SVA, PI, PT, and LL. Two authors re-measured radiographs after a 6-month interval and results were compared to original measurements. Data were compared by examining inter-rater and inter-program reliability using interclass correlation coefficient (ICC). We hypothesized that modern semi-automated software, with its promise of increased measurement efficiency, would maintain reliable results. <h3>Results</h3> The subjects' mean age was 67.9 +/- 13.8 years old, and 32% were male. The inter-program reliability was strong, with ICC values greater than 0.91 for each parameter. Similarly, there was strong inter-observer reliability with ICC values greater than 0.88. Delayed re-measurement also showed strong reliability with ICC values greater than 0.90 (p < 0.001 for all measurements). <h3>Conclusions</h3> There is excellent inter-observer and inter-program reliability between the basic PACS and semi-automated programs. These data demonstrate that the purported efficiency of semi-automated measurement programs does not come at the cost of measurement reliability. <h3>FDA DEVICE/DRUG STATUS</h3> This abstract does not discuss or include any applicable devices or drugs.