Inter- and Intraobserver Agreement in Measuring Urolithiasis Density on Nonenhanced Computed Tomography

Abstract

Introduction: Treatment choice for urolithiasis is partially based on measuring stone density in HU on nonenhanced computed tomography (NECT). Interobserver variability in these measurements could have treatment consequences. This study aims to assess the observer agreement of measuring HU and whether the use of a protocol leads to a better agreement. Materials and Methods: We retrospectively included 155 consecutive NECTs of patients with stones ≥4 mm. Five observers (two radiologists, one urologist, one urology resident, and one radiology resident) assessed all anonymized NECTs four times in randomized order. HU was measured without instruction (rounds 1 and 2) and subsequently using two protocols (A and B, rounds 3 and 4). Protocols comprised using bone setting, zoom, and measuring HU without the penumbra, in either three (A) or one (B) axial plane. The inter- and intraobserver agreement was evaluated using the intraclass correlation coefficient (ICC). Results: Interobserver agreement on HU measurement without protocol was as follows: ICC = 0.84 (confidence interval [CI]: 0.79-0.87). Agreement diminished with protocol A, ICC = 0.62 (CI: 0.37-0.76), and improved with protocol B, ICC = 0.90 (CI: 0.86-0.92). Intraobserver agreement without protocol was ICC = 0.87, with protocol A, ICC = 0.87, and with protocol B, ICC = 0.93. The biggest improvement was seen for urologists' agreement from no protocol to protocol B, where ICC improved from 0.81 (CI: 0.70-0.87) to 0.91 (CI: 0.84-0.94). Conclusions: Observer agreement of HU measurement of urolithiasis without protocol is already good but using zoom, bone setting, and measuring in a representative plane is recommended. This protocol results in higher agreement, especially among urologists. Measuring in three axial planes does not increase agreement.

Keywords: Hounsfield units; X-ray computed tomography; observer variation; stone density; urinary calculi.