Image analysis

Two musculoskeletal radiologists (reader 1 with six-year experience, reader 3 with four-year experience) and a trainee (reader 2) independently analyzed images on Picture archiving and communicating system (PACS; Infinitt Co., Ltd., Seoul, Korea). We obtained axial images parallel to intervertebral discs on CT and MRI and selected an axial plane passing through the disc center (Fig. 1). On axial CT and MRI, regions of interest (ROIs) were manually drawn along the thoracolumbar fascia (Fig. 2); this corresponded to the “total CSA” referred to in previous studies, not “fCSA” [9, 16]. We assessed inter- and intraobserver reliabilities of qualitative and quantitative measurements on CT and MRI. First, the three readers were trained by one experienced observer how to draw ROIs and measure GGs under consensus and analyzed images blinded to each other’s results. Measurements were repeated after 2 weeks for intraobserver agreement. We noted the presence of disc pathologies if there were any.

A schematic diagram defining an axial plane in sagittal plane of CT and MRI. The axial plane was obtained by looking at the point parallel to the disc passing through the center of the disc height. And consecutive images were obtained according to CT and MR parallel to the central line

Image shows an example of the region of interest (ROI) of the paraspinal back muscle on axial plane and acquisition of the quantitative measurement values on PACS workstation. On CT (a), when the ROI is drawn, mean CSA (cross sectional area) and mean HU (Hounsfield unit) are displayed on screen. On MRI (b), when the ROI is drawn on in-phase image of two-point Dixon technique, the CSA and signal intensity values are displayed as on CT, and the ROI is copied and pasted onto the fat only image (c), then CSA and signal intensity measured