X-ray image quality before and after radiographer commenting: A quantitative visual grading characteristics analysis

Introduction The Radiographer Comment and Alert (RCA) model was established to provide written alerts of abnormal radiographic appearances identified at acquisition, but its quantitative impact on image quality is unknown. This study quantifies differences in radiograph image quality before and after RCA implementation. Methods A total of 234 radiographs encompassing the pelvis, hand, wrist, elbow, knee, ankle, and foot were collected. This comprised all 117 investigations for these anatomical areas with radiographer comments over 12-months at one site, and 117 matched radiographs obtained one year before RCA implementation. Each image pair was displayed simultaneously, but the order and left-right orientation was randomised. Seventeen radiographers independently rated image pairs using a visual grading characteristics (VGC) analysis scale on seven quality criteria (rotation, angulation, centering, collimation, exposure, artifacts, diagnostic utility). Data were analysed using VGC Analyser Software v1.0.2. Non-parametric rank-invariant statistics generated an area under the VGC curve (AUC VGC ) to measure the difference in quality between images with and without radiographer comments. Results Across the combined dataset, images acquired after implementation of RCA received visual grading ratings for radiological quality that were significantly higher than images acquired prior to implementation (AUC VGC 0.52, p = 0.016, CI[0.500862,0.532967]). Statistical significance at 95% confidence was not met for individual upper (AUC VGC 0.52, p = 0.084) and lower limb (AUC VGC 0.50, p = 0.746) regions, but was achieved for the axial region (AUC VGC 0.54, p = 0.012). Quality improvement was driven by improved positioning, particularly rotation (AUC VGC 0.57, p = 0.002). Intraclass correlation coefficient scores were good or excellent (ICC3,k ≳ 0.62–0.97) and all reached significance ( p -value<0.05), indicating stable consensus and good inter-rater agreement. Conclusion Quantitative evidence demonstrates implementation of RCA significantly improves the overall quality of projection radiographs submitted for reporting. Implications for practice Radiographer image interpretation models can be utilised as tools to improve radiographic image quality.