Observer performance in the localization of tubes and catheters on digital chest images: The role of expertise and image enhancement

Rationale and Objectives. To determine whether observer performance in the localization of tubes and catheters on computed radiography (CR) chest images can be improved by using an automatic image-processing algorithm. Methods. Comparisons were made of CR hard-copy, CR soft-copy, and CR enhanced soft-copy images obtained with an image-processing algorithm. The enhanced images used gray-level optimization and nonlinear unsharp masking to emphasize the edges of the devices. Chest radiologists (n = 4), general radiologists (n = 4), and interns (n = 6) read 45 images that contained endotracheal tubes, pulmonary artery catheters, and central venous catheters. Results. Chest radiologists had the smallest mean interobserver localization variability (4 mm), followed by general radiologists (6 mm) and interns (8 mm). Localization variability was greatest for hard-copy images and least for enhanced soft-copy images. Conclusion. Use of an automatic imaging-processing algorithm reduced localization variability and enabled the medical interns to perform at approximately the same level as the chest radiologists.