Survey of chest radiography systems: Any link between contrast detail measurements and visual grading analysis?

To evaluate image quality of chest radiography for a number of systems in Belgium, using a contrast-detail (c-d) test object and Visual Grading Analysis (VGA) of an anthropomorphic phantom.

The study comprised 22 chest imaging systems in Belgium. C-d data were measured using Leeds TO20 test object, imaged using poly(methyl methacrylate) (PMMA) thicknesses of 9, 13 and 16 cm. Images of the Lungman phantom, with additional tissue-equivalent chest plates to represent different patient sizes, were then acquired. Perceived image quality was evaluated using VGA by three radiologists. Images were acquired at a patient equivalent position with system-specific exposure settings for Posterior-Anterior chest protocol. Incident air kerma (IAK) was measured using a solid-state dosemeter.

C-d results showed large differences between the systems. Total number of visible discs ranged from 38 to 83 (for 9 cm PMMA) with a consistent average drop of 10% as PMMA thickness was systematically increased. However, no correlation was found between number of visible discs and IAK. Perceived image quality scored by the readers from the Lungman images decreased with increasing phantom thickness, however no correlation of VGA score with IAK was seen. Moderate correlation was found between the VGA score of one of the readers and the TO20 results, and no correlation for the rest.

The spread in dose and image quality measures was high and no correlation was seen between either image quality measure and IAK, suggesting the need for optimization. A more powerful tool is required for task-based optimization in chest radiography.