Clinical Evaluation of Chest Radiographs Simulated from Helical Computed Tomography

This work evaluated the clinical utility and image quality of simulated chest radiographs derived from helical CT scans, compared with a CT localizer and a recent chest radiograph.

A cohort of 100 patients were retrospectively evaluated: most chest radiographs (96/100) were acquired on portable units within 24 h of the CT, where only an anterior-posterior view was acquired. The patient table was segmented from each CT dataset and cone-beam projections in the posterior-anterior and lateral orientations were generated, which were then fed into a multi-objective frequency processing (MFP) framework to obtain simulated chest radiographs (sCXR). The sCXR were qualitatively compared to the localizer and radiograph by three radiologists using Likert scale assessments. Friedman and Wilcoxon tests were used to evaluate categorical trends and statistical significance. Image quality was quantitatively evaluated using the presampled modulation transfer function (MTF).

The sCXR demonstrated significantly superior image quality when compared to the CT localizer, specifically in the lung region. The tubes/lines/devices category received the lowest ratings across both modalities. At a 30-cm display field of view (DFOV), MFP projections showed an approximate twofold improvement in limiting resolution.

The sCXR demonstrated image quality characteristics that were generally comparable to conventional chest radiography in several of the categories under investigation, highlighting the potential of sCXR as a companion image with the added flexibility of generating multiple projection views. Radiologist scoring suggests that additional edge enhancement may be required for high-frequency structures, which experience greater blurring due to table motion and longer exposure times.