Room: Exhibit Hall | Forum 8
Purpose: To explore the relation between DI and contrast-to-noise ratio (CNR) in portable digital chest radiographs (CXRs).
Methods: IRB approved this retrospective study. Data-log files corresponding to all bedside CXRs acquired over a one-year period using nine Carestream DRX-Revolution x-ray units (DRX1C image receptors) were extracted and analyzed. The vendor-reported DI and normalized CNR (noise estimated from the central mediastinum/spinal region; contrast from log exposure differences of lung region versus mediastinum) were analyzed. Analysis was split for cases falling in Â±0.5 range around DI values of 0, Â±4 and Â±9, representing 0.1 to 7.9 times the target exposure. Rejected CXR data was analyzed separately.
Results: Data was obtained from 31,923 CXRs. Mean (Â± standard deviation) DI was 0.8Â±2.1 (median: 0.8). Mean CNR was 0.5Â±0.2 (median: 0.5). Correlation between vendor-reported DI and CNR was 0.5. For examinations falling in Â±0.5 range around DI values of -9, -4, 0, 4 and 9, the CNR was 0.2Â±0.0, 0.2Â±0.1, 0.5Â±0.1, 0.6Â±0.2 and 0.4Â±0.1, respectively, indicating a gradual CNR increase with receptor exposure, and then saturation. Strikingly, even at target DI of 0Â±0.5 (18% of all CXRs) the CNR showed a wide spread ranging from 0 to 1 (partially, this may be due to inherent differences within patientsâ€™ CXRs due to underlying disease status, etc.)Í¾ this spread from 0 to 1 for CNR also extended for DI within a range of Â±3 (representing half to double the target exposure). The DI distribution of rejected CXRs overlapped with that of accepted CXRs (mean: 0.4Â±2.9Í¾ median: 0.6)Í¾ for subset of rejected images due to exposure parameters, the mean DI was -1.6 and median DI was -2.7, indicating insufficient exposure.
Conclusion: The observed wide-range in CNR, even for optimal exposures, suggests wide variability of quantitative image quality in portable CXRs which may be another target for quality improvement.