Purpose: To characterize the effect of post-processing settings on chest image quality metrology inradiographs of an anthropomorphic phantom relative to patients.
Methods: A sample patient cohort consisting of posteroanterior chest radiographs with normal anatomy and Exposure Index (EI) within an acceptable range was collected from a commercial CXR system (Philips DigitalDiagnost). An anthropomorphic chest phantom (Kyoto-Kagaku Lungman) was imaged on the same system using a similar protocol and EI. Two datasets were created by reprocessing the phantom and a representative patient’s radiograph by permuting five post-processing parameters, namely Contrast, Brightness, Enhancement, Object Detection and Noise Ratio. Eight chest-specific image quality metrics characterizing noise, detail, and contrast were measured from each rendered image using an in vivoimage quality assessment algorithm. The variations in image quality metrics were plotted as functions of post-processing parameter settings and were fitted with functions that included a patient-phantom distinction term. Regression analysis of the fits was performed to determine the statistical significance of the distinction between patient and phantom responses to post-processing settings.
Results: Image quality metrics for the patient and phantom varied similarly with post-processing settings in the majority of cases. Statistically significant differences between phantom and patient data were observed in two instances; after altering the Contrast post-processing parameter within the lung-specific image quality metrics, and after altering the Noise parameter within the lung noise metric. However, in both cases, patient and phantom curves still exhibited similar shapes, suggesting that the patient data can be inferred from the phantom.
Conclusion: Image quality of post-processed, anthropomorphic phantom images can be representative and predictive of patient chest image quality. The results affirm the value and relevance of phantom-based measures to inform clinical practice, and highlight the importance of image processing optimization for quality imaging practice.