Effect of Exposure Response Function On Flat-Field Non-Uniformity Metrics to Assess Digital Detector Performance
J Dave1*, E Gingold1, (1) Thomas Jefferson University Hospital, Philadelphia, PASU-E-I-93 Sunday 3:00PM - 6:00PM Room: Exhibit Hall
Purpose: To investigate how digital radiography flat-field image quality metrics are affected by the exposure response function.
Methods:Sets of four flat-field images were acquired from 3 Carestream digital radiography detectors at 2 institutions using manufacturer specific calibration geometry over a range of input air kerma levels (2.4-160μGy). The images were exported from the acquisition system with both logarithmic and linear data space conversions (i.e., exposure response functions). The images were also retrieved from PACS. The percent non-uniformities in three characteristic metrics i.e., signal (mean), noise (standard deviation) and SNR were calculated; local non-uniformities were based on a 100mm² region and its nearest neighbors and global non-uniformities were based on regions with extreme values of signal, noise and SNR.
Results:Images exported from the acquisition computer with logarithmic exposure response function were labeled For-Processing with Photometric Interpretation=Monochrome1, Pixel Intensity Relationship=Log and Relationship Sign=+1. Interestingly, the images with linear exposure response function also indicated a Log Pixel Intensity Relationship. For images exported from PACS, the corresponding DICOM attributes were For-Presentation, Monochrome2, log and -1. As expected, the characteristic signal, noise and SNR values showed exposure dependency consistent with the exposure response function. Using logarithmic data space conversion as reference, the absolute differences in non-uniformities calculated using linear data space conversion ranged from 0.1% to 10.7% and using images from PACS ranged from 0% to 11% for detectors configured with an appropriate gain map. For one detector configured with an inappropriate gain map, the discrepancy in calculated non-uniformities between linear and log data ranged from 0% to 34.6%. These differences in non-uniformities at only the calibration air kerma level were within 4.5%.
Conclusion:The calculated non-uniformity metrics showed a dependence on the exposure response function. Thus, for detector performance evaluation, comparisons and quality assurance testing, knowledge of the exposure response function is required.