(Sunday, 7/12/2020) [Eastern Time (GMT-4)]
Room: AAPM ePoster Library
Purpose: To evaluate the use of Gabor channelized Hotelling observer (CHO) to assess fluoroscopic image quality, and to investigate the effect of image processing on object detectability index (d’) for different object sizes, detector target dose rates, and for slow and fast-moving objects.
Methods: Fluoroscopic images of an anthropomorphic chest phantom and iodine contrast rods (0.5 to 4.0 mm diameter) were acquired at a frame rate of 15 s^(-1). The rods were contained in a PMMA disk and rotation of 2 and 20 cycles min^(-1) was used to create linear velocities of 5.2 and 52 mm s^(-1). The phantom was imaged with detector target doses (DTD) 18, 36, and 65 nGy frame^(-1). Images were recorded with system’s image-processing enabled and disabled. The images were then analyzed to estimate d’ for each rod using a framework incorporating correction of d’ estimation bias caused by finite sampling and nonstationary noise.
Results: For unprocessed images, d’ varied in proportion to the square-root of DTD, as expected for a quantum-limited system. For slow-moving objects, image processing increased d’ compared to unprocessed images, especially for larger objects and higher DTD. Image processing did not improve d’ for small objects and low DTD. Less improvement was observed in the processed images for fast-moving compared to slow-moving objects, especially for small objects. For example, d’ of the 0.75 mm diameter rod was significantly decreased (Welch’s t-test, p<0.05) from 11.4±5.8 to 8.6±5.0.
Conclusion: We demonstrated the use of CHO in evaluating fluoroscopic image quality and demonstrated the effect of image processing on d’. Results showed that image processing preferentially increased d’ for large and slow-moving objects and this enhancement increased with increasing DTD. However, image processing did not enhance d’ for relatively low d’ conditions, including small objects and low DTD.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
