Purpose: A statistical channelized Hotelling observer (CHO) was previously developed to assess x-ray angiography system performance for inherently non-stationary imaging conditions. The purpose of this work was to use this CHO to assess the influence of non-linear image processing on detectability of moving task objects in the presence of an anatomically-relevant background.
Methods: The adapted CHO model is based on multivariate statistical comparison of a single test object present image and many object absent images. It incorporates efficient Gabor channel selection and estimation of and correction for bias. Iodine contrast test objects were capsule-shaped with diameter 0.75-4 mm. The moving task objects (2.6 cm/s) were placed upon an anthropomorphic phantom to provide an anatomically-relevant background. Images were acquired both unprocessed and processed with 2 variants of spatio-temporal recursive filtering performed by the manufacturer.
Results: Preliminary experiments demonstrated that the CHO was linear with respect to changes in task object size and x-ray photon fluence. For unprocessed images, detectability (dâ€™) varied in the range dâ€™=1.7-155 dependent on object size and spatially-variable attenuation of the anthropomorphic phantom. The two levels of spatio-temporal image processing resulted in distinct levels of amplification of dâ€™ of the test objects. Studentâ€™s t-test comparison of dâ€™ for processed versus unprocessed images demonstrated that image processing significantly (p<0.05) amplified dâ€™ for test objects with dâ€™>~5. As dâ€™ of the test objects decreased below dâ€™<5, the influence of processing on dâ€™ trended toward no effect.
Conclusion: Spatio-temporal image processing can substantially increase detectability for relatively conspicuous test objects. However, detectability of smaller, lower contrast test objects was not enhanced by image processing. CHO assessment of moving test objects on an anthropomorphic background may be a useful tool for quantitative assessment of the influence of image processing on x-ray angiography image quality.
Angiography, Observer Performance