American Association of Physicists in Medicine, 29th Annual Meeting, Detroit, Michigan, 19-23 July. Medical Physics, 14(3):494, May/June 1987.
Linear filtering techniques have been utilized in digital angiography to suppress an interfering feature while enhancing a desired feature. Conventional DSA and matched filtering are successful in suppressing those features remaining constant, but will not suppress features that vary. Eigenimage filtering as initially developed will suppress a varying feature, but will simultaneously suppress a constant feature only under certain constraints. Eigenimage filtering has been expanded to suppress a multiple interfering features. The enhanced version satisfies the important requirement that it's integral is zero. The technique is demonstrated to be superior in isolating a feature of interest when there are overlapping features which interfere with its observation. A simulated study and an experimental canine cardiac study are presented which compare the improved eigenimage filter with the original filter, matched filter, and conventional DSA.