Pmj. Vanderzwet et al., AUTOMATED AND ACCURATE ASSESSMENT OF THE DISTRIBUTION, MAGNITUDE, ANDDIRECTION OF PINCUSHION DISTORTION IN ANGIOGRAPHIC IMAGES, Investigative radiology, 30(4), 1995, pp. 204-213
RATIONALE AND OBJECTIVES. Pincushion distortion continues to be a pote
ntial problem for the accurate assessment of arterial and catheter dim
ensions from x-ray angiograms. The authors investigate whether the dis
tortion of state-of-the-art intensifiers is yet small enough to be neg
lected, and whether the rotation/angulation of the x-ray system plays
a significant role. METHODS. The location and degree of distortion fro
m x-ray images of a centimeter grid, which is positioned against the i
nput screen of the image intensifier, are assessed automatically using
image processing techniques. A value for the maximum amount of change
in the distortion vector field is derived that allows the estimation
of the maximum relative error associated with a diameter measurement u
ncorrected for pincushion distortion. RESULTS. The accuracy of the alg
orithm itself was assessed by rotating and translating the centimeter
grid under the image intensifier at anteroposterior position. For the
distortion vector length, the standard deviation in the measurement of
the distortion areas was found to be 3.7 cm(2) (1.3% of the total are
a). For the gradient values, the standard deviation was 2.2 cm(2) or 0
.75% of the total image intensifier area. In the second evaluation stu
dy, the centimeter grid was fixed onto the input screen of the image i
ntensifier, and the gantry was rotated to span all possible positions
of the system. In this case, the changes in measured areas were often
much larger (up to 51.25 cm(2) for a 9-inch image intensifier, equival
ent to 15.6% of the total image intensifier area) than the standard de
viations that had been found in the first evaluation study. CONCLUSION
S. The distortion is highly dependent upon the actual spatial position
of the image intensifier, and correcting for pincushion distortion ma
y therefore introduce larger errors than leaving the measurements unco
rrected.