Signal-to-Noise:
SNRs were compared in 1902 ROI pairs of 110 images.
An SNR was calculated for each ROI pair,
and the SNRs of kinetic images were divided by the SNRs of sumDSA images.
Table 1 lists the medians of these SNR ratios and shows the fold increase in SNR of kinetic images compared with the DSA images.
Medians were calculated instead of mean values to avoid distortion caused by any outliers.
Visual comparison:
To demonstrate the effect of post-processing on kinetic image quality,
unprocessed and post-processed kinetic images were compared.
Interrater agreement in that post-processing increased kinetic image quality was more than 74%,
if all the 232 image pairs were combined (p<0.0001 for Q1,
Q2,
Q3).
Looking at each region separately (abdominal (40),
iliac (40),
femoral (28),
popliteal (53),
and talocrural (71)),
it was significant for the talocrural (p<0.0001 for Q1,
Q2,
Q3) and popliteal (p<0.0001 for Q1,
Q2,
Q3) regions.
In the other three regions (i.e.,
abdominal,
iliac,
femoral),
there was no significant agreement between the evaluators.
This could indicate that post-processing did not add value in these regions,
or it could be due to the lower numbers of images than in the talocrural and popliteal regions.
To address usability and quality differences between kinetic imaging and DSA,
238 pairs (abdominal (37),
iliac (40),
femoral (31),
popliteal (50),
and talocrural (80)) of post-processed kinetic and sumDSA images were compared.
Interrater agreement that kinetic images outperformed sumDSA was more than 73% if all 238 image pairs were combined (p<0.0001 for Q1,
Q2,
Q3).
Raters thought that the post-processed image was significantly better for the talocrural (p<0.0001 for Q1,
Q2,
Q3),
popliteal (p = 0.0016 for Q2 and p = 0.0100 for Q3) and abdominal (p = 0.0017 for Q2 and p = 0.0040 for Q3) regions.
There was no significant agreement between raters for the iliac and femoral regions.