Using the Fourth Dimension to Distinguish Between Structures for Anisotropic Diffusion Filtering in 4D CT Perfusion Scans - Longitudinal Registration and Shape Modeling - page 77

anisotropic diffusion filter we performed an observer study. The quality of the
arteriograms and venograms derived from the filtered data were compared to
the original arteriograms and venograms, by two expert observers (radiologists).
Maximum intensity projections (MIPs) of 40 mm were presented to the observers
in pairs (unfiltered versus filtered) with two different window level settings: a
window level (W/L: 380/220) for the assessment of large vessels ( > 1.5 mm)
and a window level (W/L: 200/120) for the assessment of small vessels ( < 1.5
mm). Observers were asked to score the quality (on a 5-point scale) of the large
and small arteries in the arteriogram and large and small veins in the venogram
and to specify whether more arteries and veins were visible in the arteriograms
and venograms derived from the filtered CTP data. Table 1 shows the results of
the observer study. In general the quality of the small vessels was scored to be
improved and the quality of the large vessels was scored equal in the majority of
the cases. Observer agreement [ 19 ] was very good ( ʺ : 0.81) for the large veins,
and good for the large arteries ( ʺ : 0.66) and small arteries and veins ( ʺ : 0.61).
In the majority of the cases, the observers specified that more small arteries
(Obs.1: 70 %, Obs.2: 75 % with ʺ : 0.95) and veins (Obs.1: 60 %, Obs.2: 80 %
with ʺ : 0.80) were visible in the arteriogram and venogram derived from the
filtered CTP data. Figure 2 shows two examples of cases in which the observers
scored that more small arteries and veins were visible in the arteriogram and
venogram derived from the filtered CTP data.

Table 1. Results of quality assessment (5-point scale) scores of the arteries and veins
in arteriograms and venograms derived from the filtered CTP data compared to those
derived from the original CTP data. Percentages indicate in how many percent of the
20 cases the corresponding score was assigned by the observer.

Quality assessment Large Small Large Small
Expert observer 1 Arteries Arteries Ve ins Ve ins
Much worse 0% 0% 0% 0%

Worse 0% 0% 0% 0%

Equal 65 % 30 % 85 % 20 %

Better 30 % 60 % 15 % 80 %

Much better 5% 10 % 0% 0%

Expert observer 2
Much worse 0% 0% 0% 0%

Worse 5% 0% 0% 15 %

Equal 75 % 15 % 100 % 15 %

Better 20 % 75 % 0% 65 %

Much better 0% 10 % 0% 5%

To determine the signal to noise ratio (SNR) in the original and filtered CTP
data, regions of interest (ROIs) were placed in large and small arteries in the