Home Lower Extremity CT Angiography (CTA)
Post
Cancel

Lower Extremity CT Angiography (CTA)

Posted May 31, 2009 Updated Apr 7, 2023
By Justus E. Roos MD , Tejas Rakshe PhD , David N. Tran BS , Jarrett Rosenberg PhD , Matus Straka PhD , Tamer El-Helw MD , Marc C. Sofilos RT , Sy Napel PhD , Dominik Fleischmann MD
7 min read

Rationale and Objectives

Existing density- and gradient-based automated centerline-extraction algorithms fail in severely diseased or occluded arterial segments for the generation of curved planar reformations (CPRs). We aimed to quantitatively and qualitatively assess the precision of a knowledge-based centerline-extraction algorithm in patients with occluded femoro-popliteal artery (FPA).

Material and Methods

Computed tomography angiograms of 38 FPA occlusions (mean length 120 mm) were retrospectively identified. Reference centerlines were determined as the mean of eight manual expert readings. Each occlusion was also interpolated using a new knowledge-based algorithm (partial vector space projection [PVSP]), which uses shape information extracted from a separate database of 30 nondiseased FPAs. Precision of PVSP was quantified as the maximum departure error (MDE) from the standard of reference and the proportion of the interpolated centerlines remaining within an assumed vessel radius of 3 mm. Multiple regression method was used to determine the factors predicting the precision of the algorithm. CPR quality was independently assigned by two readers.

Results

The mean MDE (in mm) for occlusion lengths of <50 mm, 50–100 mm, 100–200 mm, and >200 mm was 0.95, 1.19, 1.40, and 2.25, for manual readings and 1.68, 2.90, 9.43, and 19.95 for PVSP, respectively. MDEs of the algorithm were completely contained within 3 mm of the assumed vessel radius in 20 of 38 occlusions. CPR quality was rated diagnostic by both readers in 23 of 38 occlusions.

Conclusion

Shape-based centerline extraction of FPA occlusions in lower extremity CTA is feasible, and independent from local density and gradient information. PVSP centerline extraction allows interpolation of occlusions up to 100 mm within the variability of manually derived centerlines.

In lower extremity computed tomography angiography (CTA), cross-sectional viewing of diseased arterial segments is frequently required for visual assessment of the arterial flow channels . Longitudinal cross sections along the arterial centerlines—so called curved planar reformations (CPR) —are routinely reconstructed to complement the three-dimensional visualization methods like volume rendering or maximum intensity projection.

State-of-the-art postprocessing workstations facilitate the generation of CPRs by providing semiautomated centerline extraction methods . These algorithms, which ultimately exploit the density and gradient information within the CT datasets, generally work well in normal or minimally diseased vessels but consistently fail in the clinically relevant segments that are severely diseased or occluded. Time-consuming manual centerline extraction is then required before CPRs can be generated and the composition and length of an arterial occlusion be analyzed for treatment planning . A recently proposed knowledge-based approach to this problem (Partial Vector Space Projection [PVSP]) uses vascular shape information rather than CT density values to automatically interpolate centerlines through simulated arterial occlusions in patients with normal femoro-popliteal arteries (FPA).

Get Radiology Tree app to read full this article<

Material and methods

Patients and Image Acquisition Parameters

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Table 1

Femoro-popliteal Artery Occlusion Lengths

Fontaine Occlusion Length Occlusions Right Left IIA IIB III IV NI ≤3 cm 7 3 4 5 2 0 0 0 3–5 cm 7 4 3 6 0 0 0 1 5–10 cm 9 3 6 4 1 0 1 3 15–20 cm 2 0 2 1 1 0 0 0 ≥20 cm 3 2 4 2 2 2 0 0 Total 38 17 21 21 8 3 1 5

NI: no information available.

The occlusion lengths are shown along with the Fontaine classification method for 38 occlusions in 24 patients with peripheral arterial occlusive disease.

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Automated Centerline Extraction of Nonoccluded Vascular Segments

Get Radiology Tree app to read full this article<

Figure 1, Standard of reference: manual centerline extraction of occluded segments. The figure illustrates the process of manual centerline extraction. First ( i ), the user places initial control points (A to D) so that the patent segments (blue segments AB and CD) can be automatically tracked and the occluded segment (pink segment BC) begins as a straight line. The CPR of the centerline (segment AD) is shown. Next ( ii ), additional control points are inserted (yellow asterisk) in the center of the vessel (note the semicircular vessel wall calcification) to replace the previous untracked coordinate (pink square) and the new centerline is computed. Finally ( iii ), the curved planar reformation of the new centerline that includes the manually traced occlusion is shown.

Get Radiology Tree app to read full this article<

Manual Centerline Extraction of Occluded Vascular Segments (Reference Standard)

Get Radiology Tree app to read full this article<

Knowledge-based Centerline Estimation of Occluded Vascular Segments

Get Radiology Tree app to read full this article<

Figure 2, Principle of knowledge-based centerline interpolation of occluded arterial segments. Schematic demonstrates centerline interpolation of diseased vascular segments using a principal component analysis (PCA)-based algorithm. The knowledge-based algorithm uses a database of femoro-popliteal artery (FPA) shapes (red lines) extracted from normal individuals using PCA. The principal components (PC) (yellow lines) represent a set of standardized vascular shapes that can be combined and fitted to the patent portions (blue lines, between AB and CD) of an occluded FPA to interpolate the occlusion (green line, between BC). The residual error of fitted centerline to the patent segments of the FPA (arrows in magnified view) is called the neighborhood error (NE).

Get Radiology Tree app to read full this article<

Generation of CPR Images

Get Radiology Tree app to read full this article<

Analysis

Standard of Reference: Manual Centerline Extraction

Get Radiology Tree app to read full this article<

Quantification of Error between Automated Centerline Interpolation and Reference Standard

Get Radiology Tree app to read full this article<

Qualitative Analysis of CPRs derived from Automated Centerline Interpolation

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Statistical Analysis

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Results

Reference Standard: Manual Centerline Extraction by Expert Readers

Get Radiology Tree app to read full this article<

Figure 3, Interreader variability in manual centerline extraction. The box plot shows the variability between each of the four readers (R1 to R4) in the radial distances between the readers' manual centerlines for the 38 occlusions. The box boundaries mark the 25th and 75th percentiles, the vertical gray line is the median, the whiskers are 1.5 times the box width, and the points beyond the whiskers represent the outliers.

Get Radiology Tree app to read full this article<

Quantification of Error between Automated Centerline Interpolation and Standard of Reference

Get Radiology Tree app to read full this article<

Figure 4, Maximum departure error of automated centerline as a function of occlusion length.

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Figure 5, Proportion of centerline being outside the vessel as a function of occlusion length. Points are observations; line is a Lowess regression estimate of the mean proportion.

Get Radiology Tree app to read full this article<

Qualitative Analysis of CPRs derived from Automated Centerline Interpolation

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Figure 6, Mean maximum departure error of automated centerline as a function of quality rank score. The quality rank score is defined such that a higher score indicates better image quality. Points are observations; line is a Lowess regression estimate of the mean MDE.

Figure 7, Probability of nondiagnostic image quality as a function of occlusion length. Points are observations; line is logistic regression fit.

Figure 8, Clinical example demonstrating the knowledge-based interpolation of a femoro-popliteal artery (FPA) occlusion. Right mid FPA occlusion in an 87-year-old patient with peripheral arterial occlusive disease. Graphs demonstrate the FPA with a 3 mm radius (gray band) in x and y directions. The centerlines of the occluded segment (magnified view) were separately traced manually (colored lines) and estimated automatically using the knowledge-based algorithm (black line). Lateral and anteroposterior projection curved planar reformations demonstrate the multifocal mid-FPA occlusion (8 cm) using the knowledge-based centerline as having almost identical image quality compared to the manually extracted centerline.

Get Radiology Tree app to read full this article<

Discussion

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

Get Radiology Tree app to read full this article<

References

  • 1. Catalano C., Fraioli F., Laghi A., et. al.: Infrarenal aortic and lower-extremity arterial disease: diagnostic performance of multi-detector row CT angiography. Radiology 2004; 231: pp. 555-563.

  • 2. Fleischmann D., Hallett R.L., Rubin G.D.: CT angiography of peripheral arterial disease. J Vasc Interv Radiol 2006; 17: pp. 3-26.

  • 3. Martin M.L., Tay K.H., Flak B., et. al.: Multidetector CT angiography of the aortoiliac system and lower extremities: a prospective comparison with digital subtraction angiography. AJR Am J Roentgenol 2003; 180: pp. 1085-1091.

  • 4. Ofer A., Nitecki S.S., Linn S., et. al.: Multidetector CT angiography of peripheral vascular disease: a prospective comparison with intraarterial digital subtraction angiography. AJR Am J Roentgenol 2003; 180: pp. 719-724.

  • 5. Portugaller H.R., Schoellnast H., Hausegger K.A., et. al.: Multislice spiral CT angiography in peripheral arterial occlusive disease: a valuable tool in detecting significant arterial lumen narrowing?. Eur Radiol 2004; 14: pp. 1681-1687.

  • 6. Rubin G.D., Schmidt A.J., Logan L.J., et. al.: Multi-detector row CT angiography of lower extremity arterial inflow and runoff: initial experience. Radiology 2001; 221: pp. 146-158.

  • 7. Raman R., Napel S., Beaulieu C.F., et. al.: Automated generation of curved planar reformations from volume data: method and evaluation. Radiology 2002; 223: pp. 275-280.

  • 8. Roos J.E., Fleischmann D., Koechl A., et. al.: Multipath curved planar reformation of the peripheral arterial tree in CT angiography. Radiology 2007; 244: pp. 281-290.

  • 9. La Cruz A., Straka M., Koechl A., et. al.: Non-linear model fitting to parameterize diseased blood vessels.IEEE Visualisation.2004.IEEE Computer SocietyAustin, TX:pp. 393-400.

  • 10. Visser K., Kock M.C.J.M., Kuntz K.M., et. al.: Cost-effectiveness targets for multi-detector row CT angiography in the work-up of patients with intermittent claudication. Radiology 2003; 227: pp. 647-656.

  • 11. Rakshe T., Fleischmann D., Rosenberg J., et. al.: Knowledge-based interpolation of curves: application to femoropopliteal arterial centerline restoration. Med Image Anal 2007; 11: pp. 157-168.

  • 12. Kanitsar A., Fleischmann D., Wegenkittl R., et. al.: CPR - curved planar reformation.IEEE Visualization.2002.IEEE Computer SocietyBoston:pp. 37-44.

  • 13. Sandgren T., Sonesson B., Ahlgren A.R., et. al.: Factors predicting the diameter of the popliteal artery in healthy humans. J Vasc Surg 1998; 28: pp. 284-289.

Academic Radiology, Volume 16, Issue 6
Journals General Radiology
This post is licensed under CC BY 4.0 by the author.
Recently Updated
Contents