Automated Quantitative Imaging Measurements of Disease Severity in Patients with Nonthrombotic Iliac Vein Compression

Clinical question
Can an automated segmentation technique (AST) quantify disease severity and treatment response on CT venography for patients with lower extremity venous disease?

Take-away point
Automated segmentation technique (AST) can quantify leg volume, skin thickness, and water content of fat on CT venography of patients with lower extremity venous disease.

Reference
Automated Quantitative Imaging Measurements of Disease Severity in Patients with Nonthrombotic Iliac Vein Compression. Reposar, A.L., Mabud, T.S., Eifler, A.C., Hoogi, A., Arendt, V., Cohn, D.M., Rubin, D.L., Hofmann, L.V. Journal of Journal of Vascular and Interventional Radiology (JVIR), Volume 33, Issue 2, 270-275.

Click here for abstract

Study design
Single arm, retrospective, cohort study

Funding source
No reported funding

Setting
Academic hospital, Stanford University School of Medicine, United States.




Figure 1. a) Automated segmentation technique (AST) algorithm depicting segmented skin, muscle, subcutaneous fat, and bone. b-d) Individually segmented tissue layers for muscle, fat, and bone, respectively.

Summary

Nonthrombotic iliac vein compression and its associated myriad clinical manifestations can lead to a significant reduction in quality of life. Early diagnosis is critical to ensure a better prognosis. However, there are no consensus imaging tools or quantitative diagnostic criteria. The authors performed a retrospective study of 21 patients with left-sided nonthrombotic iliac vein compression who underwent venous stent placement with pre- and post-stent CT venography and investigated the technical feasibility of an automated segmentation technique (AST) for quantitative disease severity measurements.

Patients included in this study were diagnosed with left-sided nonthrombotic iliac vein compression without right-sided disease, underwent iliofemoral venous stenting (indicated for at least 70% stenosis or axial diameters of less than 4 mm), and had pre- and post-stent diagnostic lower extremity CT venography. Images were obtained from the institutional picture archiving and communications system (PACS), de-identified, curtailed between the inguinal ligament to the ankle, and packaged using OsiriX. An automated segmentation technique (AST) algorithm developed in MATLAB was used for segmentation according to Hounsfield unit (HU) ranges for bone, fat, muscle, and skin with subsequent calculations of leg volume, skin thickness, and water content of fat.

Significant differences were found in all 3 measures of disease severity between the left diseased and right non-diseased legs on pre-stent CT venography. The differences in skin thickness and leg volume persisted on post-stent placement though to lesser degrees while no significant difference was observed in water content of fat between the diseased and non-diseased lower extremity after stent placement. Same leg comparison pre- and post-stent demonstrated significantly lower water content of fat in both the diseased left and the non-diseased right leg; no difference in skin thickness or leg volume was observed within the same leg pre- and post-stent in both the diseased left and the non-diseased right leg.

Commentary

The authors in the paper developed an automated segmentation technique (AST) algorithm and successfully quantified 3 measures of nonthrombotic iliac vein compression disease severity based on pre- and post-stent CT venography. Results demonstrated significant differences between diseased left and non-diseased right leg on pre-stent imaging which persisted on post-stent imaging except for water content of fat. These results suggested that AST quantification of disease severity is technically feasible and may provide a nonobjective perspective on venous disease treatment response in general. This is an important and encouraging first step. Future efforts should focus on larger patient cohort, longer follow-up period, fine-tuning of existing and development of other quantitative measures. Statistical analyses will need to be more robust with the inclusion of multiple-comparison correction.

Post author
Ningcheng (Peter) Li, MD, MS
Integrated Interventional Radiology Resident, PGY-3
Department of Interventional Radiology
Oregon Health and Science University, Dotter Interventional Institute
@NingchengLi