Quantification of lower leg arterial calcifications by high-resolution peripheral quantitative computed tomography

Abstract

Vascular calcifications and bone health seem to be etiologically linked via common risk factors such as aging and subclinical chronic inflammation. Epidemiologic studies have shown significant associations between low bone mineral density (BMD), fragility fractures and calcifications of the coronary arteries and the abdominal aorta. In the last decade, high-resolution peripheral quantitative computed tomography (HR-pQCT) has emerged as in-vivo research tool for the assessment of peripheral bone geometry, density, and microarchitecture. Although vascular calcifications are frequently observed as incidental findings in HR-pQCT scans, they have not yet been incorporated into quantitative HR-pQCT analyses. We developed a semi-automated algorithm to quantify lower leg arterial calcifications (LLACs), captured by HR-pQCT. The objective of our study was to determine validity and reliability of the LLAC measure. HR-pQCT scans were downscaled to a voxel size of 250μm. After subtraction of bone volumes from the scans, LLACs were detected and contoured by a semi-automated, dual-threshold seed-point segmentation. LLAC mass (in mg hydroxyapatite; HA) was calculated as the product of voxel-based calcification volume (mm(3)) and mean calcification density (mgHA/cm(3))/1000. To determine validity, we compared LLACs to coronary artery calcifications (CACs), as quantified by multi-detector computed tomography (MDCT) and Agatston scoring in forty-six patients on chronic hemodialysis. Moreover, we investigated associations of LLACs with age, time on dialysis, type-2 diabetes mellitus, history of stroke, and myocardial infarction. In a second step, we determined intra- and inter-reader reliability of the LLAC measure. In the validity study, LLACs were present (>0mgHA) in 76% of patients, 78% of patients had CACs (>0mgHA). Median LLAC was 6.65 (0.08-24.40)mgHA and median CAC as expressed by Agatston score was 266.3 (15.88-1877.28). We found a significant positive correlation between LLAC and CAC (rho=0.6; p<0.01). Dialysis patients with type-2 diabetes mellitus (DM; 35%) and history of stroke (13%) had higher median LLAC than patients without those conditions (DM 20.0 fold greater, p=0.006; Stroke 5.1 fold greater, p=0.047). LLAC was positively correlated with time on dialysis (rho=0.337, p=0.029), there was a trend towards a positive association of LLAC and age (rho=0.289, p=0.053). The reliability study yielded excellent intra- and inter-reader agreement of the LLAC measure (intra-reader ICC=0.999, 95% CI=0.998-1.000; inter-reader ICC=0.998, 95% CI=0.994-0.999). Our study indicates that the LLAC measure has good validity and excellent reliability. The use of HR-pQCT for the simultaneous evaluation of arterial calcifications, peripheral bone geometry, bone density, and bone microarchitecture should facilitate future research on osteo-vascular interactions and potential associations with cardiovascular events.

Keywords: Agatston-score; HR-pQCT; Lower leg arterial calcifications; Quantification.

Figure 1. Lower leg arterial calcifications (LLAC) depicted by HR-pQCT

A) Anterior tibial artery calcification (arrow). B) Automatic detection of bone contours may include calcifications of the juxtaposed anterior tibial artery, requiring manual correction prior to quantitative LLAC assessment (C).

Figure 2. Non-vascular, high-density regions that should be excluded from quantitative analysis of vascular calcifications

Examples of findings that can mimic arterial calcifications are shown below and highlighted with arrows A) intra-cutaneous spotted calcifications, B) non-vascular, most likely post-traumatic soft tissue calcification, C) focal pixel artifact, D) skeletal motion artifacts.

Figure 3. Validity study comparing lower leg arterial calcifications (LLAC) and coronary artery calcifications (CAC)

A) Unranked scatter plot showing LLAC by HR-pQCT and CAC as measured by MDCT and Agatston score in n=46 patients. B) Ranked scatter plot of LLAC and CAC visualizing the Spearman correlation coefficient (rho=0.6; p<0.001).

Figure 4. Reliability study

Visualization of lower leg arterial calcification (LLAC) values obtained by repeated image analysis by a single operator (white bullets) and a second operator (black bullets).