Comparison of Image-based Calcification Scores to Total Calcium Content of Atherosclerotic Plaques

Background Peripheral Arterial Disease (PAD) is defined by the presence of atherosclerosis in the peripheral circulation, leading to ischemic injury of the extremities. PAD is a massive burden on the United States medical community, with a recorded $4.3 billion cost per year paid by Medicare towards treatment services. As the burden increases with the aging population, the unmet clinical need for prognostic PAD tools becomes more apparent. Vascular calcification complicates PAD interventional treatment and is predictive of increased morbidity and all-cause mortality; there is currently no calcification scoring system devised for the peripheral circulation to risk-stratify patients, predict PAD progression, or provide dosing information for noninvasive decalcification therapies.
Methods Twenty three calcified atherosclerotic plaque samples were removed via carotid endarterectomy. Samples underwent microCT scanning and subsequent biochemical extraction and spectrophotometric assay to directly measure the calcium weight. MicroCT images were analyzed with Fiji 2.0.0 software to obtain a volumetric estimate and a traditional estimate (based on the coronary calcification scoring method established by Agatson) of calcification burden. Score estimates were compared with biochemical measurements for validation. 
Results A strong positive correlation was observed between the measured, extracted calcium and both scoring methods, each yielding R2 values greater than 0.95. Such numbers affirm microCT-based estimates of vascular calcifications. The traditional score had an R2 value of 0.974 compared to the volumetric R2 value of 0.973, qualifying the traditional score as the most accurate and optimal scoring metric.  
Conclusions The traditional Agatston calcification score devised for the coronary circulation is applicable to the periphery. This project also investigated the scores’ accuracy when compared to direct vascular calcification measurements, which has never been done before. The mathematical correlation may be used to derive the calcium mass from a clinical CT; such information may then be used to determine drug dosage for decalcification treatments such as EDTA-loaded nanoparticles. Future directions include applying our methods to analyze clinical CTs and automating the scoring with machine learning.