Additive prognostic value of aortic arch calcification of chest X-ray and feasibility of machine learning algorithm on cardiovascular outcome; retrospective study

Abstract

Cardiovascular death is one of most common cause of death in the world. Traditional cardiovascular prognosis is based on Framingham risk scoring system or ACC/AHA Pooled Cohort Equation. Though those are excellent and convenient scoring system, we cannot deny that traditional calculation is not intuitive. Many studies revealed the relationship between aortic calcification and major cardiovascular events. Aortic arch is one of the most vulnerable segment in aorta against arterial pressure and arteriosclerosis. Although most of us agree with that aortic arch calcification on simple chest x-ray would be valuable, owing to its low reproducibility and validity of semi-quantative grading system, research focusing on aortic arch calcification and simple chest x-ray has not rigorously performed yet. In this study, we examines clinical implication of aortic arch calcification on chest x-ray to cardio-cerebrovascular outcome. After revealing its clinical importance and usefulness, then we developed machine learning algorithm to grading of aortic arch calcification on simple chest x-ray. Study population were collected patients who underwent carotid Doppler ultrasound at Gangnam Severance Hospital from 2009 March to 2012 February. Till now, total 3,080 patients were reviewed. Among them, 2,273 patients were finally enrolled in the study. Aortic arch calcification grade on chest x-ray was significantly correlated with presence of carotid artery plaques and pulse wave velocity, which represents arterial stiffness. CVA and all-cause death were significantly associated with aortic arch calcification grade on chest x-ray. The rate of admission due to heart failure aggravation was also highly related in patients whose aortic arch calcification grade was 3. In contrast, treatment with PTCA or any composite CVE was neither associated with aortic arch calcification grade on chest x ray. Predictive value of aortic arch calcification grade was also notable in case of CVA, all-cause death and some cases of admission rate due to heart failure aggravation. Through this study, we recognized that arteriosclerosis contributes to aortic arch calcification and its mechanism of action is different from atherosclerosis of coronary artery. In addition, we can also assume that aortic arch calcification grade has additive predictive value in addition to FRS for cardiovascular outcome. Further study would be elaborated to deep learning algorithm we developed so that clinicians can be instantly warned for risk of cardio-cerebrovascular outcome when they checked chest x-ray of patients.