Identification of coronary calcifications in optical coherence tomography imaging using deep learning

Coronary calcifications are an obstacle for successful percutaneous treatment of coronary artery disease patients. The optimal method for delineating calcifications extent is coronary optical coherence tomography (OCT). To identify calcification on OCT and subsequently tailor the appropriate treatment, requires expertise in both image acquisition and interpretation. Image acquisition consists from system calibration, blood clearance by a contrast agent along with synchronization of the pullback process. Accurate interpretation demands careful review by the operator of a segment of 50–75 mm of the coronary vessel at steps of 5–10 frames per mm accounting for 375–540 images in each OCT run, which is time consuming and necessitates some expertise in OCT analysis. In this paper we developed a new deep learning algorithm to assist the physician to identify and quantify coronary calcifications promptly, efficiently and accurately. Our algorithm achieves an accuracy of 0.9903 ± 0.009 over the test set at size of 1500 frames and even managed to find calcifications that were not recognized manually by the physician. For the best knowledge of the authors our algorithm achieves high accuracy which was never achieved in the past.

Postmortem coronary artery calcium score in cases of myocardial infarction

Sudden cardiac death (SCD) related to atherosclerotic coronary artery disease (ACAD) resulting in myocardial infarction is the most prevalent cause of death in western countries. In clinical practice, coronary artery calcium score (CACS) is considered an independent predictor of coronary events, closely related to atherosclerotic burden and is quantified radiologically by the Agatston score being calculated through computed tomography. Postmortem computed tomography (PMCT) allows the visualization and quantification of coronary calcifications before the autopsy. However, it was reported that some patients who died from severe ACAD had a zero CACS in PMCT. In this study, a retrospective evaluation of CACS in adult’s myocardial infarction cases related to ACAD, with available CACS and histological slides of coronary arteries, was performed in order to gain a deeper understanding of coronary calcifications and their role in myocardial infarction cases. The CACS was calculated by using the software Smartscore 4.0 after the radiological examination on a 64-row CT unit using a specific cardiac protocol. Thirty-six cases were identified out of 582 autopsies, recorded during a 2-year study period (29 men, 7 women; age 56.3 ± 11.7). CACS was 0–10 in 5 cases (5 men, 44.8 ± 13.7), 11–100 in 8 cases (6 men, 2 women, 53.1 ± 7.7), 101–400 in 13 cases (11 men, 2 women, 57.4 ± 9.6), and > 400 in 10 cases (9 men, 1 woman, 63.1 ± 11.9). Coronary thrombosis was found in 28 cases, histologically identified as plaque erosions in 6 cases and as plaque ruptures in 22 cases. Statistical analyses showed that CACS increases significantly with age (p-value < 0.05) and does not show significant correlation with gender, body weight, body mass index, and heart weight. CACS was significantly higher in plaque ruptures than in plaque erosions (p-value < 0.01). Zero or low CACS on unenhanced PMCT cannot exclude the presence of myocardial infarction related to ACAD. This paradoxical discrepancy between imaging and autopsy findings can be explained considering the histological aspect of fatal coronary plaques.

Identification of Coronary Calcifications in Optical Coherence Tomography Imaging Using Deep Learning

Coronary calcifications are an obstacle for successful percutaneous treatment of coronary artery disease patients. The optimal method for delineating calcifications extent is optical coherence tomography (coronary OCT). To identify calcification on OCT and subsequently tailor the appropriate treatment, requires expertise in both image acquisition and interpretation. Image acquisition consists from system calibration, blood clearance by a contrast agent along with synchronization of the pullback process. Accurate interpretation demands careful review by the operator of a segment of 50-75mm of the coronary vessel at steps of 0.5-1mm accounting for 75-100 images in each OCT run, which is time consuming and necessitates some expertise in OCT analysis.In this paper we developed a new deep learning algorithm to assist the physician to identify and quantify coronary calcifications promptly, efficiently and accurately. Our algorithm achieves an accuracy of 0.9903 ± 0.009 over the test set at size of 1500 frames and even managed to find calcifications that weren’t recognized manually by the physician. For the best knowledge of the authors our algorithm achieves high accuracy which was never achieved in the past.

Abstract 13132: Reclassification of CHA 2 DS 2 -VASc Thromboembolic Risk by Coronary Calcifications on Cardiac Computed Tomography in Atrial Fibrillation

Background: CHA 2 DS 2 -VASc is the most widely used thromboembolism risk score in patients with atrial fibrillation (AF). Cardiac computed tomography (CCT) routinely performed before or after pulmonary vein isolation (PVI) for AF rhythm control offers the opportunity to detect coronary calcifications (CAC) and this vascular pathology. We evaluated the frequency of CAC and the extent it affects the CHA 2 DS 2 -VASc and decision for anticoagulation for AF patients undergoing PVI. Methods: In 2014, 772 consecutive patients underwent PVI at Cleveland clinic, and 621 patients who had CCT within 1-year before or after PVI were studied. Relationships between CAC recorded as a binary variable with clinical characteristics and reclassification of CHA 2 DS 2 -VASc was analyzed. Results: Mean age was 63.1±9.8 years, 163 (26.2%) were females, 322 (51.9%) had paroxysmal AF and 264 (42.5%) having prior PVI. CAC was identified on CCT in 388 (62.5%) patients. Age was the only factor independently associated with CAC, odds ratio 1.02 (95% confidence interval 1.01-1.04), P=0.004. CAC increased the CHA 2 DS 2 -VASc in 306 (49.3%) patients, and the mean from 2.0±1.5 to 2.5±1.4 (Table). Using gender-specific cutpoints from the latest guidelines, 71 (11.4%) had CHA 2 DS 2 -VASc going from 0 to 1 in men or 1 to 2 in women (where anticoagulation may be considered), and 113 (18.2%) had CHA 2 DS 2 -VASc going from 1 to 2 in men or to 3 in women (where long-term anticoagulation is indicated). Conclusion: Almost two-thirds of patients undergoing PVI have CAC detectable on CCT, and after incorporating this information, long-term anticoagulation may or would be indicated in an additional 30% of the cohort. By assessing vascular pathology, CCT can play an important screening role for thromboembolic risk in AF patients incremental to clinical risk factors.