skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease

Abstract

A fully automated method for coronary calcification detection from non-contrast-enhanced, ECG-gated multi-slice computed tomography (CT) data is presented. Candidates for coronary calcifications are extracted by thresholding and component labeling. These candidates include coronary calcifications, calcifications in the aorta and in the heart, and other high-density structures such as noise and bone. A dedicated set of 64 features is calculated for each candidate object. They characterize the object's spatial position relative to the heart and the aorta, for which an automatic segmentation scheme was developed, its size and shape, and its appearance, which is described by a set of approximated Gaussian derivatives for which an efficient computational scheme is presented. Three classification strategies were designed. The first one tested direct classification without feature selection. The second approach also utilized direct classification, but with feature selection. Finally, the third scheme employed two-stage classification. In a computationally inexpensive first stage, the most easily recognizable false positives were discarded. The second stage discriminated between more difficult to separate coronary calcium and other candidates. Performance of linear, quadratic, nearest neighbor, and support vector machine classifiers was compared. The method was tested on 76 scans containing 275 calcifications in the coronary arteries and 335 calcifications inmore » the heart and aorta. The best performance was obtained employing a two-stage classification system with a k-nearest neighbor (k-NN) classifier and a feature selection scheme. The method detected 73.8% of coronary calcifications at the expense of on average 0.1 false positives per scan. A calcium score was computed for each scan and subjects were assigned one of four risk categories based on this score. The method assigned the correct risk category to 93.4% of all scans.« less

Authors:
; ; ;  [1];  [2];  [2]
  1. Image Sciences Institute, University Medical Center Utrecht, Utrecht (Netherlands)
  2. (Netherlands)
Publication Date:
OSTI Identifier:
20951162
Resource Type:
Journal Article
Resource Relation:
Journal Name: Medical Physics; Journal Volume: 34; Journal Issue: 4; Other Information: DOI: 10.1118/1.2710548; (c) 2007 American Association of Physicists in Medicine; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
62 RADIOLOGY AND NUCLEAR MEDICINE; AORTA; CALCIUM; CLASSIFICATION; COMPUTERIZED TOMOGRAPHY; CORONARIES; DIAGNOSIS; HEART; IMAGE PROCESSING; IMAGES; LABELLING; RISK ASSESSMENT; SKELETON

Citation Formats

Isgum, Ivana, Rutten, Annemarieke, Prokop, Mathias, Ginneken, Bram van, Radiology Department, University Medical Center Utrecht, Utrecht, and Image Sciences Institute, University Medical Center Utrecht, Utrecht. Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease. United States: N. p., 2007. Web. doi:10.1118/1.2710548.
Isgum, Ivana, Rutten, Annemarieke, Prokop, Mathias, Ginneken, Bram van, Radiology Department, University Medical Center Utrecht, Utrecht, & Image Sciences Institute, University Medical Center Utrecht, Utrecht. Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease. United States. doi:10.1118/1.2710548.
Isgum, Ivana, Rutten, Annemarieke, Prokop, Mathias, Ginneken, Bram van, Radiology Department, University Medical Center Utrecht, Utrecht, and Image Sciences Institute, University Medical Center Utrecht, Utrecht. Sun . "Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease". United States. doi:10.1118/1.2710548.
@article{osti_20951162,
title = {Detection of coronary calcifications from computed tomography scans for automated risk assessment of coronary artery disease},
author = {Isgum, Ivana and Rutten, Annemarieke and Prokop, Mathias and Ginneken, Bram van and Radiology Department, University Medical Center Utrecht, Utrecht and Image Sciences Institute, University Medical Center Utrecht, Utrecht},
abstractNote = {A fully automated method for coronary calcification detection from non-contrast-enhanced, ECG-gated multi-slice computed tomography (CT) data is presented. Candidates for coronary calcifications are extracted by thresholding and component labeling. These candidates include coronary calcifications, calcifications in the aorta and in the heart, and other high-density structures such as noise and bone. A dedicated set of 64 features is calculated for each candidate object. They characterize the object's spatial position relative to the heart and the aorta, for which an automatic segmentation scheme was developed, its size and shape, and its appearance, which is described by a set of approximated Gaussian derivatives for which an efficient computational scheme is presented. Three classification strategies were designed. The first one tested direct classification without feature selection. The second approach also utilized direct classification, but with feature selection. Finally, the third scheme employed two-stage classification. In a computationally inexpensive first stage, the most easily recognizable false positives were discarded. The second stage discriminated between more difficult to separate coronary calcium and other candidates. Performance of linear, quadratic, nearest neighbor, and support vector machine classifiers was compared. The method was tested on 76 scans containing 275 calcifications in the coronary arteries and 335 calcifications in the heart and aorta. The best performance was obtained employing a two-stage classification system with a k-nearest neighbor (k-NN) classifier and a feature selection scheme. The method detected 73.8% of coronary calcifications at the expense of on average 0.1 false positives per scan. A calcium score was computed for each scan and subjects were assigned one of four risk categories based on this score. The method assigned the correct risk category to 93.4% of all scans.},
doi = {10.1118/1.2710548},
journal = {Medical Physics},
number = 4,
volume = 34,
place = {United States},
year = {Sun Apr 15 00:00:00 EDT 2007},
month = {Sun Apr 15 00:00:00 EDT 2007}
}
  • The possibility of detecting mild coronary stenoses with positron computed tomography and nitrogen (N-13) ammonia administered during pharmacologic coronary vasodilation was previously demonstrated in chronically instrumented dogs. The feasibility of using this technique in human beings and its sensitivity in determining the degree and extent of coronary artery disease were examined in 13 young normal healthy volunteers and 32 patients with angiographically documented coronary artery disease. N-13 ammonia was administered intravenously and its distribution in the left ventricular myocardium recorded at rest and during dipyridamole-induced coronary hyperemia. In the 13 volunteers, N-13 activity was homogeneous at rest and during hyperemia,more » whereas 31 of the 32 patients had regional defects on the hyperemic images not present during rest. All six patients with double, all 10 with triple and 15 of 16 patients with single vessel disease (97 percent) were correctly identified with the technique. Two vessel involvement was correctly identified in five of the six patients with double vessel disease and three vessel disease in six of 10 patients. Of all 58 coronary stenoses, 52 (90 percent) were correctly identified. In a subgroup of 11 patients, the technique was compared with exercise thallium-201 planar images, which were abnormal in 10 (91 percent) whereas N-13 images were abnormal in all 11. Of the 19 stenosed coronary arteries in this subgroup, 11 (58 percent) were correctly identified with thallium-201 and 17 (89 percent) with tomography (p less than 0.01). It is concluded that cross-sectional imaging of the myocardial distribution of N-13 ammonia administered during pharmacologic coronary vasodilation is a highly sensitive and accurate means for noninvasive detection of coronary stenoses in human beings and for estimating the extent of coronary artery disease.« less
  • Thallium-201 myocardial imaging during dipyridamole-induced coronary hyperemia has been an accepted method for diagnosing coronary artery disease (CAD) and risk stratification. Adenosine is a powerful short-acting coronary vasodilator. Initial results of thallium imaging during adenosine infusion have been encouraging. In 132 patients with CAD and in 16 patients with normal coronary angiograms, adenosine was given intravenously at a dose of 0.14 mg/kg/min for 6 minutes and thallium-201 was injected at 3 minutes. The thallium images using single-photon emission computed tomography were abnormal in 47 of the 54 patients (87%) with 1-vessel, in 34 of 37 patients (92%) with 2-vessel andmore » in 40 of 41 patients (98%) with 3-vessel CAD. The sensitivity was 92% in the 132 patients with CAD (95% confidence intervals, 86 to 96%). In patients with normal coronary angiograms, 14 of 16 patients had normal thallium images (specificity, 88%; 95% confidence intervals, 59 to 100%). The results were very similar when subgroups of patients were analyzed: those without prior myocardial infarction, elderly patients and women. The nature of the perfusion defects (fixed or reversible) was assessed in relation to whether the 4-hour delayed images were obtained with or without the reinjection technique. In patients who underwent conventional delayed imaging, there were more fixed perfusion defects than in patients with reinjection delayed imaging (16 vs 0%, p less than 0.0001). The adverse effects were mild, transient and well tolerated. Thus, adenosine thallium tomographic imaging provides a high degree of accuracy in the diagnosis of CAD. The use of the reinjection technique enhances the ability to detect reversible defects.« less
  • In order to evaluate the detectability of coronary artery disease (CAD) with positron-emission computed tomography (PET), we performed /sup 13/N-ammonia myocardial PET scanning at rest and with exercise loading in 20 normal subjects and 40 patients with CAD, by means of a high-resolution, multi-slice, whole-body PET scanner. Myocardial PET scanning was performed 3 minutes after injection of /sup 13/N-ammonia at rest and during exercise. The circumferential profile analysis of resting PET images revealed regional hypoperfusion in 96% of CAD patients with previous myocardial infarction and in 29% of those without infarction. Exercise PET studies showed high sensitivity (93%) in detectingmore » CAD without myocardial infarction, whereas no abnormal hypoperfusion was detected in normal subjects. Segmental analysis of regional myocardial perfusion with exercise stress identified 67 of 75 stenosed vessels (89%). We conclude that /sup 13/N-ammonia myocardial PET with exercise loading provides high-quality tomographic images of regional myocardial perfusion and is a valuable technique for detecting CAD.« less
  • One hundred eight-three men underwent stress-redistribution thallium-201 myocardial perfusion tomography. After evaluation of various preprocessing filters in a phantom study, the Butterworth filter with a frequency cutoff of 0.2 cycles/pixel, order 5 (which provided optimal filter power) was used in the back projection algorithm of the patient studies. All short-axis and apical portions of vertical long-axis images were quantified by dividing each myocardial slice into 60 equal sectors and displaying the maximal count per sector as a linear profile. In a pilot group consisting of 20 normal men (less than 5% likelihood of coronary artery disease) and 25 men withmore » coronary artery disease (greater than or equal to 50% coronary stenosis by angiography), profiles representing the lowest observed value below the mean normal profiles provided the best threshold for defining normal limits. Abnormal portions of the patient profiles were plotted on a two-dimensional polar map. The polar map was divided into 102 sectors, and sectors with a probability of greater than or equal to 80% for disease of each one of the three major coronary arteries were clustered to represent specific coronary artery territories. Receiver operating characteristic curve analysis for defect size showed that the optimal threshold for defining a definite perfusion defect was 12% for the left anterior descending and left circumflex and 8% for the right coronary artery territories. These criteria were prospectively applied to an additional 92 patients with angiographic coronary artery disease, 18 patients with normal coronary arteriograms and 28 patients with less than 5% likelihood of coronary disease. Sensitivity, specificity and normalcy rate for overall detection of coronary disease were 96%, 56% and 86%, respectively.« less
  • The diagnostic value of maximal pharmacologic coronary vasodilation with intravenously administered adenosine in conjunction with thallium-201 single-photon emission computed tomography (SPECT) for detection of coronary artery disease was investigated in 101 consecutive patients who had concomitant coronary arteriography. Tomographic images were assessed visually and from computer-quantified polar maps of the thallium-201 distribution. Significant coronary artery disease, defined as greater than 50% luminal diameter stenosis, was present in 70 patients. The sensitivity for detecting patients with coronary artery disease using quantitative analysis was 87% in the total group, 82% in patients without myocardial infarction and 96% in those with prior myocardialmore » infarction; the specificity was 90%. The sensitivity for diagnosing coronary artery disease in patients without infarction with single-, double-and triple-vessel disease was 76%, 86% and 90%, respectively. All individual stenoses were identified in 68% of patients with double-vessel disease and in 65% of those with triple-vessel disease. The extent of the perfusion defects, as quantified by polar maps, was directly related to the extent of coronary artery disease. In conclusion, quantitative thallium-201 SPECT during adenosine infusion has high sensitivity and specificity for diagnosing the presence of coronary artery disease, localizing the anatomic site of coronary stenosis and identifying the majority of affected vascular regions in patients with multivessel involvement.« less