[Analysis of the diagnostic value of endovascular ultrasound with roc curves. Practical implications]

Arch Mal Coeur Vaiss. 1997 Jan;90(1):59-66.
[Article in French]

Abstract

Intravascular ultrasound enables detection of the components of atherosclerotic plaques. The diagnostic value was assessed by ROC (receiver operating characteristic) curves on images acquired in vitro and correlated with the histological findings in 61 arteries. Five questions were asked of each operator; the reply was represented by a continuous variable in order to express all nuances of judgement. The area under the ROC curve, Az, was the criterion of performance (0.5 : chance response : 1.0 : all replies were accurate). Detection of plaque was satisfactory (Az = 0.89). The three layer appearance of muscular arteries was well recognised (Az = 0.94). The fibrous composition of a plaque was only just satisfactory (Az = 0.88) with 38.7% interindividual variability. The lipid composition of the plaque was poorly recognised (Az = 0.76) with large interindividual variability (52.8%) : hypoechogenicity was too ambiguous a sign from the acoustic point of view. A hypoechogenic zone must not be synonymous with a lipid plaque but a cellular zone. Calcium can almost always be detected (Az = 0.98) with a very low interindividual variability (10.7%), fibrohyaline progression of some plaques can be confusing. The authors present a more objective description of endovascular ultrasonographic images. They conclude that the diagnostic performance of 30 MHz intravascular ultrasound is satisfactory but several limitations are apparent in the interpretation of images, especially hypoechogenic zone and hyper-reflective zones with high attenuation.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Arteries / diagnostic imaging
  • Arteries / pathology
  • Arteriosclerosis / diagnosis*
  • Calcinosis / pathology
  • Coronary Vessels / diagnostic imaging
  • Coronary Vessels / pathology
  • Fibrosis / pathology
  • Humans
  • Image Processing, Computer-Assisted
  • In Vitro Techniques
  • Lipids / analysis
  • Models, Biological
  • Muscle, Smooth, Vascular / pathology
  • Observer Variation
  • ROC Curve*
  • Sensitivity and Specificity
  • Ultrasonography, Interventional* / methods

Substances

  • Lipids