An objective method for grading ultrasound images of carotid artery plaques

Plaque composition may improve identification of patients at risk of stroke. A new method of grading ultrasound (US) images to assess plaque composition is described. B-mode US images were obtained from 50 carotid specimens. Image analysis parameters were entered into a discriminant analysis package and compared retrospectively with histology. Discriminant functions were derived and then applied prospectively to image-analysis data obtained from a further 50 plaque specimens. For the prospective analysis, US images were graded according to the relative contribution of calcium, fibrous tissue, haemorrhage and lipid. The accuracy for retrospective classification of calcium was 100%, for fibrous tissue 97%, for lipid 76% and 68% for haemorrhage (kappa = 0.81). Prospective classification showed an overall agreement of 65% (kappa = 0.47). Significant intraplaque haemorrhage was identified with an 81 % sensitivity and 83 % specificity. The US method described demonstrated improved accuracy compared to previous studies. Further study is required to establish the use of this method for in vivo images and its correlation with patient symptoms. (E-mail: soundrie.padayachee @kcl.ac.uk) (C) 2001 World Federation for Ultrasound in Medicine & Biology.