Automated coronary plaque characterisation with intravascular ultrasound backscatter: ex vivo validation

Atherosclerosis is considered both a systemic and focal disease. Current diagnostic tools do not allow adequate in vivo identification and characterisation of lesions. Advanced spectral analysis of IVUS backscatter has displayed the potential for real-time plaque characterisation. The aim of this study is to determine the ex vivo accuracy of automated plaque characterisation by spectral analysis of intravascular ultrasound (IVUS) backscatter.

Plaques (n=184) from 51 coronary arteries were imaged by IVUS. The arteries were then pressure fixed and matching histology collected. Regions were selected from histology and corresponding IVUS data were used to build the plaque classification system using spectral analysis and classi-fication trees. Tissue-maps were validated ex vivo by comparison with histology via 899 selected regions (n=94 plaques) that comprised 471 fibrous tissue (FT), 130 fibro-fatty (FF), 132 necrotic-core (NC) and 156 dense-calcium (DC) regions. The overall predictive accuracies were 93.5% for FT, 94.1% for FF, 95.8% for NC, and 96.7% for DC with sensitivities and specificities ranging from 72% to 99%. The Kappa statistic was calculated to be 0.845 indicating very high agreement with histology.

Automated spectral analysis of IVUS backscatter provides accurate ex vivo information on plaque composition, with considerable potential for assessment of plaque vulnerability in real-time.