Application of neural networks for the analysis of intravascular ultrasound and histological aortic wall appearance-an in vitro tissue characterization study

Abstract

The role of tissue characterization by intravascular ultrasound (IVUS) imaging of the aortic wall has not been well established. The artificial neural networks (ANNs) are a promising tool for image classification. The aim of the study was to assess the texture correlation between matching IVUS and histologic images of the aortic wall. The computer-based discrimination of pathology within the data sets was also evaluated. In vitro IVUS images and histologic sections from 36 aortic segments were compared using texture parameters that produced the best correlation or the highest discriminative value. The images were classified as normal or abnormal with variable degrees of pathology. Tissue characterization was performed by a nearest neighbor classifier, linear discriminant analysis (LDA) and the ANN-based approach. Good agreement was observed between IVUS and the histologic reference with a correlation coefficient of r = 0.89, r = 0.76 and r = 0.71 for the three most successful texture parameters. The ANN-based approach was the most effective in discriminant analysis, with a correct classification rate of 87.5% for histologic images and 79.2% for IVUS data. The study shows that ANNs are a potentially effective tool for assessment of IVUS aortic images.