Including in-vivo IVUS data in ex-vivo validated dataset

Intravascular Ultrasound (IVUS) data validation is usually performed by comparing postmortem (ex-vivo) IVUS data and corresponding histological analysis of the tissue, obtaining a reliable ground truth. The main drawback of this method is the few amount of available training dataset. In this work we propose a method to include examples from in-vivo labeled plaque to the ex-vivo training dataset. In-vivo plaques selection is performed by Sequential Floating Forward Selection (SFFS) algorithm and results are validated by the Leave-One-Patient-Out crossvalidation technique. Classification task is performed by using AdaBoost exploiting the Error Correcting Output Code (ECOC) technique to infer the final decision. New data inclusion improves global classification accuracy from 89.56% to 91.77%.