Abstract 151: Optical Coherence Tomography post-processing towards an automated quantification of coronary artery wall alteration subsequent to Kawasaki Disease

Abstract

Introduction: Statistical mathematical processing of medical imaging promotes operator-independent interpretation. We sought to identify discriminative image features for a histological disambiguation of KD related coronary (CA) lesions. The ultimate goal is to develop a fully automated algorithm for the quantification of the degradation/healing state of the CA wall structure post KD. Methods: We analyzed OCT CA recordings from two KD patients with (KD + AN + ) and two patients with no history of CA aneurysm (KD + AN [[Unable to Display Character: –]] ), vs a non KD patient (KD [[Unable to Display Character: –]] ). Results: In KD [[Unable to Display Character: –]] , regardless of the radial region of interest (ROI) position in the image, the OCT mean signal intensity presents centrifugally two peaks corresponding to the intima and media, constantly separated (17.1± 2.0 pixels) (fig. 1A). In KD + AN [[Unable to Display Character: –]] the peaks may disappear (Fig. 1-B left medial hyperplasia) and the distance between remaining crests vary between ROIs (22.7±6.9 Pixels). In KD + AN + these peaks disappear (Figure1-C) and the signal intensity changes drastically between ROIs due to wall restructuration. Figure1-C shows a hatched signal that maybe that of a laminar structure. In this case the variation of gradient intensity may be a discriminative feature. Conclusion: Mathematical modeling of CA wall layers is feasible. While the consistency of the distance between media and intima peaks may discriminate KD [[Unable to Display Character: –]] from KD + and while the gradient intensity may detect restructuration in KD + AN + , ongoing investigation to discriminate CA lesions include signal homogeneity, energy and contrast. Texture analysis with anatomical correlates (e.g., calcium, fibrosis and clots) may allow automated diagnoses.