From medical image processing to in-vivo mechanical characterization: A framework based on IVUS studies

Abstract

Cardiovascular diseases are the principal cause of mortality and morbidity worldwide mostly due to myocardial infarction and stroke. The understanding of the genesis, development and progression of such diseases is key for effective diagnosis, treatment and surgical risk assessment. Notorious advances have been performed in the histological characterization of culprit plaque for such events, although in-vivo techniques for tissue characterization still comprise an extremely active area of research. In this work, a framework is proposed targeting the in-vivo characterization of the arterial wall tissues. The set of methodologies involves: novel image processing methods for medical image enhancement (gating, registration and denoising of high frequency ultrasonic images) and optical flow estimation; detailed mechanical models for coronary arteries; and an effcient data assimilation method for tissue characterization. The thesis is structured in three parts: i) medical image processing; ii) material parameter estimation and iii) medical applications. Particularly, this work makes use of Intravacular Ultrasound (IVUS) as medical image acquisition technique, even though, the second part of the thesis is generic and can be straightforwardly extended to other imaging techniques. In the first part, different methods are presented to enhance and retrieve data of arterial vessel deformations and spatial description of anatomical structures. A novel gating method is proposed to obtain the vessel description at each instant along the cardiac cycle. Due to the intrinsic motion of the sensors during the image acquisition, we propose a registration method that corrects the sensor displacement in the transversal plane of acquisition and along the axis of the vessel. To improve the signal-to-noise ratio of the ultrasound, we propose a denoising method based on the speckle noise (ultrasound characteristic noise) statistics which outperforms classic denoising strategies. Using the three previous methods, we present a methodology to obtain the optical flow of the vessel cross-section during the whole cardiac cycle. In the second part, we scrutinize state-of-the-art literature about the arterial anatomy and mechanical behavior of the arterial wall with particular focus on coronary arteries. Hence, we describe the pathophysiology of the atherosclerosis and the mechanical alterations of the components of the tissues in affected vessels. Then, the tissue characterization problem is addressed by estimating the constitutive parameters of constitutive mechanical models for arterial tissues with a reduced-order unscented Kalman filter. Using the surveyed data and adequate constitutive models, the appropriate setup for the data assimilation problem is studied, and the capabilities of the proposed strategy for tissue estimation are assessed. Then, optical flow techniques are employed to characterize the tissues in-vivo. The third part of the thesis presents a side contribution related to the first part of this work, that is a multimodality comparison for the generation of geometric arterial models from medical images. Specfi cally, we compare coronary computed tomography angiography (CCTA) versus coronary angiography fused with intravascular ultrasound in terms of geometric descriptors and hemodynamic indexes derived from the geometric models. In such study the gating and registration techniques developed in the first part of the thesis are employed.