The functional performance of membrane oxygenators is directly related to the perfusion dynamics of blood flow through the fiber bundle. Non-uniform flow and design characteristics can limit gas exchange efficiency and influence susceptibility of thrombus development in the fiber membrane. Computational fluid dynamics (CFD) is a powerful tool for predicting properties of the flow field based on prescribed geometrical domains and boundary conditions. Validation of numerical results in membrane oxygenators has been predominantly based on experimental pressure measurements with little emphasis placed on confirmation of the velocity fields due to opacity of the fiber membrane and limitations of optical velocimetric methods. A novel approach was developed using biplane X-ray digital subtraction angiography to visualize flow through a commercial membrane artificial lung at 1–4.5 L/min. Permeability based on the coefficients of the Ergun equation, α and β, were experimentally determined to be 180 and 2.4, respectively, and the equivalent spherical diameter was shown to be approximately equal to the outer fiber diameter. For all flow rates tested, biplane image projections revealed non-uniform radial perfusion through the annular fiber bundle, yet without flow bias due to the axisymmetric position of the outlet. At 1 L/min, approximately 78.2% of the outward velocity component was in the radial (horizontal) plane verses 92.0% at 4.5 L/min. The CFD studies were unable to predict the non-radial component of the outward perfusion. Two-dimensional velocity fields were generated from the radiographs using a cross-correlation tracking algorithm and compared with analogous image planes from the CFD simulations. Velocities in the non-porous regions differed by an average of 11% versus the experimental values, but simulated velocities in the fiber bundle were on average 44% lower than experimental. A corrective factor reduced the average error differences in the porous medium to 6%. Finally, biplane image pairs were reconstructed to show 3-D transient perfusion through the device. The methods developed from this research provide tools for more accurate assessments of fluid flow through membrane oxygenators. By identifying non-invasive techniques to allow direct analysis of numerical and experimental velocity fields, researchers can better evaluate device performance of new prototype designs.

Over the last decade, there has been a growing interest in assessing cerebral aneurysmal wall motion, because of its potential connections to the biomechanical conditions of the vessel wall, which could eventually aid the prediction of aneurysmal rupture risk. Such quantification could provide a valid surrogate for the vascular wall status and integrity. However, the vast majority of current morphological indices used in the literature to predict growth and rupture in cerebral aneurysms do not take into account the temporal changes that occur during the cardiac cycle. This is because these indices are derived from image modalities that do not provide sufficient temporal and/or spatial resolution to obtain dynamic aneurysm information, which is expected to be similar to or below image resolution. Among currently available vascular imaging techniques, 3D rotational angiography (3DRA) and digital subtraction angiography (DSA) have the highest spatial (and temporal) resolution. Still, for a human operator relying solely on qualitative visual observation, even when using images from these modalities, to objectively analyze the small motion and shape changes of the cerebrovasculature of an individual throughout the cardiac cycle is difficult, if not impossible. Therefore, the availability of a robust morphodynamic analysis tool is needed. In this context, this thesis focuses on developing techniques to estimate, quantify and analyze cerebrovascular wall motion, particularly aneurysmal wall motion, using such modalities. The main contributions of the thesis are: 1) a first methodology to estimate and model patient-specific cerebrovascular morphodynamics over one cardiac cycle, through a proposed multiple 2D to 3D image registration framework; 2) an extension of this methodology to provide robust and efficient estimates of cerebrovascular wall motion for clinical evaluation and for further biomechanical modeling of the cerebrovascular wall; 3) a patient study that demonstrates the validity of the developed techniques from clinical practice, through an analysis of 3DRA and DSA images. Each of these contributions is published in or submitted to a peerreviewed international journal. / Durante la última década se ha dado un creciente interés en la evaluación del movimiento de la pared vascular en aneurismas cerebrales. Éste hecho ha sido motivado en gran medida por la relación existente entre dicha motilidad y sus condiciones biomecánicas, pudiendo éstas llegar a ser útiles en la predicción del riesgo de ruptura del aneurisma cerebral analizado. De este modo, de ésta cuantificación, se podría llegar a derivar un indicador indirecto del estado e integridad de la pared vascular. Sin embargo, la gran mayoría de los índices morfológicos utilizados en la actualidad para predecir crecimiento y ruptura de aneurismas cerebrales no consideran los cambios que se producen en el tiempo a lo largo del ciclo cardíaco. Esto se debe a que dichos índices se obtienen a partir de modalidades de imagen que no proporcionan suficiente resolución espacial y/o temporal para obtener información dinámica del aneurisma, cuyo rango de variación se espera sea similar o inferior a la resolución de la imagen. Entre las técnicas de imagen vascular disponibles en la actualidad, la angiografía rotacional 3D (3DRA) y la angiografía de substracción digital (DSA) son las que ofrecen la mayor resolución espacial (y temporal). De todos modos, aún utilizando imágenes de estas modalidades, el análisis objetivo de pequeñas diferencias de forma y movimiento en los vasos cerebrales de un individuo a lo largo de un ciclo cardíaco es difícil, si no imposible para un operador humano utilizando únicamente medidas cualitativas guiadas por inspección visual. Por lo tanto, la disponibilidad de herramientas robustas para el análisis morfodinámico de la vasculatura cerebral resulta necesaria. En este contexto, la investigación de esta tesis se concentra en el desarrollo de técnicas para estimar, cuantificar y analizar el movimiento de las paredes de los vasos cerebrales, con particular énfasis en el movimiento de la pared en aneurismas, utilizando las modalidades indicadas anteriormente. En líneas generales, esta tesis presenta tres contribuciones principales: 1) una primera metodología de estimación y modelado morfodinámico de vasos cerebrales a lo largo de un ciclo cardíaco, utilizando una técnica de registrado de imágenes 2D-3D; 2) una metodología extendida para proporcionar una estimación robusta y eficiente del movimiento de las paredes de los vasos cerebrales para su evaluación clínica y posterior modelado biomecánico de dichas paredes; 3) un estudio sobre una población de pacientes que demuestra la validez de las técnicas desarrolladas en la práctica clínica, a través del análisis en imágenes de 3DRA y DSA. Cada una de estas contribuciones ha sido publicada o se encuentra en fase de revisión en revistas internacionales indexadas.