Tracer kinetic modeling of tumor microcirculation
Date of Issue2016
School of Electrical and Electronic Engineering
This thesis focuses on the investigation of Dynamic Contrast-Enhanced (DCE) based parametric assessment methods for tumor characterization, diagnosis and prognosis. In the recent few decades, extensive studies have been done on parametric assessments of tumor microcirculation characteristics. The introduction of DCE technique further enables the determination of relevant tumor microcirculation parameters, such as blood flow, blood volume, and capillary permeability. Early empirical model-free methods using DCE imaging are fast and simple in parameter assessment but their theoretical ground is insufficient. To model tracer concentration-time curves in a more systematic way, tracer kinetic models have been developed. However, practical applications of existing models are constrained by imprecise mathematical representation, unreliable parameter estimation, and high computational complexity. To addresses these issues, the research reported in this study proposes novel solutions using theoretical, numerical, and experimental techniques. Firstly, an Infinite-Pathways Distributed Parameter (IPDP) model was developed. It originates from the multiple-pathway model by Bassingthwaighte et al. and inherits the advantage of blood flow variability assessment. Through reconstructing the mathematical representation of the vasculature network, the IPDP model eliminates the discontinuities and normalization errors in multiple-pathway model, and at the same time relieves the dependency on hardware performance. The reconstruction of mathematical representation also enables in vivo assessment of blood flow distribution with DCE imaging. The IPDP model was assessed by applying on patient data with cerebral tumors. Estimation results clearly show the properties of capillary-tissue units and distinguish tumor tissues from normal tissues effectively. Secondly, a study using the proposed IPDP model was made to evaluate the performance of empirical model-free methods. The results obtained show that the accuracy of empirical methods is condition-dependent and some conditions may be difficult to meet in practice. The results also suggest that the peak enhancement metrics are not linearly related to any intrinsic tissue parameters. Thirdly, an improvement was made on the reliability of parameter estimation using the IPDP model. Studies have shown that the model-fitting process is sometimes trapped in a local minimum that is far from the global minimum, and it generates undesired results. A dual-phase model-fitting method was proposed to drive the model-fitting process towards global minimum for tracer enhancement curves with a washout pattern. Numerical results show that the coefficient of variation using the proposed approach can be reduced by up to 83% and the estimation bias is also alleviated in general. The proposed approach is applicable in blood flow dominated cases and could be implemented in existing tracer kinetic models for more accurate and stable parameter extraction. Finally, an attempt was made to reduce the computational complexity of the IPDP model. Specifically, a simplified Infinite-Pathway Conventional Compartmental (IPCC) model was developed by incorporating a conventional two-compartment model into each of the multiple pathways. The IPCC model was used to generate perfusion parameter maps for three DCE computed tomography patient cases to study its clinical applicability. The simplified IPCC model takes approximately 0.1s to generate an impulse response function compared to the 81s processing time by the original multiple-pathway software. Results also show that the simplified IPCC model has better fitting performance than the standard two-compartment model in most cases. In summary, the research conducted has led to the proposal of advanced tracer kinetic models. IPDP model improves the mathematic representation of the existing multiple-pathway kinetic models. Using the IPDP model, an in-depth study has been conducted on the performance and physiological implications of existing empirical model-free methods. Furthermore, the dual-phase model-fitting method and the IPCC model were proposed to reduce the coefficient of variations, estimation bias, and computational complexity of the IPDP model. The work reported in this thesis has offered desirable tracer kinetic models, modeling approach and algorithm for the characterization and diagnosis of cerebral tumor.