A New Paradigm for the Personalized Delivery of Iodinated Contrast Material at Cardiothoracic, Computed Tomography Angiography

In North America more than 40 million doses of iodinated X-Ray contrast medium are delivered to patients undergoing CT imaging every year. This particular pharmaceutical is necessary to enable Computed Tomography of soft tissue, tumors, and vasculature. Very few of the contrast enhanced procedures are performed with the dose of the drug tailored to the individual patient or procedure and nearly every patient receives the same dose of contrast material. This dissertation presents a methodology to allow the routine administration of a personalized dose of contrast material to generate contrast enhancement sufficient for diagnosis during cardiothoracic CT Angiography imaging. Parameter estimation of a patient specific model is performed using Maximum Likelihood Estimation (MLE) with data generated from the scanner during a pre-diagnostic "test" injection of contrast agent. A non-parametric system identification technique, using the truncated Singular Value Decomposition, is also developed for deriving a patient specific prediction of contrast enhancement. The MLE technique produces contrast enhancement predictions with less error than the tSVD method. It is also shown that the MLE method is less sensitive to data length and has greater noise immunity. A novel, patient-specific contrast protocol generation algorithm is also presented. It is based upon a constrained minimization (Sequential Quadratic Programming) that enforces constraints on the input parameters while minimizing the volume of contrast sufficient to achieve a prospectively chosen enhancement target. A physiologically based pharmacokinetic (PBPK) numeric model is developed and used to validate the contrast prediction and protocol generation techniques. Finally, a novel, instrumented, flow phantom is developed and used to validate the identification and protocol generation techniques.