Fall Seminar Series  October 4, 2007
University of Minnesota
School of Statistics
College of Liberal Arts

Statistical Methods for Indirect Estimation of Physiological Parameters: Case Studies in Viral Kinetics

Brian Caffo
Department of Biostatistics
Johns Hopkins Univeersity

Thursday, October 4, 2007
3:30 PM, 115 Ford Hall
Minneapolis, East Bank Campus
Social at 3:00 PM, 300 Ford Hall


Abstract

In this talk we consider statistical methods for indirect estimation of physiological parameters. The problems arose from investigations of the ability to isolate the location and mechanisms of infection of the human immunodeficiency virus (HIV). Single photon emission computed tomography (SPECT) was used to image a radiolabeled lubricant in the colon after having been displaced by forces similar to those in anal intercourse. The benefit of SPECT imaging was the ability to image the tracer/lubricant mixture, rather than the largely irrelevant anatomy that other imaging modalities would highlight.  A novel application of fitting three dimensional statistical curves via a modified principal curve algorithm was implemented to solve the relevant scientific problem. The algorithm will be shown to have been tested and debugged on a battery of challenging two dimensional shapes. Potential extensions of the methodology to brain imaging will be discussed. Time permitting, a second complimentary application will be investigated that involves indirect estimation of the distribution of the HIV virus within the male genital tract. All examples will be geared towards a general audience without specialized statistical or medical background knowledge.