Kimberly Hsu, Freshman Bioengineering student
Developing Classification Algorithms to Provide Real-time Data for Spectroscopic Tools
with Dr. Rebekah Drezek
Research in Dr. Drezek's laboratory is focused on the development of novel spectroscopic technologies for in situ disease detection and monitoring, with particular emphasis on cancer applications. Projects includes both development of imaging tools to sense endogenous optical signals which vary with disease progression and development of tools to be used in conjunction with molecular-specific contrast agents.
Kimberly’s specific project focuses on work towards development of classification algorithms which could ultimately provide real-time data to clinicians or surgeons using the spectroscopic tools currently under development in our laboratory. Algorithms for automated assessment of in vivo spectroscopic data have been most typically accomplished via statistical approaches such as principal component analysis. Although promising classification results have been achieved in a number of organ sites using statistically driven algorithms, we believe algorithms based on understanding the basic mechanisms which cause light to interact differently with normal and neoplastic tissue may ultimately offer a more sensitive and specific diagnostic approach. Moreover, elucidating the relationships between tissue biochemistry and morphology and optical spectra should help facilitate translation of technology developed for the ovary to the detection of other epithelial neoplasms. In addition, as we begin to quantitatively understand the biochemical and biophysical nature of the optical spectra we measure, optical spectroscopy will progress from a novel tool for the detection of precancers to a method increasingly relevant to fundamental studies of cancer biology, offering a non-invasive means to monitor a variety of processes such as angiogenesis, extracellular matrix degradation, adhesion, and epithelial-stromal communication. Such studies will advance present knowledge of the development of cancer, a health problem which notwithstanding significant medical advances over the past fifty years, remains the second leading cause of death in the United States as we enter the new millennium.
In particular, despite extensive research in new diagnostic and treatment methods for ovarian cancer, we have not changed the overall survival in the last 50 years. Kimberly’s project has focused on algorithm development for the interpretation of in vivo fluorescence spectra. The development of an inverse model of ovarian tissue fluorescence can proceed in several different directions. A pure analytical model is easiest to invert and could be achieved using single-scattering or diffusion theory approaches. However, the validity of either of these two theories must be rigorously demonstrated for our specific measurement conditions (probe geometry, ovarian tissue optical properties over the spectral range of interest).
Using MATLAB, Kimberly has developed an initial diffusion theory model which provides a mean to accurately fit measured data. Sample results are shown below. She will soon be beginning the process of validating her code using Monte Carlo simulations of photon transport.
The derivation of an appropriate analytic expression for remitted fluorescence is lengthy but not complicated. The far greater challenge lies in assessing which parameters in the model will provide the most useful diagnostic information in order to determine which terms will be fixed and which will be free parameters. Our preliminary results indicate some of the most relevant parameters include the intrinsic NAD(P)H and collagen contributions which are modulated by both by the disease process and menopausal status. We are optimistic that we will be able to provide point-of-care quantification of these endogenous signals as this project matures.
Department of Bioengineering
Back to Undergraduate Research page
