Transcript Document
Diagnostic Decision Making using High Frequency Bioresponses and Medical Imaging Wavelet-based 3-D MFS in BMRI Project with Dean, Park, and Ziegler (Div. Pulmonary and Critical Care Med. Emory). Project with CBIS and Emory (Winship Cancer Institute) Communicated at ISBRA 2008, and part of NIH grant proposal 2008 About BESTA Multifractal Analysis of H-NMR Aim: To classify BMRI images to benign and malignant using wavelet-based multifractal spectrum (MFS) of the image background Preliminary results published Journal of Data Science 2008 Aim: To connect fractality descriptors to measures of sulfuramino acid (SAA) deficiency (cysteine) Aims Description of Data The center aims are to promote research and consulting in all aspects involving the planning of statistical experiments and statistical modeling of results, with an emphasis on biomedical data. Members Case Control Wavelet spectrum/Analysis Melinda Higgins Sky Lee • Descriptors and realizations of multifractal spectrum Xavier Le Faucheur In NMR spectra, a wealth of information is ignored. From the resolution of tens of thousands of metabolites, traditional analysis focuses on a few peaks. The idea is to look at the spectrum as a (multi)fractal and summarize (multi)fractal properties. Brani Vidakovic Hin Kyeol Woo Lucy Petrova • The extended three dimensional concept of wavelet-based multifractal spectrum is used in classification of BMRI Karan Raturi Contact Us BESTA - Center for Bioengineering Statistics Wallace Coulter Department of Biomedical Engineering Georgia Institute of Technology 1213 Whitaker Building. Atlanta, GA 30332 Wavelets on Surfaces Wavelet Enhancement of Mammograms Xavier Le Faucheur (joint with Delphine Nain, Allen Tannenbaum, and Brani Vidakovic) Project with Dubois Bowman (RSPH, Emory) Preliminary results published in SPIE 6763, 2007 Communicated at ISBRA 2007 and Georgia Cancer Coalition seed grant award Shape signal (x,y,z) Encoding using Spherical Wavelets • Wavelet Image Interpolation (WII) is a wavelet-based approach to enhancement of digital mammography images. Wavelet Coefficients Error Bayesian Wavelet Shrinkage using Shape Features Shrunk Wavelet Coefficients Smooth Shape Signal Inverse Wavelet Transform Procedure Original Shape VOCs Detecting Breast Cancer Noisy Shape 0.9 0.8 0.7 0.6 • WII involves the application of an inverse wavelet transformation to a coarse or degraded image and constructed detail coefficients to produce an enhanced higher resolution image. 1. One performs k wavelet decomposition steps on empty image. The transform is linear and the resulting smooth and detail sub-matrices are all zero-matrices. 2. The degraded image from a digital mammogram is inserted into the position of the smooth matrix containing zeros. 0.5 0.4 3. The object in 2 is back-transformed by k steps. 0.3 0.2 0.1 Recovered Shape after Shrinkage Squared Error This process increases the resolution of the degraded image and contains 4 k times the number of pixels in the original input. Project with Charlene Bayer (GTRI), Sheryl GabramMendola (Winship), and Boris Mizaikoff (University of Ulm) Aim: To diagnose subject with cancer based on the VOC (Volatile Organic Compound) content of their breath Description of Data: 383 VOCs per subject; 35 subjects (24 controls, 11 cases) Dimension Reduction/Analysis • Dimension reduction from 383 VOCs to 2 informative components • Nonlinear dimension reduction is very discriminative