Derivatives of Raman Spectra for Breast Cancer Diagnosis

• Activity Number: 359
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Richard Charnigo*+ and Jing Guo and Cidambi Srinivasan and Ramachandra Dasari and Maryann Fitzmaurice and Abigail Haka
• Companies: University of Kentucky and University of Kentucky and University of Kentucky and Massachusetts Institute of Technology and Case Western Reserve University and Cornell University
• Keywords: nonparametric regression ; compound estimation ; pattern recognition ; cross validation ; confidence bands

Abstract

Raman spectra have been successfully employed in the classification of normal, benign, and malignant breast tissue, based on coefficient estimates from a linear combination model involving basis spectra for chemical constituents of breast tissue. Motivated by recent work showing that nanoparticles can be classified using not only Mueller scattering profiles but also their derivatives, we investigate whether derivatives of Raman spectra can be effectively employed in the classification of breast tissue. We consider two approaches. The first entails calculating the cross-validated distance between the estimated derivative of a Raman spectrum for a breast tissue to be diagnosed and a reference curve defined as the average estimated derivative for breast tissues known to have a particular diagnosis. The second is based on constructing simultaneous confidence bands about the estimated derivative of a Raman spectrum for a breast tissue to be diagnosed and examining the aforementioned reference curves in relation to the confidence bands. Numerical results from real ex vivo data are presented, and potential implications for the non-invasive diagnosis of breast cancer are discussed.