Abstract

A transillumination breast spectroscopy (TiBS) system used for breast cancer risk assessment is being modified to facilitate large-scale trials and to simply use. A proposed change involves switching from a broadband light source to several laser sources cycled through in sequence, which will allow for a wavelength-independent detection system. The effect of the reduction of the spectral content of the system on the ability to predict mammographic density (a known breast cancer risk factor) from the spectra was assessed. Wavelengths for the laser sources were chosen based on their contribution to the loading vectors obtained from a principal components analysis of spectra from a study correlating TiBS spectra with mammographic density. 12 wavelengths were selected based on the principal component loads. Principal component scores were obtained using both full-spectrum and 12-wavelength-spectrum data. No significant loss of predictive ability was found when the broadband spectra were reduced to only 12 wavelengths–for both data sets, 3 principal component scores were significantly able to distinguish between high- and low-mammographic density groups.

© 2009 OSA/SPIE