Contrast Features of Breast Cancer in Frequency-Domain Laser Scanning Mammography

Abstract

Frequency-domain optical mammography has been advocated to improve contrast and thus cancer detectability in breast transillumination. To the best of our knowledge, this report provides the first systematic clinical results of a frequency-domain laser scanning mammograph (FLM). The instrument provides monochromatic light at 690 and 810 run, whose intensity is modulated at 110.0010 and 110.0008 MHz, respectively. The breast is scanned by stepwise positioning of source and detector, and amplitude and phase for both wavelengths are measured by a photomultiplier tube using heterodyne detection. Images are formed representing amplitude or phase data on linear gray scales. Furthermore, various algorithms carrying on more than one signal (amplitude ratio, phase difference, μ a, μ′ s, N) were essayed. Twenty visible cancers out of 25 cancers in the first 59 investigations were analyzed for their quantitative contrast with respect to the whole breast or to defined reference areas. Contrast definitions refer to the signal itself (definition 1), to the signal noise (definition 2), or were based on nonparametric comparison (definition 3). The amplitude signal provides better contrast than the phase signal. Ratio images between red and infrared amplitudes gave variable results; in some cases the tumor contrast was canceled. The algorithms to determine μ a and μ′ s from amplitude and phase data did not significantly improve upon objective contrast. The N algorithm, using the phase signal to flatten the amplitude signal did significantly improve upon contrast according to contrast definitions 1 and 2, however, did not improve upon nonparametric contrast. Thus, with the current instrumentation, the phase signal is helpful to correct for the complex and variable geometry of the breast. However, an independent informational content for tumor differentiation could not be determined. The flat field algorithm did greatly enhance optical contrast in comparison with amplitude or amplitude ratio images. Further evaluation of FLM will have to be based on the N -algorithm images. © 1998 Society of Photo-Optical Instrumentation Engineers.