IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 16, NO. 4, JULY/AUGUST 2010 893 Distinguishing Cancer and Normal Breast Tissue Autofluorescence Using Continuous Wavelet Transform Anita H. Gharekhan, Siddharth Arora, Prasanta K. Panigrahi, and Asima Pradhan Abstract—We study the spectral features of the polarized flu- orescence spectra of normal and cancerous human breast tissues through continuous wavelet transform, which clearly identifies dis- tinguishing features between the tissue types. After pinpointing these robust features in the wavelet scalogram, we systematically study the autocorrelation property of the wavelet coefficients of the fluorescence spectra, which is found to differentiate normal and malignant tissues with high sensitivity. The intensity differ- ence of parallel and perpendicularly polarized fluorescence spectra is subjected to investigation, since the same is relatively free of the diffusive background. Index Terms—Autocorrelation, continuous wavelet transform (CWT), Morlet wavelet. I. INTRODUCTION B REAST cancer is a disease in which malignant cells form in the tissues of the breast. It often starts as a small lump, which can occur in both women and men. Among women worldwide, breast cancer is the most common cancer and the most common cause of cancer death. In the United States, it is the third most common cause of cancer death, after lung and colon cancers. Although the risk factor for Asian women has been estimated to be one-fifth to one-tenth that of women in North America and Western Europe, it is still the second most fatal cancer in women after lung cancer [1], [2]. Because the breast is composed of identical tissues in males and fe- males, breast cancer also occurs in males, though it is less common. Autofluorescence spectroscopy has the potential to provide real-time, nondestructive, and quantitative means for charac- terizing tissue pathology. Fluorescence techniques are being increasingly employed to investigate both morphological and biochemical changes in different tissue types, for eventual ap- plication in the detection of tumors at an early stage [3], [4]. Fluorescence spectroscopy is well suited for the diagnosis of Manuscript received August 1, 2009; revised September 16, 2009; accepted September 16, 2009. Date of publication January 8, 2010; date of current version August 6, 2010. A. H. Gharekhan is with the Department of Physics, C. U. Shah Sci- ence College, Gujarat University, Ahmedabad 380 009, India (e-mail: anitaghare@gmail.com). S. Arora is with the Department of Mathematics, University of Oxford, Ox- ford OX1 2HS, U.K. (e-mail: arora@maths.ox.ac.uk). P. K. Panigrahi is with the Physical Research Laboratory, Navrangpura, Ahmedabad 380 009, India, and also with the Indian Institute of Science Edu- cation and Research, Kolkata 700 106, India (e-mail: prasanta@prl.res.in). A. Pradhan is with the Department of Physics and Centre for Laser Technology, Indian Institute of Technology, Kanpur 208016, India (e-mail: asima@iitk.ac.in). Digital Object Identifier 10.1109/JSTQE.2009.2033018 cancerous tissues because of its sensitivity to minute variations in the amount and the local environment of the native fluo- rophores present in the tissues [5]–[10]. Morphological changes prevalent in tumors, such as enlargement and hyperchromasia of nuclei, overcrowding, and irregular cellular arrangement are known to alter light propagation and scattering properties in such media, and hence, affect the fluorescence spectra [11]–[13]. A number of fluorophores, e.g., NADH and flavins producing aut- ofluorescence in the visible regime have proved extremely use- ful for bioimaging. Flavins are the derivatives of riboflavin, the most common of them being flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD). FMN with emission maxima at 530 nm, exhibits much brighter fluorescence efficiency than FAD. Intracellular riboflavin, flavin coenzymes, and flavopro- teins show slightly shifted fluorescence (540–560 nm) compared to flavins. Some of the present authors have earlier studied autofluores- cence through discrete wavelet transform and principal com- ponent analysis (PCA) [14]–[18]. It was found that spectral fluctuations can significantly differ in the diseased and nor- mal tissues. Weak emission peaks corresponding to porphyrin emission can be extracted through discrete wavelets. The spec- tral correlation extracted through PCA also shows different be- havior in the cancer and normal tissue fluorescence indicating the differences in flourophore activities as well as absorption behavior in the two tissues. Here, we investigate the subtle differences in the fluorescence spectra of normal and cancer breast tissues through continuous Morlet wavelets. The fact that continuous wavelets provide an overcomplete basis as com- pared to the complete orthonormal basis set in the discrete wavelets, makes them ideal to extract subtle changes in the spectral features [19]. Morlet wavelet has been found to be quite effective in many areas of data analysis, and hence, the same is employed here. It is found that Morlet wavelet coeffi- cients clearly identify distinguishing features between the tis- sue types. The intensity differences of parallel and perpendicu- larly polarized fluorescence spectra is subjected to investigation, since the same is relatively free of the diffusive background. After pinpointing these robust features in the wavelet scalo- gram, we systematically study the autocorrelation property of the wavelet coefficients of the fluorescence spectra, since these coefficients show different periodic modulations with decreas- ing amplitude for cancer and normal cases. It is found that the correlation properties strongly differ in the two tissue types, which can differentiate normal and malignant tissues with high sensitivity. 1077-260X/$26.00 © 2010 IEEE