(698f) Rapid Diagnosis and Discrimination of Healthy and Breast Cancer Tissues Using Classical and Imaging FTIR

Rapid characterization of normal and malignant breast tissues at the molecular level was aimed using classical and imaging Fourier Transform Infrared (FTIR) spectroscopy. Differences in spectral response due to compositional changes in breast tissue after cancer development constitute the basis of this study. Macroscopic size samples from 17 patients, who diagnosed with invasive ductal or lobular carcinoma at Acibadem Maslak Hospital, Istanbul, Turkey, cut in cubes and spectral responses collected from each face of cubes in duplicate by attenuated internal reflection (ATR) mode operating FTIR spectroscopy. Spectral investigation of breast carcinoma showed that three major responses can be obtained depending on the fat content of any spot contacted with the diamond ATR crystal (0.5 m diameter). We divided the spectrum into three groups as, Type I , Type II and Type II according to the intensity of fat related IR bands at 2800-300 cm^-1 of lipid CH₃ chains and at 1750 cm^-1 C=O of fatty acid ester functional groups (Figure 1). Spectra exhibiting strong fat related IR peaks were found to be typical of normal breast tissue and denoted as spectrum Type I. Mid to low intensity fatty IR bands rich spectra were nominated as spectrum type II which was spectral trend of some healthy and a few malign tissue samples. No fat bands containing spectra were classified as spectrum type III and this was the common observation for almost all malignant tissues.

The talk will cover spectral and chemometric investigation of IR data to develop a classification tool to discriminate tumors from healthy tissues. For this, spectral responses collected from healthy and malignant tissues were first treated with multivariate statistical tools to develop a model to distinguish healthy and malignant tissues. Spectroscopic data was first reduced into factors and loading applying partial least square (PLS) and factors of PLS were then used to develop discrimination models to separate healthy and tumor tissues on a 2D canonical variate plane using canonical variate analysis (CVA). This hybrid PLS-CVA analysis was employed to fingerprint, DNA/RNA or Amide I&II IR absorption bands portions of the original FTIR spectra separately. Since the fingerprint region covers DNA/RNA and Amide absorption bands regions in the 100-1650 cm^-1 wavenumber, the best discrimination results were obtained using from this region. Figure 2 shows discrimination of 38 healthy and 66 tumor tissues in three groups.

After successful classification of breast tissues into healthy and tumor groups, discrimination of tumors with respect to their clinical histopathologic evaluations was performed. Dicrimination of tumors based on their pathologic grading (Figure 3), staging with respect to tubulus formation (T score), spread to lymph nodes (N score) and metastasis of cancer (M score). Discrimination of tumor based on histopathological information will be introduced and successful discrimination models will be shown.

Finally, IR image of a breast tissue will be introduced and tumors’ compositional differentiation depended spectral response changes will be presented to prove the rapid discrimination nature of the proposed technique using molecular level information.