Predictive potential of photoacoustic spectroscopy in breast tumor detection based on xenograft serum profiles

Breast cancer is the second most common cancer all over the world. Heterogeneity in breast cancer makes it a difficult task to detect with the existing serum markers at an early stage. With an aim to detect the disease early at the pre-malignant level, MCF-7 cells xenografts were developed using female nude mice and blood serum were extracted on days 0^th, 10^th, 15^th & 20^th post tumor cells injection (N=12 for each time point). Photoacoustic spectra were recorded on the serum samples at 281nm pulsed laser excitations. A total of 144 time domain spectra were recorded from 48 serum samples belonging to 4 different time points. These spectra were then converted into frequency domain (0-1250kHz) using MATLAB algorithms. Subsequently, seven features (mean, median, mode, variance, standard deviation, area under the curve & spectral residuals after 10^th degree polynomial fit) were extracted from them and used for PCA. Further, using the first three Principal components (PCs) of the data, Linear Discriminate Analysis has been carried out. The performance of the analysis showed 82.64% accuracy in predicting various time points under study. Further, frequency-region wise analysis was also performed on the data and found 95 - 203.13 kHz region most suitable for the discrimination among the 4 time points. The analysis provided a clear discrimination in most of the spectral features under study suggesting that the photoacoustic technique has the potential to be a diagnostic tool for early detection of breast tumor development.