Photoacoustic spectroscopy based evaluation of breast cancer condition

摘要

Photoacoustic spectroscopy, a hybrid of optics and acoustics has been gaining popularity in the biomedical field very fast. The main aim in the present study was to apply this technique to detect and distinguish breast tumor tissues from normal and hence develop a tool for clinical applications. There were 224 photoacoustic spectra recorded from 28 normal and 28 breast tumor tissues using PZT detector at 281nm pulsed laser excitations from Nd-YAG laser pumped frequency doubled dye laser system. The recorded time domain photoacoustic spectra were fast Fourier transformed into frequency domain patterns in the frequency region 0-1250kHz and from each pattern, 7 features (mean, median, mode, variance, standard deviation, area under the curve & spectral residual after fitting with 10th degree polynomial) were extracted using MATLAB algorithms. These features were then tested for their significance between normal and malignant conditions using Student T-test and two of them (variance, std. deviation) showing significant variation were selected for further discrimination analysis using supervised quadratic discriminate analysis (QDA). In QDA, 60 spectra from each of the normal and malignant were used for making the respective calibration sets and the remaining 52 spectra from each were used for the validation. The performance of the analysis tested for the frequency region 406.25 - 625.31 kHz, showed specificity and sensitivity values of 100% and 88.46% respectively suggesting possible application of the technique in breast tumor detection.