This study investigates the feasibility of using frequency analysis of multispectral PA (Photoacoustics) signals generated by excised human thyroid tissue to differentiate between malignant and normal thyroid regions. Multispectral PA imaging was performed on freshly excised thyroid tissue from 6 patients undergoing thyroidectomy or thyroid lobectomy. The regions of interests in the PA images corresponding to malignant and normal tissue have been selected with the help of histopathology slides. The calibrated power spectrum of each PA signal from each region of interest was fit to a linear model for extracting the values of slope, midband fit and intercept parameters. The results show that mean values of intercept and midband fit parameters are significantly different between malignant and normal regions for all five wavelengths and mean values of slope are significantly different for two wavelengths.