Hyperspectral imaging is an emerging technology for biomedical applications. In this study, an advanced image processing and classification method is proposed to analyze hyperspectral image data for prostate cancer detection. Least squares support vector machines (LS-SVMs) were developed and evaluated for classifying hyperspectral data in order to enhance the detection of cancer tissue. The method was used to detect prostate cancer in tumor-bearing mice. Spatially resolved images were created to highlight the differences of the reflectance properties of cancer versus those of normal tissue. Preliminary results in mice show that the hyperspectral imaging and classification method was able to reliably detect prostate tumors in the animal model. The hyperspectral imaging technique may provide a new tool for optical diagnosis of cancer.