Cancer Classification of Freshly Excised Murine Tumors with Ordered Orthogonal Projection

This paper proposes a low dimension feature extraction method for cancer classification with terahertz (THz) images of freshly excised murine tumors. In THz imaging, the properties of various tissue regions, such as cancer, fat, and muscle, are embedded in the shape of reflected THz pulses. We propose to extract such information by projecting the frequency domain THz waveforms onto a series of ordered orthogonal basis. Experiment results demonstrated that the proposed algorithm can achieve reasonably accurate detection results for fresh tissues with three regions.