Detecting abnormal vasculature from photoacoustic signals using wavelet-packet features

Abstract

Photoacoustic systems can produce high-resolution, high-contrast images of vascular structures. To reconstruct images at very high-resolution, signals must be collected from many transducer locations, which can be time consuming due to limitations in transducer array technology. A method is presented to quickly discriminate between normal and abnormal tissue based on the structural morphology of vasculature. To demonstrate that the approach may be useful for cancer detection, a special simulator that produces photoacoustic signals from 3D models of vascular tissue is developed. Results show that it is possible to differentiate tissue classes even when it is not possible to resolve individual blood vessels. Performance of the algorithm remains strong as the number of transducer locations decreases and in the presence of noise.