Enhancing the Performance of Pulse Position Coded Excitation for Photoacoustic Imaging by Denoising Autoencoder

Abstract

Photoacoustic (PA) imaging is a hybrid technique that combines light illumination and ultrasound detection to generate images of tissue. Advances in laser technology allow laser diodes that are low cost, compact, and have a high pulse repetition frequency (PRF) to improve the frame rate and signal-to-noise ratio (SNR) of PA imaging. This improvement is achieved by employing PA-coded excitation techniques. However, PA-coded excitation is limited by side-lobes and artifact signals, particularly when the code length is short. Pulse position modulation (PPM) is a type of coded excitation that achieves the highest code gain with a short code length. This study explores a signal-processing approach that integrates PPM-coded excitation with a denoising autoencoder to reduce the generated side lobs and artifact signals and enhance the SNR of the PA signals. The denoising autoencoder is designed to address the varying shapes of side lobes that occur with different PPM code lengths, resulting in improved attenuation and removal of artifacts and background noise. The results show that the denoising autoencoder is particularly effective when the amplitude of the decoded PA signal exceeds that of the background noise, enabling reduced acquisition time and memory requirements for RF data collection. This work offers a promising approach to overcoming the limitations of PPM-coded excitation in PA imaging, supporting further improvements in the quality and reliability of PA signals for various medical applications.