Denoising signals for photoacoustic imaging in frequency domain based on empirical mode decomposition

As a newly developed biomedical imaging technology, photoacoustic imaging (PAI) has attracted great attention worldwide. Traditional frequency domain PAI modality requires tremendous calculations by involving back-projection algorithm in time domain. A pure frequency domain method referred as model-based frequency domain PAI modality based on amplitude and phase analysis was introduced recently. However, the noises in signals may distort the reconstruction images severely. Since noises are normally un-stationary and non-linear, the degree of difficulty increases for restraining the impact of noises by Fourier transform. Also, impacts of different types of noise vary hugely. Moreover, inappropriate sifting process may even aggravate image distortion. In this study, a frequency domain PAI system based on chirp modulation signals has been introduced, in which the influence introduced by two main types of noise have been analyzed by simulation to show how different types of noises may damage reconstructed image in practical and to what extent. White Gaussian noise acts as an additive noise and distort the image gradually while the stochastic noise distort the image very rapidly as the fluctuation increase. Then a new method based on the empirical mode decomposition (EMD) for signal-sifting is demonstrated. By carefully choosing the thresholds and iterative steps, the reconstructed image quality has been promoted, the relative intensity of the target and noises has increased a lot. Further discussion about this method is also made, which certifies the availability of the proposed method. (C) 2018 Elsevier GmbH. All rights reserved.