^{a}, Yuanyuan Wang

^{b}and Chen Zhang

^{a}yzhang.fudan@gmail.com

^{b}yywang@fudan.edu.cn

Photoacoustic Imaging (PAI) is an emerging biomedical imaging technique, which involves the reconstruction of an object’s absorbed laser energy deposition from the measured ultrasonic data. The PAI has great potentials in many clinical applications for its non-invasive, high spatial resolution, and functional imaging ability. Many reconstruction algorithms have been developed for the PAI reconstruction. Comparing to the traditional analytical reconstruction algorithms, the model-based image reconstruction algorithm is able to provide a more accurate and high resolution photoacoustic image. However, the relative heavy computational complexity and huge memory storage often impose restrictions on the use of the model-based reconstruction. Here we proposed an efficient PAI reconstruction algorithm which is based on the Discrete Cosine Transform (DCT) of the measured ultrasonic data. We establish a new photoacoustic forward model, in which the new measurement matrix is obtained by multiplying the DCT matrix. Taking into account of the sparse property of the signal in the DCT domain, the significant parts of the measured signals are used for the reconstruction instead of the entire signal. In this way, much unnecessary calculation can be avoided and the computational complexity will be reduced. During the reconstruction, the measured ultrasonic signal is firstly processed through the discrete cosine transform. Then the significant elements in the processed signal are chosen for the image reconstruction. Usually we set a threshold and neglect the small elements less than the threshold. Correspondingly, the relative insignificant columns in the measurement matrix are removed. Finally the image can be iteratively reconstructed on the basis of the reduced signal and the measurement matrix. The proposed algorithm is verified through the numerical simulations. In the simulations, the threshold is set as 0.01. By neglecting the minor parts of the signal in the DCT domain, the number of elements need to be processed is reduced to 981 from 6750, so that the calculation is reduced dramatically. We record the time cost in the reconstruction, the proposed algorithm takes 190 seconds to reconstruct an image with the size of 150 pixels x 150 pixels, while the traditional algebraic reconstruction algorithm without the DCT processing takes 830 seconds. Additionally, to compare the reconstruction quality, we calculate the Peak Signal to Noise Ratio (PSNR) of the reconstructed image. The PSNR of the image reconstructed by the proposed algorithm is 24.0 dB while that of the reconstructed with algebraic reconstruction algorithm without the DCT processing is 24.3 dB. It is revealed in the simulation results that the proposed algorithm is able to provide an equivalent PAI reconstruction quality with costing much less time. It can be concluded that the proposed algorithm may be useful to enhance the practical applicability of the model-based PAI reconstruction.