Compressive-sensing like grating-lobe suppressed image reconstruction for photoacoustic linear array imaging

To avoid large grating lobes, using a small element-to-element pitch of ultrasound array transducers for photoacoustic (PA) imaging is necessary. Such constraint introduces higher system cost and complexity, especially when greater than 20-MHz high-frequency arrays are used. As a result, to reduce fabrication difficulties and obtain better signal sensitivity in PA imaging, ultrasound linear array transducers are commonly used in practice instead of phased arrays. However, the field-of-view (FOV) is limited to the full aperture size because linear arrays do not have the ability to steer PA receive beams without the introduction of large grating lobes. In addition, strong PA signals are commonly generated in the near field in the back-ward mode where grating-lobe clutters can even hamper the image contrast seriously. In this study, we proposed a novel compressed-sensing-like grating-lobe suppressed image reconstruction method for PA linear array imaging. To overcome the tradeoff between FOV and grating lobe clutters introduced by using a linear array, compressive sensing (CS) concept is adopted here to reduce the grating lobes. The CS theory relies on an important principle: sparsity. Fortunately, unlike ultrasound imaging, absorption distribution in PA imaging intrinsically owns sparsity in the spatial domain. In consequence, a sparsity constraint minimizing the L1 norm of energy deposition can be introduced to the conventional reconstruction method. By adopting such a constraint and using the nonlinear recovery algorithm based on convex optimization, PA linear array imaging can be reconstructed with grating lobe clutters greatly suppressed. Simulation results demonstrated that the proposed method can reduce the grating lobes caused by using a linear array with large FOV. In the meantime, compared with the image reconstructed by the traditional back-projection method, the image reconstructed by the proposed method has fewer artifacts.