LV‐GAN: A deep learning approach for limited‐view optoacoustic imaging based on hybrid datasets

Abstract

The optoacoustic imaging (OAI) methods are rapidly evolving for resolving optical contrast in medical imaging applications. In practice, measurement strategies are commonly implemented under limited‐view conditions due to oversized image objectives or system design limitations. Data acquired by limited‐view detection may impart artifacts and distortions in reconstructed optoacoustic (OA) images. We propose a hybrid data‐driven deep learning approach based on generative adversarial network (GAN), termed as LV‐GAN, to efficiently recover high quality images from limited‐view OA images. Trained on both simulation and experiment data, LV‐GAN is found capable of achieving high recovery accuracy even under limited detection angles less than 60°. The feasibility of LV‐GAN for artifact removal in biological applications was validated by ex vivo experiments based on two different OAI systems, suggesting high potential of a ubiquitous use of LV‐GAN to optimize image quality or system design for different scanners and application scenarios.

A jointed feature fusion framework for photoacoustic image reconstruction

“…7 shows the output of RGC-Net, which indicates the image post-processing scheme. Noting that some prevailing end-to-end deep learning solutions for reconstruction are often implemented by arbitrarily changing this backbone [15] , [17] , [37] , [38] , which is also a comparative experiment. Fig.…”

“…DL enables PACT reconstruction in both image and signal domains. In image domain, a straightforward way of applying DL is to reduce image artifacts as a post-processing step [14] , [15] , [16] . For instance, Austin Reiter identified the point source locations from pre-beamformed PA data using a CNN [17] .…”

Sounding out the hidden data: A concise review of deep learning in photoacoustic imaging

“…55 Similarly, Lu et al proposed the use of a GAN model for a ring-array PACT system with a limited view. 56 Rather than using simulated data for training, Davoudi et al have directly utilized experimental data to train a U-Net for removing both sparse-sampling and limited-view artifacts. 17 Full- view reconstructed images were used as the ground truth to train a CNN model to improve sub-aperture reconstructed images.…”

A practical guide for model-based reconstruction in optoacoustic imaging

“…OA is inherently a tomographic imaging modality. This implies that OA images can only be accurately reconstructed if signals are acquired at a set of locations enclosing the sample with sufficient angular coverage [35][36][37]. OA excitation is mainly performed with short (10 −10 -10 −7 s) light pulses.…”