Deep neural network approach to ghost imaging-optical coherence tomography imaging

• An innovative imaging technique that combines ghost imaging and optical coherence tomography (GI-OCT) has been proposed to reconstruct clear images of targets in scattering media. • Leveraging the deep neural network's (DNN) characteristics of low sensitivity to natural signals and high resistance to noise, a DNN approach termed the GIOCT algorithm has been implemented both in simulation and experimentally. • The number of measurements required to obtain a clear image is reduced to 12.5 % of the previous requirement. Target imaging in scattering media is a crucial task for advancements in biomedical imaging and environmental measurements, offering a non-contact and non-invasive approach. We have designed and developed a novel sensing technique that integrates ghost imaging with optical coherence tomography (GI-OCT) to reconstruct real images of targets within scattering media. This combination effectively mitigates the limitations of OCT, such as the need for extensive scanning time for two-dimensional measurements, and the inability of GI to capture cross-sectional images. Previously, we achieved a target image in scattering media with a resolution of 10 × 10 pixels over an area of 2 × 2 mm 2 . However, practical applications demand higher resolution and shorter measurement times. A significant challenge in GI-OCT measurements is the substantial computational demand associated with increased image resolution. In this paper, we introduce a novel computational method to address this challenge. By leveraging deep learning techniques, we successfully reconstructed images in simulation with a 12.5 % sampling rate at a resolution of 64 × 64 pixels, significantly reducing the required measurements. This method is computationally verified, enhancing image quality after 200 iterations. In experiments, the GIOCT algorithm at an optimal learning rate of 0.01 showed an average improvement of 40 % compared to initial reconstructions using the differential ghost imaging method. This advancement represents a significant step forward in the practical application of GI-OCT for high-resolution target imaging in scattering media.