Megahertz-rate down-sampled optical coherence tomography based on high-speed k-linear swept laser and deep learning

Megahertz-rate optical coherence tomography (MHz-OCT) is an optical imaging technology that has attracted considerable attention in clinical practice. Its advantages, such as ultra-high speed, noninvasiveness, and high resolution, endow it with broad application prospects in various clinical fields. However, MHz-OCT systems place high demands on the sampling rate and bandwidth of acquisition and data transmission systems, greatly increasing the system cost. Based on a high-speed k-linear swept laser with the acousto-optic deflector (AOD), this paper proposes a hardware-based down-sampling method. Sweeping a narrowband spectrum and utilizing the linear wavenumber characteristic of the laser enables an equivalent down-sampling of the original interference signal. Deep learning is employed to recover high-resolution images from the down-sampled signals. High-quality imaging results have been successfully achieved at a high sweep speed of 1 MHz while using acquisition and data transmission systems with lower bandwidth and sampling rate. The novel down-sampled OCT system proposed in this paper helps to reduce the cost of MHz-OCT systems in clinical settings and promotes their popularization and application.