An integrated temporal-spatial co-design framework for accelerated swept-source OCT angiography

Optical coherence tomography angiography (OCTA) enables label-free visualization of microvascular networks. However, conventional OCTA relies on repeated B-scan acquisition and dense spatial sampling, resulting in prolonged imaging time and limited real-time applicability. In this work, we propose an integrated temporal-spatial co-design framework for information-enriched and accelerated swept-source OCTA. The acquisition and reconstruction modules address two different but complementary bottlenecks. Variable-interval (VI) scanning does not shorten the four-repeat acquisition itself; instead, it increases temporal sampling efficiency by expanding the effective inter-scan interval coverage within the same acquisition time. Hardware-level lateral undersampling directly reduces the number of fast-axis sampling positions from 1000 to 500, thereby decreasing the actual acquisition time from approximately 40 s to 20 s, while an attention-enhanced residual network compensates for the resulting spatial information loss. Experimental results on mouse cerebral vasculature demonstrate improved vascular visibility and quantitative OCTA metrics under VI acquisition. Real-hardware validation in a stroke mouse model further shows that the complete framework preserves subtle vascular structures within a 20 s acquisition. Using the nominal timing of a conventional eight-repeat sequential acquisition required to include a 7Δt separation (∼80 s) as a theoretical benchmark, the proposed 20 s workflow corresponds to an effective fourfold reduction in acquisition burden. These results support a practical route toward fast, high-fidelity, and information-efficient OCTA imaging.