Chinese Journal of Lasers, Volume. 51, Issue 9, 0907011(2024)
Inverse SNR and Complex‑Valued Decorrelation OCTA Real‑Time Imaging Based on GPU High‑Speed Parallel Computing
Fig. 1. OCTA real-time data processing flow chart. The middle region contains the operations performed and the data stored in the GPU, and the upper and lower sides contain the data stored in the CPU. The data are represented by irregular rectangles, and multiple irregular rectangles represent multi-frame data. In the middle region, rectangular boxes represent the steps performed in global memory, rounded rectangular boxes represent the steps performed in texture memory, and dashed boxes represent processing steps using CUDA streams
Fig. 2. Mouse retinal data acquired by real-time en face OCTA image guidance at 250 kHz line scanning speed. (a) Retina; (b) SVP layer; (c) ICP layer; (d) DCP layer
Fig. 3. Real-time display guiding refraction and scanning area adjustment. (a)(b)(c)(d) Real-time display guiding refraction adjustment; (e)(f)(g)(h) real-time display guiding scanning area adjustment; (a)(c)(e)(g) data acquired with real-time cross section OCT image; (b)(d)(f)(h) data acquired with real-time en face OCTA image; (c)(d)(g)(h) correspond successively to the cross section image at the position of the dashed line in (a)(b)(e)(f); (a)(e) images processed after data acquisition, and the rest are real-time display images
Fig. 4. Mouse eye movement and mouse jitter as shown by real-time OCTA images in dynamic OCTA imaging. The first and the third lines are en face OCTA image real-time display, the second and the fourth lines are real-time cross section OCT image corresponding to the position of dashed line, each column corresponding dynamic imaging in order of the 0th s, 6th s and 12th s. The upper dashed rectangular displays eye movement of mouse, and the lower rectangular displays mouse jitter
Fig. 5. Flicker light-induced functional retinal hyperemia experiment in mice is carried out with real-time en face OCTA image. The images and hemodynamic response curves are processed after data acquisition. En face OCTA images of SVP layer (a)‒(c), ICP layer (d)‒(f) and DCP layer (g)‒(i) at baseline stage, stimulation stage and turn-off stimulation stage are shown. Illustrations (Ⅰ‒Ⅻ) are enlarged views of the areas corresponding to the dotted rectangular boxes. The hemodynamic response quantization curves of SVP layer large blood vessels (j), SVP layer capillaries (k), ICP layer (l) and DCP layer (m) are on the right side. En face OCTA image corresponds to different colors according to the decorrelation value. Dotted lines in the illustration indicate the same location, and triangle points to the vessels that respond clearly. Baseline is the baseline stage, FLS is the flicker light stimulation stage, and Post-FLS is the turn-off stimulation stage. Scale bar: 250 μm
Fig. 6. Screenshot of real-time en face OCTA video generated by the system. (a) Baseline stage; (b) stimulus stage; (c) turn-off stimulus stage. The OCTA image corresponds to different colors according to the decorrelation value. The large blood vessels indicated by the arrow and the capillaries in the rectangular box area respond obviously. Scale: 200 μm
|
|
|
Get Citation
Copy Citation Text
Dayou Guo, Kaiyuan Liu, Huiying Zhang, Tengxiang Lin, Zhihua Ding, Peng Li. Inverse SNR and Complex‑Valued Decorrelation OCTA Real‑Time Imaging Based on GPU High‑Speed Parallel Computing[J]. Chinese Journal of Lasers, 2024, 51(9): 0907011
Category: biomedical photonics and laser medicine
Received: Oct. 19, 2023
Accepted: Nov. 22, 2023
Published Online: Apr. 28, 2024
The Author Email: Li Peng (peng_li@zju.edu.cn)