Deep learning acceleration of multiscale superresolution localization photoacoustic imaging
Problem
Superresolution photoacoustic imaging achieves very high spatial resolution but requires superimposing many raw image frames, sacrificing temporal resolution and making imaging slow.
Why it matters
Slow imaging limits practical biological and medical use of superresolution photoacoustic imaging.
What the AI did
Deep neural networks reconstruct high-density superresolution images from far fewer raw frames, applied to both 3D label-free OR-PAM and 2D labeled PACT.
Results
Required raw volumetric frames reduced from tens to fewer than ten for OR-PAM; PACT result truncated in abstract
Impact
Demonstrated substantial reduction in required frames, enabling faster superresolution photoacoustic imaging.
Record
| Venue | Light Science & Applications |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Jongbeom Kim, Gyuwon Kim, Lei S. Li, Pengfei Zhang, Jin Young Kim, Yeonggeun Kim, Hyung Ham Kim, Lihong V. Wang et al. |
| Citations | 126 (OpenAlex, at import time) |
| References | 62 |
| Model or method | deep neural networks (DNNs), not further specified |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41377-022-00820-w |
| Open access | gold |
| Keywords | photoacoustic imaging, superresolution, deep learning, OR-PAM, PACT, temporal resolution |
Importance rating
real but narrower advance
Speeds superresolution photoacoustic imaging for potential preclinical/clinical use; a real but narrower advance within one imaging modality, with no scale stated.