Hybrid diffusion deep learning restores full-fidelity multispectral optoacoustic tomography images using 32 detectors instead of 256

https://www.eurekalert.org/news-releases/1140487

"multispectral optoacoustic tomography provides deep tissue anatomical/ functional data, but constrained by high cost, complex hardware, processing delays, severe noise/ artifacts caused by sparse data sampling... overcome combining 10-step short-chain diffusion with edge-aware self-attention preserving spatial resolution/ anatomical boundaries... roughly 150 ms/frame inference speed, 0-shot generalization on human clinical data... high-resolution deep-tissue optoacoustic imaging at point-of-care clinics"