In the realm of optical imaging, computational imaging has emerged as a groundbreaking technique, promising wide field-of-view and high-resolution capabilities. A key advancement in this area is coherent imaging, also known as holographic imaging, which allows optical systems to capture an enormous number of optically resolvable spots, providing crucial insights into cellular and molecular structures for biomedical research.
While the potential is vast, existing large-scale coherent imaging methods face challenges in practical clinical applications. Many of these techniques require multiple scanning or modulation processes, resulting in lengthy data collection times needed to achieve high resolution and signal-to-noise ratio. This tradeoff between speed, resolution, and quality hinders the feasibility of widespread clinical use.
Recently, researchers from the Beijing Institute of Technology, the California Institute of Technology, and the University of Connecticut have presented a novel solution in the form of a complex-domain neural network. Their work, published in Advanced Photonics Nexus, demonstrates significant improvements in large-scale coherent imaging. By using denoising algorithms during iterative reconstruction, they enhance imaging quality even with sparse data, overcoming computational complexities and limitations of traditional deep learning-based techniques.
The key innovation lies in exploiting latent coupling information between amplitude and phase components, enabling multidimensional representations of complex wavefronts. This approach exhibits strong generalization and robustness across various coherent imaging modalities.
The researchers developed a network comprising a two-dimensional complex convolution unit and complex activation function. They also devised a comprehensive multi-source noise model that encompasses speckle noise, Poisson noise, Gaussian noise, and super-resolution reconstruction noise. This noise model enhances the domain-adaptation ability from synthetic data to real data.
The reported technique was tested on several coherent imaging modalities, including Kramers-Kronig relations holography, Fourier ptychographic microscopy, and lensless coded ptychography. Extensive simulations and experiments demonstrated that the approach maintains high-quality reconstructions and efficiency while significantly reducing exposure time and data volume by an order of magnitude.
The implications of these high-quality reconstructions are substantial, especially for subsequent high-level semantic analysis, such as high-accuracy cell segmentation and virtual staining, which could advance intelligent medical care.
The potential for rapid, high-resolution imaging with reduced exposure time and data volume opens up exciting possibilities for real-time cell observation. Combining this technology with artificial intelligence diagnosis could unlock the mysteries of complex biological systems and push the boundaries of medical diagnostics. This innovative approach has the potential to revolutionize optical imaging and significantly impact biomedical research and healthcare.