Recently, the team led by Qin Wenjian from the Institute of Biomedical and Health Engineering at the Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, published their research findings in the journal "Optics and Lasers in Engineering" with the title "Fast physic-informed mixer architecture for color Lensfree holographic reconstruction." The team proposed an unsupervised complex-valued network architecture based on physical information for efficient, high-quality color lens-free holographic reconstruction.

Accurate reconstruction of color images from multi-wavelength holograms is crucial in biomedical imaging applications. Currently, data-driven deep learning methods have made significant progress in the performance of biomedical image reconstruction. In particular, untrained neural network methods can effectively address the requirements for the number of samples in the dataset and generalization issues in imaging models. However, existing methods still require more iterative calculations to improve reconstruction quality, resulting in longer convergence times for the models.

The team proposed an efficient complex-valued attention mixer (ECA-Mixer) architecture for fast and accurate physical information-based color holographic reconstruction. This architecture consists of three core modules—encoder, nonlinear transformer, and decoder. Each module integrates efficient attention mechanisms and mixer layers for channel feature extraction and spatial information transformation. To preserve high-frequency information, the team also introduced 2D Haar wavelets and their inverse transformations for encoding and decoding features.

The results of this achievement on a large number of simulated and experimental samples demonstrate that this approach has achieved excellent color reconstruction results in terms of computational time and image quality. More importantly, the proposed solution can rapidly image large-size wide-field samples at higher resolution within just a few minutes. The aforementioned technical achievements provide new ideas and methods for the application of computational holographic imaging in biomedical microscopic imaging.

The paper is available at here. The Shenzhen Advanced Institute is the first completing insitute of the paper.