Deep design produces 'butterfly' phase mask for light-sheet fluorescence microscopy
(A) When a traditional Gaussian beam is used, diffraction causes the illumination beam to widen at the edges. (B) Joint optimization scheme. A network predicts the positions of beads in the focal plane. The deep learning network and input phase mask are simultaneously updated based on the loss function. Credit: Chen Li et al.
Researchers have introduced a solution to the problem of light-sheet fluorescence microscopy: novel illumination beams designed based on deep learning using a trainable phase mask. Their study eliminates the need for sophisticated optical design tools, allowing optimization to be directly applied to improve image contrast.
The core of this approach lies in the integration of optics propagation modeling and a deep neural network. This optimization updates both the parameters of the deep learning network and the illumination beam simultaneously, resulting in superior image quality. The group's research was published July 4 in Intelligent Computing .
The authors demonstrated the efficacy of their approach through both simulations and optical experiments. The results showcase substantial enhancements in image quality compared to traditional Gaussian light sheets. This method has the potential to simplify the design of new illumination beams, even for those without extensive optics expertise.
The approach is analogous to an assembly line; the traditional deep learning approach represents a skilled worker operating within the established assembly line. In contrast, the new joint optimization approach involves the worker's input during the assembly line design to rapidly obtain superior results.
Instead of merely analyzing images, the new deep learning model designs unexpected shapes for the illumination beam to achieve better results. Specifically, the model generated a butterfly-shaped beam by optimizing "hundreds of thousands of variables" within the phase mask.
Light-sheet fluorescence microscopy has become the leading method for imaging large, tissue-cleared samples in 3D, owing to its optical sectioning, reduced photodamage, and rapid acquisition. Image quality heavily depends on the characteristics of the illumination beam. Recent designs of slender, non-diffracting beams, such as Bessel, Airy, and lattice light-sheets, have achieved uniform and high-contrast images, yet new beam shapes have the potential to improve image quality , a great need in biological samples .
The study contributors included Chen Li, Mani Ratnam Rai, Troy Ghashghaei, Yuheng Cai, Adele Moatti, and Alon Greenbaum from the UNC/NCSU Joint Department of Biomedical Engineering.
More information: Chen Li et al, Intelligent Beam Optimization for Light-Sheet Fluorescence Microscopy through Deep Learning, Intelligent Computing (2024). DOI: 10.34133/icomputing.0095