Physics-enhanced machine learning for virtual fluorescence microscopy
This work addresses the challenge of improving virtual fluorescence microscopy for biological imaging researchers, representing an incremental advance through hybrid physics-ML integration.
The paper tackles the problem of inferring fluorescence images from unstained transmission microscopy images by introducing a method that incorporates illumination modeling into deep neural networks, resulting in consistent performance improvements across different experimental setups with learned LED patterns showing greater benefits for higher-information targets.
This paper introduces a new method of data-driven microscope design for virtual fluorescence microscopy. Our results show that by including a model of illumination within the first layers of a deep convolutional neural network, it is possible to learn task-specific LED patterns that substantially improve the ability to infer fluorescence image information from unstained transmission microscopy images. We validated our method on two different experimental setups, with different magnifications and different sample types, to show a consistent improvement in performance as compared to conventional illumination methods. Additionally, to understand the importance of learned illumination on inference task, we varied the dynamic range of the fluorescent image targets (from one to seven bits), and showed that the margin of improvement for learned patterns increased with the information content of the target. This work demonstrates the power of programmable optical elements at enabling better machine learning algorithm performance and at providing physical insight into next generation of machine-controlled imaging systems.