Mueller matrix imaging with improved resolution by deep learning method

Polarization imaging is widely applied in the biomedical field. Mueller matrix imaging is an effective way to extract optical properties of biomedical samples. The mainstream technology is to use Division of Focus Plane (DoFP) system to obtain Mueller matrix images, because Mueller matrix microscope based on dual Division of Focal Plane (DOFP) polarimeters can realize fast full polarization imaging. However, DoFP system has a trade-off between resolution and imaging speed. Although using high numerical aperture objectives can improve resolution, it leads to a decrease of the field of view. To address this issue, we propose a deep learning network named MMRE-GAN (Mueller Matrix Resolution Enhancement GAN), aimed at improving the resolution of Mueller matrix images while also keeping the big field of view of low numerical aperture objectives. Our method does not require paired images and effectively improves the resolution of Mueller matrix images through Generative Adversarial Networks (GANs). In detail, we used a dual Division of Focal Plane (DoFP) polarimeters-based full Mueller Matrix Microscope (DoFPs-MMM) to quickly get the Mueller matrix images of thyroid slices. Then we used the dataset to validate the effectiveness of MMRE-GAN. The results show that our method successfully improves the resolution of Mueller matrix images while keeping the field of view in 4x objective, meaning a shortened scanning time for the whole slides. And our method also enhances the details and signal-to-noise ratio of imaging. It provides support for the application of DoFPs-MMM in biomedical imaging.