Abstract:Deep learning shows great potential for super-resolution microscopy, offering biological structures visualization with unprecedented details and high flexibility. An effective pathway toward this goal is structured illumination microscopy (SIM) augmented by deep learning because of its ability to double the resolution beyond the light diffraction limit in real-time. Although the deep-learning-based SIM technique works effectively, it is generally a black box that is difficult to explain the latent principle. Thus, the generated super-resolution biological structures contain unconvinced information for clinical diagnosis. This limitation impedes its further applications in safety-critical fields like medical imaging. In this paper, we report a reliable deep-learning-based SIM technique with uncertainty maps. These uncertainty maps characterize imperfections in various disturbances, such as measurement noise, model error, incomplete training data, and out-of-distribution testing data. Specifically, we employ a Bayesian convolutional neural network to quantify uncertainty and explore its application in SIM. The backbone of the reported neural network is the combination of U-net and Res-net with three low-resolution images from different structured illumination angles as inputs. The outputs are high-resolution images with double resolution beyond the numerical aperture and the pixel-wise confidence intervals quantification of reconstruction images. A series of simulations and experiments validate that the reported uncertainty quantification framework offers reliable uncertainty maps and high-fidelity super resolution images. Our work may promote practical applications of deep-learning-based super-resolution microscopy.

Adaptive Optics For Structured Illumination Microscopy Based On Deep Learning

Aberrations in Structured Illumination Microscopy: A Theoretical Analysis

Real-time reconstruction using electro-optics modulator-based structured illumination microscopy

Regularized Deconvolution for Structured Illumination Microscopy Via Accelerated Linearized ADMM

Untrained Neural Network Enabling Fast and Universal Structured-Illumination Microscopy

Structured Illumination Microscopy with Modified Joint Richardson-Lucy Deconvolution

Wavefront Reconstruction Based on Deep Transfer Learning for Microscopy.

A Sensorless Adaptive Optics Control System for Microscopy Based on Extreme Learning Machine

3D Super-Resolution Microscopy Based on Nonlinear Gradient Descent Structured Illumination.

Machine learning guided rapid focusing with sensor-less aberration corrections.

Accelerated Phase Shifting for Structured Illumination Microscopy Based on Deep Learning

Super-Resolution Structured Illumination Microscopy Reconstruction Using a Least-Squares Solver

Structured Illumination Microscopy with Uncertainty Quantification

Deep super-resolution imaging of thick tissue using structured illumination with adaptive optics

Deep learning enables structured illumination microscopy with low light levels and enhanced speed

Correction of Refractive Index Mismatch-Induced Aberrations under Radially Polarized Illumination by Deep Learning.

Super-resolution reconstruction of structured illumination microscopy using deep-learning and sparse deconvolution

Optical Aberration Correction via Phase Diversity and Deep Learning

Untrained Neural Network Enhances the Resolution of Structured Illumination Microscopy under Strong Background and Noise Levels

Surpassing the resolution limitation of structured illumination microscopy by an untrained neural network

Enhancing structural illumination microscopy with hybrid CNN-transformer and dynamic frequency loss