Abstract:Recovering 3D phase features of complex, multiple-scattering biological samples traditionally sacrifices computational efficiency and processing time for physical model accuracy and reconstruction quality. This trade-off hinders the rapid analysis of living, dynamic biological samples that are often of greatest interest to biological research. Here, we overcome this bottleneck by combining annular intensity diffraction tomography (aIDT) with an approximant-guided deep learning framework. Using a novel physics model simulator-based learning strategy trained entirely on natural image datasets, we show our network can robustly reconstruct complex 3D biological samples of arbitrary size and structure. This approach highlights that large-scale multiple-scattering models can be leveraged in place of acquiring experimental datasets for achieving highly generalizable deep learning models. We devise a new model-based data normalization pre-processing procedure for homogenizing the sample contrast and achieving uniform prediction quality regardless of scattering strength. To achieve highly efficient training and prediction, we implement a lightweight 2D network structure that utilizes a multi-channel input for encoding the axial information. We demonstrate this framework's capabilities on experimental measurements of epithelial buccal cells and Caenorhabditis elegans worms. We highlight the robustness of this approach by evaluating dynamic samples on a living worm video, and we emphasize our approach's generalizability by recovering algae samples evaluated with different experimental setups. To assess the prediction quality, we develop a novel quantitative evaluation metric and show that our predictions are consistent with our experimental measurements and multiple-scattering physics.

Physics-informed neural networks for diffraction tomography

Universal and High-Fidelity Resolution Extending for Fluorescence Microscopy Using a Single-Training Physics-Informed Sparse Neural Network

Deep-learning Projector for Optical Diffraction Tomography.

Physical model simulator-trained neural network for computational 3D phase imaging of multiple-scattering samples

Physics-informed neural networks for solving nonlinear diffusivity and Biot's equations

Reconstruction of nanoscale particles from single-shot wide-angle FEL diffractions patterns with physics-informed neural networks

A neural network reconstruction approach for obtaining parallax-free X-ray powder diffraction computed tomography data from large samples

Video frame interpolation neural network for 3D tomography across different length scales

Refractive index tomography with a physics based optical neural network

Hybrid Multi-Head Physics-informed Neural Network for Depth Estimation in Terahertz Imaging

Real-time 3D Nanoscale Coherent Imaging via Physics-aware Deep Learning

Neural rendering enables dynamic tomography

Physics Embedded Weight-Sharing Neural Network for Electrical Impedance Tomography

Amplitude/Phase Retrieval for Terahertz Holography With Supervised and Unsupervised Physics-Informed Deep Learning

Experimental neural network enhanced quantum tomography

The Forward and Inverse Models in Time-Resolved Optical Tomography Imaging and Their Finite-Element Method Solutions

Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations

Solving Optical Tomography with Deep Learning

High-resolution tomographic reconstruction of optical absorbance through scattering media using neural fields

Reconstruction of partially occluded objects with a physics-driven self-training neural network