Dynamic multiplexed intensity diffraction tomography using a spatiotemporal regularization-driven disorder-invariant multilayer perceptron.

Multiplexed intensity diffraction tomography (mIDT) is a technique that reconstructs the three-dimensional refractive index (3DRI) of a sample solely through intensity measurements. Using an array of light sources to generate multi-directional and multiplexed illumination eliminates the need for mechanical scanning, allowing for quantitative 3DRI reconstruction of label-free samples. However, acquiring multiple images under different illumination conditions limits its use in dynamic scenes. Here, we propose a dynamic 3DRI reconstruction method based on a spatiotemporal regularization-driven disorder-invariant multilayer perceptron (STR-DI-MLP). This method reconstructs the 3DRI distribution of a sample from multiple images with motion-induced shifts. Simulations show it offers faster reconstruction speeds and fewer artifacts. Moreover, it maintains robust reconstruction performance under varying degrees of motion. Experimental validation of our system demonstrates its capability to reconstruct 3DRI in dynamic scenes with motion speeds below approximately 16 µm/s, proving it effective for 3DRI reconstruction of living microorganisms.