Two photon adaptive microscopy using Walsh mode based neural network aberration estimation

Existing adaptive optics microscopy methods correcting aberrations with complex high-order shapes commonly require a time-consuming optimization process. A neural network encapsulating the mathematics of Walsh mode lattices was built. Results suggested the method had the potential to outperform conventional methods. Furthermore, based on the aberration information extraction power of convolutional neural networks and the image pre-processing method called pseudo-PSF generation, a new image-based neural network was proposed, which will further improve the correction accuracy and robustness against noise and will shorten the convergence duration.