Image Reconstruction Through Multimode Polymer Optical Fiber for Potential Optical Recording of Neural Activity

Abstract

Despite the growing demand for high-resolution imaging techniques in neuroscience, traditional methods are limited in terms of flexibility and spatial resolution. We explored an approach using multimode polymer optical fiber (POF) and employing a neural network for image reconstruction and studied the ability of multimode POF to effectively capture and reconstruct high-quality images. Here, a conventional U-Net model within the framework of convolutional neural networks (CNNs) is applied to the reconstruction of speckle images obtained via POF. The model was trained on an experimental dataset consisting of MNIST graphs and successfully reconstructed high-quality images that closely resemble the original undistorted scene. This study not only highlights the potential of POF in biomedical imaging but also paves the way for more sophisticated optical recording techniques.