Abstract
Purpose: Probe-based Confocal Laser Endomicroscopy (pCLE) enables performingan optical biopsy, providing real-time microscopic images, via a probe. pCLEprobes consist of multiple optical fibres arranged in a bundle, which takentogether generate signals in an irregularly sampled pattern. Current pCLEreconstruction is based on interpolating irregular signals onto an over-sampledCartesian grid, using a naive linear interpolation. It was shown thatConvolutional Neural Networks (CNNs) could improve pCLE image quality. Althoughclassical CNNs were applied to pCLE, input data were limited to reconstructedimages in contrast to irregular data produced by pCLE. Methods: We compare pCLEreconstruction and super-resolution (SR) methods taking irregularly sampled orreconstructed pCLE images as input. We also propose to embed a Nadaraya-Watson(NW) kernel regression into the CNN framework as a novel trainable CNN layer.Using the NW layer and exemplar-based super-resolution, we design an NWNetSRarchitecture that allows for reconstructing high-quality pCLE images directlyfrom the irregularly sampled input data. We created synthetic sparse pCLEimages to evaluate our methodology. Results: The results were validated throughan image quality assessment based on a combination of the following metrics:Peak signal-to-noise ratio, the Structural Similarity Index. Conclusion: Bothdense and sparse CNNs outperform the reconstruction method currently used inthe clinic. The main contributions of our study are a comparison of sparse anddense approach in pCLE image reconstruction, implementing trainable generalisedNW kernel regression, and adaptation of synthetic data for training pCLE SR.