Probe-based confocal laser endoscopy (pCLE) has emerged as a powerful tool for disease diagnosis, yet it faces challenges such as the formation of hexagonal patterns in images due to the inherent characteristics of fiber bundles. Recent advancements in deep learning offer promise in image denoising, but the acquisition of clean-noisy image pairs for training networks across all potential scenarios can be prohibitively costly. Few studies have explored training denoising networks on such pairs. Here, we propose an innovative self-supervised denoising method. Our approach integrates noise prediction networks, image quality assessment networks, and denoising networks in a collaborative, jointly trained manner. Compared to prior self-supervised denoising methods, our approach yields superior results on pCLE images and fluorescence microscopy images. In summary, our novel self-supervised denoising technique enhances image quality in pCLE diagnosis by leveraging the synergy of noise prediction, image quality assessment, and denoising networks, surpassing previous methods on both pCLE and fluorescence microscopy images.
Probe-based confocal laser endomicroscopy (pCLE) is a subcellular in vivo imaging technique that generates diagnostic images revealing malignant structural modifications in epithelial tissues. In the clinical diagnosis of probe confocal laser endomicroscopy (pCLE), the image background generally has the problems of dynamic blur or information loss, which is not conducive to achieving high-resolution and clear pCLE imaging. In recent years, deep learning technology has achieved remarkable results in image deblurring. For the task of recovering high-resolution pCLE images, the current methods still suffer from the following drawbacks: it is difficult to choose a strategy to make CNN converge at a deeper level and mainstream methods cannot handle the complex balance between spatial details and high-level feature information well when reconstructing clear images. In order to solve the problem, we propose a new cross-channel attention, multistage, high-resolution pCLE image deblurring structure. This methodology improves the supervised attention mechanism, enhances the ability of feature extraction and fusion capabilities, and improves the quality of image deblurring by adding cross-channel attention module (CAM) into the multistage neural networks’ architecture. The experimental results show that the average peak signal-to-noise ratio (PSNR) of the proposed model on the dataset is as high as 29.643 dB, and the structural similarity (SSIM) reaches 0.855. This method is superior to the prior algorithms in the visualization of recovered images, and the edge and texture details of the restored pCLE images are clearer.