Improvement of nerve imaging speed with coherent anti-Stokes Raman scattering rigid endoscope using deep-learning noise reduction

Abstract

A coherent anti-Stokes Raman scattering (CARS) rigid endoscope was developed to visualize peripheral nerves without labeling for nerve-sparing endoscopic surgery. The developed CARS endoscope had a problem with low imaging speed, i.e. low imaging rate. In this study, we demonstrate that noise reduction with deep learning boosts the nerve imaging speed with CARS endoscopy. We employ fine-tuning and ensemble learning and compare deep learning models with three different architectures. In the fine-tuning strategy, deep learning models are pre-trained with CARS microscopy nerve images and retrained with CARS endoscopy nerve images to compensate for the small dataset of CARS endoscopy images. We propose using the equivalent imaging rate (EIR) as a new evaluation metric for quantitatively and directly assessing the imaging rate improvement by deep learning models. The highest EIR of the deep learning model was 7.0 images/min, which was 5 times higher than that of the raw endoscopic image of 1.4 images/min. We believe that the improvement of the nerve imaging speed will open up the possibility of reducing postoperative dysfunction by intraoperative nerve identification.

A Framework for Biosensors Assisted by Multiphoton Effects and Machine Learning

“…Raman microscopy is another option for label-free imaging; however, because of the poor effectiveness of Raman scattering, neuron imaging with ordinary spontaneous Raman scattering needs a considerable exposure period [ 237 ]. Plasmonic materials have been employed to boost the Raman technique’s sensitivity.…”

Convolution Network with Custom Loss Function for the Denoising of Low SNR Raman Spectra

“…During the training phase, the deep learning model adjusts its inner weights so that the MSE is reduced towards zero. While this loss function has been extensively used in regression and denoising tasks, recent studies suggest adding the structured similarity index (SSIM) to enhance the quality of the recovered signal [13,14]. One point to consider in the Raman denoising problem is that the dominant peaks in the signal are of high importance in analyzing the material structure.…”

“…In recent years, deep learning Convolutional Neural Networks (CNNs) have been shown to perform well at denoising coherent Raman spectra when compared to established denoising techniques [12][13][14]. In these contributions deep learning was applied to threedimensional hyperspectral image sets.…”

Advanced Label-Free Laser Scanning Microscopy and Its Biological Imaging Application

“…Then a CARSM with imaging speed of video-rate was developed for vibrational imaging of tissues in vivo by Xie's group via detecting strong backscattering of forward CARS signal with video-rate scanning microscopy [15]. Today, higher speed imaging has been achieved by employing more techniques, such as Fourier-transform [16] and deep-learning noise reduction [17]. The spectral bandwidth of CARSM has also shown great increase by multiplex [18,19] and broadband CARSM techniques [20,21].…”