3D Neighbor2Neighbor-based unsupervised deep learning for noise reduction in OCT imaging: insights from multiple clinical datasets

Optical coherence tomography (OCT), a non-invasive three-dimensional imaging technique, plays a crucial role in the early diagnosis and precise treatment within the fields of ophthalmology and dermatology. However, speckle and electrical noise unavoidably interfere with images during OCT imaging. In this study, we propose a three-dimensional (3D) Neighbor2Neighbor-based unsupervised deep-learning strategy for noise reduction in OCT imaging, which can leverage the inherent properties of OCT 3D volumetric data. In addition, we have also optimized the ResNet network in our 3D framework to improve noise reduction performance and enhance the clinical capabilities of OCT. Experimental results demonstrated that our method can effectively reduce noise and obtain clearer tissue construct details, providing the possibility for the broader and deeper application of OCT technology in ophthalmology and dermatology.