Framework of Unsupervised based Denoising for Optical Coherence Tomography

Abstract

Optical Coherence Tomography (OCT) is a newly established imaging technology, now widely adopted in various medical settings such as ophthalmology and dermatology, though to a lesser but emerging extent in dentistry. Its conventional acceptance for den-tistry, particularly, is hindered by speckle noise, inherent in the methodology of image capture. A degraded signal-to-noise ratio accentuates ambiguity in feature extraction and contributes to the introduction of artefacts. This ultimately impacts its clinical utility where clear diagnostic detail is sort. This paper proposes a deep learning based denoising technique for OCT images. The approach is an unsupervised denoising framework in which the training data was created from one OCT image. This ensures fast processing as it is focused on essential data removal. Additionally, there are limited clean datasets for OCT available. The approach was analysed quan-titatively and visually against state-of-the-art denoising algorithms. The experimental results show that the approach verifiably removes speckle noise. The method improved the PSNR (dB) by 23.5, CNR (dB) by 7.7 and ENL (dB) by 585.5.