A Deep Learning Approach to Denoise Optical Coherence Tomography Images of the Optic Nerve Head
A Deep Learning Approach to Denoise Optical Coherence Tomography Images of the Optic Nerve Head
复制标题
DOI：
10.1038/s41598-019-51062-7
复制
发表时间：
2019-10-08
影响因子：
4.6
通讯作者：
Girard, Michael J. A.
中科院分区：
文献类型：
--
作者：
Devalla, Sripad Krishna;Subramanian, Giridhar;Girard, Michael J. A.
光学相干断层扫描（OCT）已成为一个既定的临床常规的视神经乳头（ONH）组织的体内成像，这是至关重要的诊断和管理的各种眼部和神经-眼部病变。然而，散斑噪声的存在影响OCT图像的质量及其解释。虽然最近的帧平均技术已被证明可以提高OCT图像质量，但它们需要更长的扫描持续时间，导致患者不适。使用一个自定义的深度学习网络，训练了2,328个“干净的B扫描”（多帧B扫描;信号平均），以及相应的“嘈杂的B扫描”（干净的B扫描+高斯噪声），我们能够成功地对1,552个看不见的单帧（没有信号平均）B扫描进行降噪。去噪B扫描在定性上与其相应的多帧B扫描相似，ONH组织的可见性增强。平均信噪比（SNR）从4.02 +/- 0.68 dB（单帧）增加到8.14 +/- 1.03 dB（去噪）。对于所有ONH组织，平均对比噪声比（CNR）从3.50 +/- 0.56（单帧）增加到7.63 +/- 1.81（去噪）。与相应的多帧B扫描相比，平均结构相似性指数（MSSIM）从0.13 +/- 0.02（单帧）增加到0.65 +/- 0.03（去噪）。我们的深度学习算法可以在20 ms内对ONH的单帧OCT B扫描进行降噪，从而提供了一个框架，以获得上级质量的OCT B扫描，减少扫描时间，最大限度地减少患者不适。
Optical coherence tomography (OCT) has become an established clinical routine for the in vivo imaging of the optic nerve head (ONH) tissues, that is crucial in the diagnosis and management of various ocular and neuro-ocular pathologies. However, the presence of speckle noise affects the quality of OCT images and its interpretation. Although recent frame-averaging techniques have shown to enhance OCT image quality, they require longer scanning durations, resulting in patient discomfort. Using a custom deep learning network trained with 2,328 'clean B-scans' (multi-frame B-scans; signal averaged), and their corresponding 'noisy B-scans' (clean B-scans + Gaussian noise), we were able to successfully denoise 1,552 unseen single-frame (without signal averaging) B-scans. The denoised B-scans were qualitatively similar to their corresponding multi-frame B-scans, with enhanced visibility of the ONH tissues. The mean signal to noise ratio (SNR) increased from 4.02 +/- 0.68 dB (single-frame) to 8.14 +/- 1.03 dB (denoised). For all the ONH tissues, the mean contrast to noise ratio (CNR) increased from 3.50 +/- 0.56 (single-frame) to 7.63 +/- 1.81 (denoised). The mean structural similarity index (MSSIM) increased from 0.13 +/- 0.02 (single frame) to 0.65 +/- 0.03 (denoised) when compared with the corresponding multiframe B-scans. Our deep learning algorithm can denoise a single-frame OCT B-scan of the ONH in under 20 ms, thus offering a framework to obtain superior quality OCT B-scans with reduced scanning times and minimal patient discomfort.