Deep‐learning‐based motion correction in optical coherence tomography angiography

Abstract

Optical coherence tomography angiography (OCTA) is a widely applied tool to image microvascular networks with high spatial resolution and sensitivity. Due to limited imaging speed, the artifacts caused by tissue motion can severely compromise visualization of the microvascular networks and quantification of OCTA images. In this article, we propose a deeplearning-based framework to effectively correct motion artifacts and retrieve microvascular architectures. This method comprised two deep neural networks in which the first subnet was applied to distinguish motion corrupted B-scan images from a volumetric dataset. Based on the classification results, the artifacts could be removed from the en face maximum-intensity-projection (MIP) OCTA image. To restore the disturbed vasculature induced by artifact removal, the second subnet, an inpainting neural network, was utilized to reconnect the broken vascular networks. We applied the method to postprocess OCTA images of the microvascular networks in mouse cortex in vivo. Both image comparison and quantitative analysis show that the proposed method can significantly improve OCTA image by efficiently recovering microvasculature from the overwhelming motion artifacts.

Self-Supervised Bulk Motion Artifact Removal in Optical Coherence Tomography Angiography

“…To evaluate the performance of the proposed CABR framework, we compare it with the state-of-the-art OCTA motion correction model by Li et al [14]. To the best of our knowledge, there is no other publications that pertain directly to removing BMA in OCTA.…”

“…We evaluate the proposed BMA removal framework on mouse cortex OCTA datasets. The results are then compared with the state-of-the-art motion correction model [14] and GatedConv [28]. Here we re-implement Gated-Conv [28] for inpainting missing masks.…”

“…Then, we will briefly introduce some general denoising methods followed by OCT denoising methods. Note that the state-of-the-art method [14] is omitted here, as it has been discussed above.…”

“…This is an important yet underdeveloped topic in computer vision. The current sate-of-the-art model [14] corrects BMA with a combination of a BMA detection module, a vessel segmentation module and a mask inpainting module [28]. The BMA detection module predicts BMA affected rows in OCTA.…”

“…Considering the paucity of data and annotations, we correct motion artifacts via structural denoising, or BMA removal, rather than image denoising which usually demands numerous training data or corresponding clear and noisy image pairs. Inspired by the work of Li et al [14], our approach is formulated as a BMA recovery framework that solves the aforementioned issues and learns in a selfsupervised manner. First, the gradient-based structural information and appearance feature are extracted from the BMA area and injected into the model to capture more connectivity.…”