P-OCT (Peripapillary OCT Images)

用于视网膜光学相干断层扫描 (OCT) 图像的准确和自动化的组织分割算法对于青光眼的诊断至关重要。然而，由于视盘的存在，视网膜视乳头周围区域的解剖结构复杂，分割难度大。为了解决这个问题，我们开发了一种新的图卷积网络（GCN）辅助的两阶段框架，以同时标记九个视网膜层和视盘。具体来说，在 U 形神经网络的编码器和解码器之间插入了一个多尺度全局推理模块，以利用解剖学先验知识并执行空间推理。我们对人视乳头周围视网膜 OCT 图像进行了实验。所提出的分割网络的 Dice 得分为 0.820 ± 0.001，像素精度为 0.830 ± 0.002，两者均优于其他最先进技术。

Accurate and automated tissue segmentation algorithms for retinal optical coherence tomography (OCT) images are critical for the diagnosis of glaucoma. However, due to the presence of the optic disc, the anatomical structure of the peripapillary retinal region is complex, making segmentation challenging. To address this issue, we developed a novel graph convolutional network (GCN)-assisted two-stage framework for simultaneously annotating nine retinal layers and the optic disc. Specifically, a multi-scale global reasoning module was inserted between the encoder and decoder of the U-shaped neural network to leverage anatomical prior knowledge and perform spatial reasoning. We conducted experiments on human peripapillary retinal OCT images. The proposed segmentation network achieved a Dice score of 0.820 ± 0.001 and a pixel accuracy of 0.830 ± 0.002, both outperforming other state-of-the-art methods.

搜集汇总

背景与挑战

背景概述

P-OCT数据集包含人视乳头周围视网膜光学相干断层扫描（OCT）图像，用于开发自动分割算法以辅助青光眼诊断。该数据集支持一种基于图卷积网络的两阶段框架，能同时标记九个视网膜层和视盘，在实验中表现出优于现有技术的分割性能。