Retinal vessel optical coherence tomography images for anemia screening

Anemia is a disease that leads to low oxygen carrying capacity in the blood. Early detection of anemia is critical for the diagnosis and treatment of blood diseases. We find that retinal vessel optical coherence tomography (OCT) images of patients with anemia have abnormal performance because the internal material of the vessel absorbs light. In this study, an automatic anemia screening method based on retinal vessel OCT images is proposed. The method consists of seven steps, namely, denoising, region of interest (ROI) extraction, layer segmentation, vessel segmentation, feature extraction, feature dimensionality reduction, and classification. We propose gradient and threshold algorithm for ROI extraction and improve region growing algorithm based on adaptive seed point for vessel segmentation. We also conduct a statistical analysis of the correlation between hemoglobin concentration and intravascular brightness and vascular shadow in OCT images before feature extraction. Eighteen statistical features and 118 texture features are extracted for classification. This study is the first to use retinal vessel OCT images for anemia screening. Experimental results demonstrate the accuracy of the proposed method is 0.8358, which indicates that the method has clinical potential for anemia screening.

贫血是一种导致血液携氧能力降低的疾病。贫血的早期检测对血液疾病的诊断和治疗至关重要。我们发现贫血患者的视网膜血管光学相干断层扫描（OCT）图像由于血管内部物质吸收光线而表现异常。在这项研究中，提出了一种基于视网膜血管OCT图像的自动贫血筛查方法。该方法包括七个步骤，即去噪、感兴趣区域（ROI）提取、层分割、血管分割、特征提取、特征降维和分类。我们提出了用于ROI提取的梯度和阈值算法，并改进了基于自适应种子点的区域生长算法用于血管分割。在特征提取之前，我们还对血红蛋白浓度与OCT图像中血管内亮度和血管阴影之间的相关性进行了统计分析。提取了18个统计特征和118个纹理特征用于分类。这项研究首次使用视网膜血管OCT图像进行贫血筛查。实验结果表明，所提方法的准确率为0.8358，这表明该方法在贫血筛查方面具有临床潜力。