Hu Songqi, Tang Hongying, Luo Yuemei
School of Information Engineering, Shanghai University of Maritime, 1550 Haigang Avenue, Shanghai 201306, China.
School of Information, Mechanical and Electrical Engineering, Shanghai Normal University, 100 Haisi Road, Shanghai 201418, China.
Biomed Opt Express. 2024 Jul 31;15(8):4980-4994. doi: 10.1364/BOE.532482. eCollection 2024 Aug 1.
Retinopathy detection using optical coherence tomography (OCT) images has greatly advanced with computer vision but traditionally requires extensive annotated data, which is time-consuming and expensive. To address this issue, we propose a novel contrastive graph regularization method for detecting retinopathies with less labeled OCT images. This method combines class prediction probabilities and embedded image representations for training, where the two representations interact and co-evolve within the same training framework. Specifically, we leverage memory smoothing constraints to improve pseudo-labels, which are aggregated by nearby samples in the embedding space, effectively reducing overfitting to incorrect pseudo-labels. Our method, using only 80 labeled OCT images, outperforms existing methods on two widely used OCT datasets, with classification accuracy exceeding 0.96 and an Area Under the Curve (AUC) value of 0.998. Additionally, compared to human experts, our method achieves expert-level performance with only 80 labeled images and surpasses most experts with just 160 labeled images.
利用光学相干断层扫描（OCT）图像进行视网膜病变检测在计算机视觉的助力下取得了巨大进展，但传统方法需要大量带注释的数据，既耗时又昂贵。为解决这一问题，我们提出了一种新颖的对比图正则化方法，用于在使用较少标注OCT图像的情况下检测视网膜病变。该方法将类别预测概率和嵌入的图像表示相结合进行训练，其中这两种表示在同一训练框架内相互作用并共同进化。具体而言，我们利用记忆平滑约束来改进伪标签，这些伪标签由嵌入空间中的邻近样本聚合而成，有效减少了对错误伪标签的过度拟合。我们的方法仅使用80张带标注的OCT图像，在两个广泛使用的OCT数据集上优于现有方法，分类准确率超过0.96，曲线下面积（AUC）值为0.998。此外，与人类专家相比，我们的方法仅用80张带标注的图像就能达到专家级性能，而仅用160张带标注的图像就能超越大多数专家。
