Department of Biomedical Engineering, The Chinese University of Hong Kong, Hong Kong Special Administrative Region of China.
Department of Biomedical Engineering, The Chinese University of Hong Kong, Hong Kong Special Administrative Region of China.
Comput Biol Med. 2024 Jun;175:108459. doi: 10.1016/j.compbiomed.2024.108459. Epub 2024 Apr 9.
Diabetic retinopathy (DR) is the most common diabetic complication, which usually leads to retinal damage, vision loss, and even blindness. A computer-aided DR grading system has a significant impact on helping ophthalmologists with rapid screening and diagnosis. Recent advances in fundus photography have precipitated the development of novel retinal imaging cameras and their subsequent implementation in clinical practice. However, most deep learning-based algorithms for DR grading demonstrate limited generalization across domains. This inferior performance stems from variance in imaging protocols and devices inducing domain shifts. We posit that declining model performance between domains arises from learning spurious correlations in the data. Incorporating do-operations from causality analysis into model architectures may mitigate this issue and improve generalizability. Specifically, a novel universal structural causal model (SCM) was proposed to analyze spurious correlations in fundus imaging. Building on this, a causality-inspired diabetic retinopathy grading framework named CauDR was developed to eliminate spurious correlations and achieve more generalizable DR diagnostics. Furthermore, existing datasets were reorganized into 4DR benchmark for DG scenario. Results demonstrate the effectiveness and the state-of-the-art (SOTA) performance of CauDR. Diabetic retinopathy (DR) is the most common diabetic complication, which usually leads to retinal damage, vision loss, and even blindness. A computer-aided DR grading system has a significant impact on helping ophthalmologists with rapid screening and diagnosis. Recent advances in fundus photography have precipitated the development of novel retinal imaging cameras and their subsequent implementation in clinical practice. However, most deep learning-based algorithms for DR grading demonstrate limited generalization across domains. This inferior performance stems from variance in imaging protocols and devices inducing domain shifts. We posit that declining model performance between domains arises from learning spurious correlations in the data. Incorporating do-operations from causality analysis into model architectures may mitigate this issue and improve generalizability. Specifically, a novel universal structural causal model (SCM) was proposed to analyze spurious correlations in fundus imaging. Building on this, a causality-inspired diabetic retinopathy grading framework named CauDR was developed to eliminate spurious correlations and achieve more generalizable DR diagnostics. Furthermore, existing datasets were reorganized into 4DR benchmark for DG scenario. Results demonstrate the effectiveness and the state-of-the-art (SOTA) performance of CauDR.
糖尿病性视网膜病变（DR）是最常见的糖尿病并发症，通常会导致视网膜损伤、视力丧失，甚至失明。计算机辅助 DR 分级系统对帮助眼科医生进行快速筛查和诊断有重大影响。眼底摄影的最新进展促使新型视网膜成像相机的发展及其在临床实践中的后续应用。然而，大多数基于深度学习的 DR 分级算法在不同领域的泛化能力有限。这种较差的性能源于成像协议和设备的差异导致的领域转移。我们假设，模型在不同领域之间的性能下降是由于数据中存在虚假相关性。将因果分析中的 do-operations 纳入模型架构中可能会缓解这个问题并提高泛化能力。具体来说，我们提出了一种新的通用结构因果模型（SCM）来分析眼底成像中的虚假相关性。在此基础上，我们开发了一种基于因果关系的糖尿病性视网膜病变分级框架，命名为 CauDR，用于消除虚假相关性并实现更具泛化性的 DR 诊断。此外，我们重新组织了现有的数据集，形成了用于 DG 场景的 4DR 基准。结果表明，CauDR 的有效性和达到了最先进水平（SOTA）。
