Multi-style conversion for semantic segmentation of lesions in fundus images by adversarial attacks
This addresses the challenge of expensive and incompatible segmentation databases for diabetic retinopathy diagnosis, though it appears incremental as it builds on existing adversarial methods.
The paper tackles the problem of inconsistent annotation styles across different fundus image databases for diabetic retinopathy segmentation by introducing adversarial style conversion, which allows a single model to adapt its segmentation style based on input and achieve improved generalization through dataset combination.
The diagnosis of diabetic retinopathy, which relies on fundus images, faces challenges in achieving transparency and interpretability when using a global classification approach. However, segmentation-based databases are significantly more expensive to acquire and combining them is often problematic. This paper introduces a novel method, termed adversarial style conversion, to address the lack of standardization in annotation styles across diverse databases. By training a single architecture on combined databases, the model spontaneously modifies its segmentation style depending on the input, demonstrating the ability to convert among different labeling styles. The proposed methodology adds a linear probe to detect dataset origin based on encoder features and employs adversarial attacks to condition the model's segmentation style. Results indicate significant qualitative and quantitative through dataset combination, offering avenues for improved model generalization, uncertainty estimation and continuous interpolation between annotation styles. Our approach enables training a segmentation model with diverse databases while controlling and leveraging annotation styles for improved retinopathy diagnosis.