Search papers, labs, and topics across Lattice.
This study explores the impact of input resolution on the segmentation of diabetic retinopathy lesions, revealing that higher resolutions improve detection of microaneurysms but can degrade performance on larger hemorrhages. By systematically testing various architectures, including U-Net and Vision Transformers at different resolutions, the authors uncover a counter-intuitive relationship between resolution and lesion type. To address these conflicting demands, they introduce a Multi-Resolution Feature Stem that allows for parallel processing of multiple scales, effectively balancing fine detail capture with contextual integrity.
Higher input resolution can hinder segmentation performance for larger diabetic retinopathy lesions, challenging the assumption that more detail is always better.
Diabetic Retinopathy (DR) is a leading cause of preventable blindness worldwide, requiring automated lesion segmentation using deep learning models for early detection and monitoring. However, DR lesions vary dramatically in size from tiny microaneurysms to large hemorrhages and exudates. This variability creates conflicting demands on the model architecture and input resolution, posing a challenge for effective design. This work investigates the impact of input resolution on different lesion types. Through systematic experimentation with multiple architectures (U-Net, UNet++, Vision Transformers, DeepLabV3+) at $512 \times 512$ and $1024 \times 1024$ resolutions, we identify a critical, counter-intuitive phenomenon where increasing input resolution has opposing effects on different lesion types. We demonstrate that while higher resolution is essential for resolving fine-grained microaneurysms, it can unexpectedly degrade performance on larger hemorrhages. This finding challenges the common assumption that higher resolution is uniformly beneficial. To address this, we propose a novel Multi-Resolution Feature Stem, an input-level pyramid integrated with a UNet++ backbone. This architecture processes multiple scales in parallel, capturing fine-grained details without sacrificing contextual information. This work contributes crucial empirical evidence of this complex, resolution-dependent behavior and a practical, parameterefficient architecture that successfully resolves this trade-off. Our code is available at https://github.com/taeheej/Multi-Resolution-Feature-Stem-for-Diabetic-Retinopathy-lesion-segmentation.
