Medical image segmentation improved by dual-domain optimization framework
O2D-MedNet: An Optimization-Aware Dual-Domain Framework for Robust Medical Image Segmentation
Computer Vision and Pattern Recognition
Summary
Medical images can be hard to analyze because they often have noise and details that are difficult to separate. This paper presents a new approach called O2D-MedNet that looks at images in two ways, using different mathematical tools to reduce noise and keep important details. It also uses a technique to better combine different image features for clearer boundaries. Finally, it introduces a new way to help the computer learn steadily even when some parts are harder to recognize. The authors tested their method on many image types and found it performs better than existing ones.
What this means in practice
- •For medical imaging software developers: Implement better segmentation tools that handle noisy images and complex structures for improved diagnostic accuracy.
- •For radiology data teams: Improve robustness and consistency of lesion detection workflows across multiple scanning technologies and patient data conditions.
- •For automated pathology laboratories: Enhance automated lesion boundary detection systems to reduce manual correction time and improve throughput.$Commercial implications: This method enables commercial-grade automation of lesion segmentation that could be sold to pathology labs for faster analysis.
Authors
Shipeng Hu, Miaohui Wang, Jing Qin, Sukun Tian, Yuchun Sun
Abstract
Medical image segmentation (MIS) remains challenging due to composite degradations and multi-scale geometric variations, which jointly impair lesion-background separability, structural consistency, and optimization stability. Existing single-stream frameworks entangle heterogeneous cues within a single representation pathway, while hybrid-stream approaches often rely on generic or static fusion, limiting effective coordination across domains and failing to explicitly address optimization instability. To overcome these limitations, we propose a new optimization-aware dual-domain framework for MIS (O2D-MedNet). First, a degradation-aware dualdomain interaction (DDI) module integrates Fourier - and Wavelet-domain features to suppress degradation-sensitive responses while preserving structural and fine-detail information. Second, a cross-backbone contour-guided fusion (CGF) module enables selective interaction between CNN-based local detail modeling and visual state-space-based global structure modeling, improving boundary accuracy and representation consistency. Third, a gradient-context optimization (GCO) strategy stabilizes training by adaptively adjusting learning dynamics based on loss trends and gradient statistics, mitigating instability caused by class imbalance and heterogeneous sample difficulty. Extensive experiments on six benchmarks demonstrate consistent improvements over state-of-the-art MIS methods across multiple imaging modalities. Further analyses confirm enhanced sensitivity, structural consistency, and optimization stability, validating the effectiveness of the proposed O2D-MedNet for robust MIS under heterogeneous conditions.