MGFI-Net: A Multi-Grained Feature Integration Network for Enhanced Medical Image Segmentation

TL;DR

MGFI-Net addresses the core MIS challenge of accurately delineating complex structures in noisy, low-contrast images by integrating multi-grained contextual and local features while preserving edge details. The two novel modules, MGFI and Adaptive Edge, enable hierarchical feature fusion and deformable-edge refinement, trained with a hybrid loss that emphasizes both pixel-wise accuracy and boundary precision. Across CVC-ClinicDB, ISIC-2018, and the 2018 Data Science Bowl, MGFI-Net achieves state-of-the-art segmentation performance with favorable efficiency, demonstrating real-time viability. This work advances boundary-aware MIS and offers a practical, scalable approach for clinical imaging tasks.

Abstract

Medical image segmentation plays a crucial role in various clinical applications. A major challenge in medical image segmentation is achieving accurate delineation of regions of interest in the presence of noise, low contrast, or complex anatomical structures. Existing segmentation models often neglect the integration of multi-grained information and fail to preserve edge details, which are critical for precise segmentation. To address these challenges, we propose a novel image semantic segmentation model called the Multi-Grained Feature Integration Network (MGFI-Net). Our MGFI-Net is designed with two dedicated modules to tackle these issues. First, to enhance segmentation accuracy, we introduce a Multi-Grained Feature Extraction Module, which leverages hierarchical relationships between different feature scales to selectively focus on the most relevant information. Second, to preserve edge details, we incorporate an Edge Enhancement Module that effectively retains and integrates boundary information to refine segmentation results. Extensive experiments demonstrate that MGFI-Net not only outperforms state-of-the-art methods in terms of segmentation accuracy but also achieves superior time efficiency, establishing it as a leading solution for real-time medical image segmentation.

Figures (6)

• Figure 1: This set of images shows segmentation results before and after applying the Adaptive Edge module. (a) The original image, (b) the ground truth, (c) the result without the Adaptive Edge module, with blurred edges, and (d) the result with the module, showing better edge preservation and accuracy. This figure demonstrates the effectiveness of the module in capturing complex edge details and improving segmentation performance.
• Figure 2: The architecture of the proposed MGFI-Net for medical image segmentation. The model starts with a convolutional encoder that extracts hierarchical features, which are then processed through the MGFI module. This module selectively focuses on the most relevant information for accurate region delineation. Meanwhile, the AE module, which applies deformable convolutions, improves edge segmentation by refining boundary details.
• Figure 3: The structure of the MGFI module. The upper part of the module is responsible for extracting local features and global context information. The lower part includes three different convolution types: deformable convolution, atrous convolution, and standard convolution. These are applied in parallel to capture multi-grained features, which are then fused through channel concatenation.
• Figure 4: Sample results of polyp segmentation. From left to right: input image, ground truth, SOTA results obtained by MGFI-Net, CE-Net, Attention U-Net, U-Net++, U-Net, KiU-Net, MedFormer.
• Figure 5: Sample results of nuclei segmentation. From left to right: input image, ground truth, SOTA results obtained by MGFI-Net, CE-Net, Attention U-Net, U-Net++, U-Net, KiU-Net, MedFormer.