EEA-UNet: An efficient element-wise adaptive attention-based network for abdominal multi-organ segmentation

Area of Science:

• Medical image analysis
• Computer vision
• Radiology

Background:

• Accurate segmentation of abdominal organs in X-ray computed tomography (CT) is vital for clinical applications.
• Existing methods struggle with long-range dependencies and boundary segmentation due to anatomical complexity.

Purpose of the Study:

• To develop an efficient abdominal multi-organ segmentation model (EEA-UNet) that addresses limitations in current approaches.
• To enhance the modeling of contextual dependencies and improve boundary segmentation accuracy.

Main Methods:

• Proposed an efficient element-wise adaptive (EEA) attention mechanism in skip connections for improved feature interaction and receptive field.
• Introduced an enhanced multi-scale feature fusion (EMF) module to boost decoding capabilities.
• Utilized an edge-awareness composite loss function for precise segmentation of small organs and boundaries.

Main Results:

• EEA-UNet achieved a Dice score of 84.45% and HD95 of 0.16 on the Synapse dataset.
• Demonstrated a strong balance between segmentation accuracy and computational efficiency.
• Outperformed several existing methods in both visual and quantitative evaluations.

Conclusions:

• EEA-UNet offers a competitive and efficient solution for abdominal multi-organ segmentation in CT images.
• The proposed attention and fusion modules effectively handle complex anatomical variations and boundary challenges.
• This model shows significant potential for improving automated medical image analysis and clinical decision-making.