LIAUNet: Rethinking the Application of Loss Information in Medical Image Segmentation

Area of Science:

• Medical Imaging
• Deep Learning
• Computer Vision

Background:

• Medical image segmentation is vital for clinical diagnosis and treatment.
• Encoder-decoder networks are common but lose critical details during down-sampling.
• Effective utilization of lost information is key to improving segmentation performance.

Purpose of the Study:

• To propose a novel network, the Loss Information Aggregation Network (LIAUNet), for medical image segmentation.
• To address the information loss issue in encoder-decoder architectures.
• To enhance segmentation accuracy by effectively aggregating lost details.

Main Methods:

• Designed a MultiScale Hybrid Pooling Loss Extraction Module (MHPLE) to extract and adapt loss information across four scales.
• Developed an Information-Preserving multiscale Attention Aggregator (IMAA) to aggregate extracted loss information, preserving feature integrity.
• Incorporated a Spectral-Gated Convolution Block (SGCB) to enhance joint modeling of frequency and spatial domain features.

Main Results:

• LIAUNet demonstrated superior segmentation accuracy compared to existing methods on multiple medical image datasets.
• The proposed modules (MHPLE, IMAA, SGCB) effectively addressed information loss during segmentation.
• Experimental results validate the efficacy of LIAUNet in medical image segmentation tasks.

Conclusions:

• LIAUNet offers an effective approach to leverage lost information in medical image segmentation.
• The network architecture enhances feature representation and reduces information sparsity.
• This method provides significant support for advancing medical imaging and clinical diagnosis.