Related Experiment Video
Updated: Mar 6, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
851
MAAR-Net: Multi-scale attention-assisted residual neural network for renal microvascular structure segmentation
Tingting Wang1, Baoguang Lin1, Tong Jiang1
1Department of Radiationtherapy, General Hospital of Northern Theater Command, Shenyang, Liaoning, China.
Plos One
|March 4, 2026
Summary
A new deep learning model, MAAR-Net, accurately segments renal microvasculature in kidney histology images. This advancement aids in evaluating renal disease progression and enables real-time clinical diagnostics.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Nephrology
Background:
- Renal disease poses a significant public health challenge, with microvascular lesions driving progression.
- Accurate segmentation of renal microvasculature is critical for pathological evaluation.
- Existing deep learning models face challenges in segmenting complex renal microvessels, affecting accuracy, continuity, and boundary delineation.
Purpose of the Study:
- To develop a novel deep learning architecture for accurate renal microvessel segmentation.
- To address limitations of existing models in handling complex vessel structures and background noise.
- To create a computationally efficient model suitable for real-time clinical applications.
Main Methods:
- Proposed Multiscale Attention-Assisted Residual Neural Network (MAAR-Net) based on a U-Net architecture.
- Integrated multiscale residual blocks, high-semantic feature extraction, and depth-separable convolutional attention blocks.
- Employed additional segmentation branches for multi-receptive-field information aggregation.
- Validated on the HuBMAP dataset of 2D PAS-stained kidney histology images.
- Optimized the model using structured pruning and quantification for real-time performance.
Main Results:
- MAAR-Net achieved an Intersection over Union (IoU) of 0.5063 and an F1-score of 0.6751 on the HuBMAP dataset.
- The model outperformed mainstream segmentation methods in accuracy and efficiency.
- Optimized MAAR-Net demonstrated suitability for real-time diagnostic applications without requiring specialized hardware.
Conclusions:
- MAAR-Net offers a robust and practical solution for accurate renal microvessel segmentation.
- The model's performance and efficiency facilitate its deployment in clinical settings for improved renal disease assessment.
- The developed approach enhances the potential for precise pathological evaluation and timely diagnosis of kidney diseases.
