Related Experiment Video
Updated: Sep 22, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
APU-Net: An Attention Mechanism Parallel U-Net for Lung Tumor Segmentation
Tao Zhou1,2, YaLi Dong1,2, HuiLing Lu3
1School of Computer Science and Engineering, North Minzu University, Yinchuan, Ningxia 750021, China.
Biomed Research International
|May 19, 2022
Summary
A novel APU-Net model accurately segments lung nodules from PET/CT scans, improving early lung cancer diagnosis. This advanced segmentation aids in identifying complex lesions, enhancing computer-aided diagnostic capabilities.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Lung cancer poses significant mortality risks, with early detection crucial for effective treatment.
- Pulmonary nodules represent early-stage lung cancer, but subtle symptoms often lead to delayed diagnosis.
Purpose of the Study:
- To develop an advanced deep learning model for precise segmentation of lung nodules in multimodal medical images.
- To improve the accuracy of identifying complex lesions and their adhesion to normal tissues for better computer-aided diagnosis.
Main Methods:
- Proposed APU-Net, a parallel U-Net architecture for multimodal feature extraction from CT, PET/CT, and PET images.
- Incorporated multimodal feature extraction blocks, a hybrid attention mechanism, and a multiscale feature aggregation block.
Main Results:
- Achieved high segmentation accuracy with DSC of 96.86% and Recall of 97.53% on a lung tumor 18FDG PET/CT dataset.
- Demonstrated improved segmentation of complex lesion shapes and their boundaries with normal tissues.
Conclusions:
- APU-Net significantly enhances lung nodule segmentation accuracy in multimodal medical imaging.
- The model holds positive implications for computer-aided diagnosis, facilitating earlier and more accurate lung cancer detection.
