Related Experiment Video
Updated: May 8, 2026
14:08
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
Published on: April 13, 2013
42.8K
A novel residual network based on multidimensional attention and pinwheel convolution for brain tumor classification
Jincan Zhang1, Rongfu Lv1, Wenna Chen2
1College of Information Engineering, Henan University of Science and Technology, Luoyang, China.
Scientific Reports
|August 23, 2025
Summary
A new Res-MAPNet model improves brain tumor classification using advanced attention and convolution techniques. This AI approach achieves high accuracy, offering a robust tool for medical diagnosis.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in healthcare
- Neuro-oncology
Background:
- Accurate brain tumor classification is crucial for effective treatment.
- Convolutional Neural Networks (CNNs) in medical imaging face limitations in feature extraction and focusing on critical details.
- Existing methods require enhancement for improved diagnostic accuracy.
Purpose of the Study:
- To introduce a novel Residual Network based on Multi-dimensional Attention and Pinwheel Convolution (Res-MAPNet) for Magnetic Resonance Imaging (MRI) based brain tumor classification.
- To enhance the feature extraction capabilities and focus on critical information in medical image analysis.
- To develop an efficient and robust solution for computer-aided diagnosis systems.
Main Methods:
- Development of the Res-MAPNet model incorporating Coordinated Local Importance Enhancement Attention (CLIA) and Pinwheel-Shaped Attention Convolution (PSAConv) modules.
- CLIA module integrates channel attention, spatial attention, and direction-aware positional encoding for lesion focus.
- PSAConv module utilizes asymmetric padding and grouped convolution to improve spatial feature perception and receptive field.
Main Results:
- The Res-MAPNet model achieved 99.51% accuracy in three-class brain tumor classification and 98.01% in four-class classification.
- Ablation studies demonstrated the significant contributions of CLIA and PSAConv modules, outperforming the ConvNeXt baseline by 4.41% and 4.45%, respectively.
- The proposed model outperformed existing mainstream models in brain tumor classification tasks.
Conclusions:
- Res-MAPNet offers an efficient and robust solution for brain tumor classification using MRI data.
- The novel CLIA and PSAConv modules effectively enhance feature extraction and focus on critical diagnostic information.
- The study presents a promising tool with potential for clinical applications in computer-aided diagnosis systems.
