Related Experiment Video
Updated: Mar 10, 2026
06:08
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
1.8K
Adaptive multi-feature fusion architecture with optimized learning for high-fidelity brain tumor classification in
Mohammed Safy1, Mahmoud Khaled Abd-Ellah2, Esraa Salah Bayoumi3,4
1College of Computing and Information Technology, Arab Academy for Science, Technology and Maritime Transport (AASTMT), Smart Village, B 2401, Giza, Egypt.
Scientific Reports
|March 9, 2026
Summary
This study introduces an advanced computer-aided diagnostic framework for brain glioma detection. The novel method significantly improves the accuracy of distinguishing between high-grade glioma, low-grade glioma, and healthy brain tissue using MRI scans.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Brain gliomas are aggressive cancers with diagnostic challenges.
- Current computer-aided diagnostic (CAD) methods struggle to accurately differentiate glioma grades and healthy tissue on MRI.
- Early and precise diagnosis is crucial for effective treatment of brain gliomas.
Purpose of the Study:
- To develop and evaluate a novel multi-stage framework for improved classification of high-grade glioma (HG-G), low-grade glioma (LG-G), and healthy brain tissue.
- To enhance the accuracy and robustness of computer-aided diagnosis for brain gliomas using magnetic resonance images (MRI).
- To integrate advanced image processing techniques with deep learning and traditional machine learning for superior diagnostic performance.
Main Methods:
- A multi-stage framework incorporating Adaptive Gamma Correction (AGC) for contrast enhancement and a Denoising Convolutional Neural Network (DnCNN) for noise reduction.
- Feature extraction using deep representations from three fine-tuned transfer learning CNNs (TRCNNs) combined with Gray-Level Co-occurrence Matrix (GLCM) texture features, creating nine CNN-GLCM Fused Feature (CGFF) sets.
- Evaluation of CGFF sets using classifiers like Random Forest (RF), Extreme Gradient Boosting (XGBoost), LightGBM, and Support Vector Machine (SVM), with a stacked ensemble for stability.
Main Results:
- The proposed framework achieved high performance metrics: 99.05% accuracy, 98.99% recall, 99.52% specificity, 99.08% positive predictive value (PPV), and 99.54% negative predictive value (NPV).
- The hybrid feature sets (CGFF) combined with robust classifiers demonstrated superior performance compared to state-of-the-art (SOTA) methods across all evaluated metrics.
- Statistical verification using the Friedman test (p < 0.05) confirmed the reliability and significance of the performance improvements.
Conclusions:
- The developed multi-stage framework offers a highly accurate and robust solution for computer-aided diagnosis of brain gliomas from MRI.
- The integration of enhanced image preprocessing, deep learning features, and texture analysis provides a powerful approach for challenging medical image classification tasks.
- This advanced CAD system has the potential to significantly aid clinicians in the early and precise diagnosis of brain gliomas, improving patient outcomes.
