DB-FGA-Net: Dual Backbone Frequency Gated Attention Network for Multi-Class Brain Tumor Classification with Grad-CAM Interpretability

2026-04-21 · Source: cs.AI updates on arXiv.org · Field: Science & Research — Health & Medical Research, Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences · Depth: Expert, extended

Summary

DB-FGA-Net is a novel deep learning framework for multi-class brain tumor classification from MRI images, integrating VGG16 and Xception dual backbones with a Frequency-Gated Attention (FGA) block. This model captures complementary local and global features, achieving high performance without data augmentation, which enhances its robustness and generalizability. It attained 99.24% accuracy on the 7K-DS dataset for 4-class classification, 98.68% for 3-class, and 99.85% for 2-class settings. Cross-dataset validation on the independent 3K-DS dataset showed 95.77% accuracy. The framework also incorporates Grad-CAM for interpretability, visualizing tumor regions, and includes a Python Tkinter-based graphical user interface (GUI) for real-time classification and tumor localization, bridging the gap between model prediction and clinical application.

Key takeaway

For Computer Vision Engineers developing medical imaging solutions, DB-FGA-Net demonstrates that high accuracy and interpretability in brain tumor classification can be achieved without relying on data augmentation. You should consider integrating dual-backbone architectures and frequency-gated attention mechanisms to enhance feature extraction and ensure robust generalization across diverse datasets, while also prioritizing explainable AI tools like Grad-CAM for clinical trust.

Key insights

DB-FGA-Net offers robust, interpretable, and augmentation-free brain tumor classification using dual-backbone and frequency-gated attention.

Principles

• Combine diverse CNN backbones for richer feature extraction.
• Integrate frequency domain attention for enhanced texture analysis.
• Prioritize interpretability for clinical trust and adoption.

Method

The DB-FGA-Net methodology involves dual VGG16 and Xception backbones, enhanced by a Frequency-Gated Attention (FGA) block, followed by global average pooling, dropout, and a softmax classifier. It is trained with Adam optimizer and categorical cross-entropy loss.

In practice

• Use Grad-CAM to visualize model's decision-making regions.
• Develop GUIs for real-time, interactive clinical deployment.
• Evaluate models on independent datasets to confirm generalization.

Topics

• DB-FGA-Net
• Frequency-Gated Attention
• Brain Tumor Classification
• Grad-CAM Interpretability
• Dual Backbone Network

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• BrainFusionNet: a deep learning and XAI model to understand local, global, and sequential features of MRI images for improved brain tumour detection — BrainFusionNet combines CNNs, ViT, and GRUs with XAI for highly accurate and interpretable brain tumor detection from MRI.
• Frequency-Enhanced Dual-Subspace Networks for Few-Shot Fine-Grained Image Classification — FEDSNet improves few-shot fine-grained image classification by integrating frequency domain features to enhance structural stability and reduce overfitting.
• Multi-layer feature aggregation network with residual module and attention mechanism for jaw cyst image segmentation — MFA-Net enhances medical image segmentation by integrating residual modules, channel-spatial attention, and multi-layer feature aggregation.
• Entropy-Guided Self-Supervised Learning for Medical Image Classification — Combining ImageNet and entropy-guided MAE pre-training via ensemble significantly boosts medical image classification accuracy and robustness.
• DDF2Pol: A Dual-Domain Feature Fusion Network for PolSAR Image Classification — DDF2Pol fuses real and complex PolSAR features with attention for efficient, high-accuracy classification.
• DualGate-Net: A Prior-Gated Dual-Encoder Framework for Histopathology Cell Detection — Adaptive, gated fusion of local and global features with contextual priors improves histopathology cell detection.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.