BrainFusionNet: a deep learning and XAI model to understand local, global, and sequential features of MRI images for improved brain tumour detection

2026-06-17 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Computer Vision & Pattern Recognition · Depth: Expert, quick

Summary

BrainFusionNet is a novel deep learning and explainable AI (XAI) model designed to enhance brain tumor detection from Magnetic Resonance Imaging (MRI) by addressing noise and complex tumor characteristics. It integrates Convolutional Neural Networks (CNNs), Vision Transformers (ViT), and Gated Recurrent Units (GRUs) to extract spatial, contextual, and sequential features. The model achieved 98% accuracy on two public MRI datasets using K-fold validation, outperforming six state-of-the-art CNNs, including DenseNet121 and VGG16 which reached 96%. BrainFusionNet's novelty lies in its effective extraction of local and global features, even from small tumor regions, and its customized ViT for stable gradient flow. XAI techniques like SHAP, LIME, and GradCAM are incorporated to visualize decision-making. Additionally, the study found that MRI pixel intensity distribution significantly affects deep learning performance.

Key takeaway

For AI Scientists or Machine Learning Engineers developing medical imaging solutions, BrainFusionNet offers a robust blueprint for improving diagnostic accuracy and model transparency. Its hybrid architecture, combining CNNs, ViT, and GRUs, effectively handles complex MRI data, even for small tumors. You should consider adopting similar multi-modal deep learning and XAI integration strategies to enhance the reliability and interpretability of your brain tumor detection systems.

Key insights

BrainFusionNet combines CNNs, ViT, and GRUs with XAI for highly accurate and interpretable brain tumor detection from MRI.

Principles

• Hybrid DL models can overcome MRI noise challenges.
• XAI enhances interpretability of medical image decisions.
• Pixel intensity distribution impacts DL performance.

Method

BrainFusionNet uses CNNs for spatial features, ViT for contextual/local features, and GRUs for sequential features, feeding CNN/ViT outputs into GRU for final classification, with XAI for visualization.

In practice

• Integrate CNNs, ViT, GRUs for complex image analysis.
• Apply SHAP, LIME, GradCAM for model interpretability.
• Consider MRI pixel intensity distribution in model design.

Topics

• BrainFusionNet
• Deep Learning
• Explainable AI
• MRI
• Brain Tumor Detection
• Hybrid Models

Best for: Computer Vision Engineer, AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• DB-FGA-Net: Dual Backbone Frequency Gated Attention Network for Multi-Class Brain Tumor Classification with Grad-CAM Interpretability — DB-FGA-Net offers robust, interpretable, and augmentation-free brain tumor classification using dual-backbone and frequency-gated attention.
• Toward Reliable Tea Leaf Disease Diagnosis Using Deep Learning Model: Enhancing Robustness With Explainable AI and Adversarial Training — Deep learning with adversarial training and XAI enhances tea leaf disease detection robustness and interpretability.
• Made and Published a Paper Comparing Analysis of CNN and Vision Transformer Architectures for Brain Tumor Detection [R] — CNNs and Vision Transformers were compared for brain tumor detection in MRI scans.
• IEEE Transactions on Neural Networks and Learning Systems, Volume 37, Issue 3, March 2026 — The issue highlights advancements across neural networks, computer vision, and machine learning applications.
• ResNet-34 with Lightweight Decoder for Accurate and Efficient Segmentation of Fetal Brain MRI — A ResNet-34 encoder with a lightweight MLP decoder achieves accurate and efficient fetal brain MRI segmentation, outperforming baselines.
• AI algorithm enables tracking of vital white matter pathways — An AI tool precisely segments brainstem white matter bundles in live MRI, revealing disease-specific structural changes and recovery.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.