NeuroEnsemble8
Hybrid Brain MRI Tumor Analysis Framework
A multi-model deep learning framework for tumor classification, segmentation, explainability, and reliability analysis from MRI.
Repository Links
Source Code
https://github.com/tharunsridhar/Brain-Tumor-mri-AI-Analysis-System
Model Repository
This Hugging Face repository contains trained weights and inference artifacts.
Overview
MRI Input
↓
EfficientNetV2-S
MobileNetV3
ConvNeXt Tiny
↓
Adaptive Ensemble Fusion
↓
Tumor Segmentation
↓
Grad-CAM Explainability
↓
Diagnostic Reliability Index (DRI)
↓
Final Diagnostic Output
Core Models
| Model | Architecture |
|---|---|
| EfficientNetV2-S | EfficientNetV2 Small |
| MobileNetV3 | MobileNetV3 Large |
| ConvNeXt Tiny | ConvNeXt Tiny |
Classes:
- Glioma
- Meningioma
- Pituitary Tumor
- No Tumor
Segmentation
BRISC-EffUNet
Experimental Results
Training Curves
- docs/ConvNext tiny graphs.png
- docs/mobilenet graph.png
- docs/v2s graph.png
- docs/Segmentation graph.png
Confusion Matrices
- docs/convNext tiny confustion matrix.png
- docs/mobilenet cm.png
- docs/v2s confustion matrix.png
Performance
Classification
| Model | Accuracy | Precision | Recall | F1-Score |
|---|---|---|---|---|
| EfficientNetV2-S | 97.78% | 98.09% | 97.24% | 98.00% |
| MobileNetV3 | 96.57% | 97.21% | 94.67% | 96.00% |
| ConvNeXt Tiny | 94.85% | 96.35% | 93.57% | 95.00% |
Segmentation Performance
| Metric | Score |
|---|---|
| Dice Score | 87.98% |
| IoU | 78.60% |
| Segmentation Loss | 0.1162 |
Advanced Features
- Adaptive Ensemble Fusion
- Grad-CAM Explainability
- Diagnostic Reliability Index (DRI)
- Lesion-aware Risk Scoring
Repository Structure
models/
docs/
examples/
inference.py
requirements.txt
README.md
Quick Start
pip install -r requirements.txt
python inference.py
Intended Use
Research and educational use only. Not for standalone clinical diagnosis.
Citation
@software{tharun_neuroensemble8,
title={NeuroEnsemble8: Hybrid Brain Tumor Analysis Framework},
author={Tharun Sridhar},
year={2026}
}
Author
Created and released openly for research and societal benefit.