---
license: mit
tags:
- medical-imaging
- image-segmentation
- white-matter-hyperintensities
- mri
- flair
- deep-learning
- tensorflow
- keras
- neurology
- multiple-sclerosis
datasets:
- custom
- msseg2016
metrics:
- dice-coefficient
- hausdorff-distance
library_name: tensorflow
pipeline_tag: image-segmentation
---

# **Neuro-AI: AI-Driven MS Lesion Analysis Framework**

# Ventricles & WMH Segmentation:

Pre-trained models for **ventricles and white matter hyperintensity (WMH) segmentation** with explicit distinction between normal periventricular changes (normal WMH) and pathological lesions (abnormal WMH).

## Model Description

This repository should contain 6 pre-trained deep learning models (6 architectures) for automated, simultaneous Ventricles and WMH segmentation from FLAIR MRI images. The models implement a novel **four-class approach** that distinguishes between:

- **Class 0**: Background
- **Class 1**: Ventricular system
- **Class 2**: Normal WMH (aging-related periventricular changes)
- **Class 3**: Abnormal WMH (pathologically significant lesions)


## Model Architectures

| Architecture | mean DSC | mean IoU | mean HD95 | mean Precision | mean Recall |
|--------------|----------|----------|-----------|----------------|-------------|
| **Baseline U-Net (WCE)** | 0.714 ±0.018 | 0.601 ±0.020 | 6.50 ±0.46 | 0.616 ±0.020 | **0.937 ±0.002** |
| **Baseline Pix2Pix (WCE)** | 0.823 ±0.011 | 0.721 ±0.014 | 5.31 ±0.20 | 0.805 ±0.005 | 0.848 ±0.010 |
| **Baseline Pix2Pix (UFL)** | 0.817 ±0.010 | 0.714 ±0.012 | 5.50 ±0.35 | 0.791 ±0.021 | 0.850 ±0.019 |
| **Pix2Pix + Attention Discriminator** | 0.824 ±0.008 | 0.723 ±0.011 | 5.23 ±0.31 | 0.807 ±0.014 | 0.843 ±0.009 |
| **Pix2Pix + Adaptive Hybrid Loss** | 0.844 ±0.002 | 0.750 ±0.003 | **4.81 ±0.05** | 0.845 ±0.009 | 0.843 ±0.010 |
| **Pix2Pix + Attention Discriminator + Adaptive Hybrid Loss** ⭐| **0.852 ±0.004** | **0.760 ±0.006** | 4.87 ±0.13 | **0.856 ±0.006** | 0.850 ±0.006 |

⭐ **Recommended**: Pix2Pix + Attention Discriminator + Adaptive Hybrid Loss (V5) for best performance

## Repository Structure

```
results_fold_2_var_5_bet_zscore/
└── models/standard_4class/fold_2
    └── best_dice_generator.h5      # 4-Class: Background, Ventricles, Normal, Abnormal

```

## Quick Start

### Installation

```bash
pip install huggingface_hub tensorflow numpy nibabel
```

### Download Models

```python
from huggingface_hub import hf_hub_download

# Download best performing model (V5)
model_path = hf_hub_download(
    repo_id="Bawil/neuro-ai",
    filename="results_fold_2_var_5_bet_zscore/models/standard_4class/fold_2/best_dice_generator.h5"
)

# Load model
from tensorflow.keras.models import load_model
model = load_model(model_path)
```

### Inference Example

```python
import numpy as np
from tensorflow.keras.models import load_model

# Load pre-trained model
model = load_model(model_path)

# Prepare input (256x256 grayscale FLAIR MRI, normalized)
# input_image shape: (batch_size, 256, 256, 1)
input_image = preprocess_flair(your_flair_image)

# Run inference
predictions = model.predict(input_image)

# Get class predictions
predicted_classes = np.argmax(predictions, axis=-1)
# 0: Background
# 1: Ventricles
# 2: Normal WMH (periventricular)
# 3: Abnormal WMH (pathological)

# Extract pathological lesions only
abnormal_mask = (predicted_classes == 2).astype(np.uint8)
```

## Training Data

### Dataset Composition

- **Local Dataset**: 300 MS patients (6,000 FLAIR MRI slices)
  - Demographics: 78 males, 222 females
  - Age range: 18-68 years
  - Scanner: 1.5-Tesla TOSHIBA Vantage
  
- **Public Dataset**: MSSEG2016 (15 patients, 750 FLAIR slices)

### Annotations

- Expert annotations by board-certified neuroradiologists (20+ years experience)
- Four-class labeling: Background, Ventricles, Normal WMH, Abnormal WMH
- Approved by Ethics Committee (IR.TBZMED.REC.1402.902)

### Data Split

- **Training**: 70% patients (local)
- **Validation**: 10% patients (local)
- **Testing**: 20% patients (local) + 40% patients (public)
- **Strategy**: Patient-level stratified split (no slice-level leakage)

## Model Training

### Configuration

- **Framework**: TensorFlow 2.11, Keras
- **Optimizer**: Adam (learning rate: 0.0002)
- **Loss Functions**:
  - Option 1: Weighted categorical cross-entropy
  - Option 2: Unified Focal Dice
- **Epochs**: 60 (with early stopping)
- **Batch Size**: 4
- **Input Size**: 256×256×1

### Hardware

- **GPU**: NVIDIA RTX 3060 (12GB VRAM)
- **Training Time**: 3-4 hours per model (5-fold CV)
- **Inference Time**: ~35-40ms per image

## Model Performance


## Use Cases

### Clinical Applications

- **MS Lesion Quantification**: Accurate measurement of disease burden
- **Differential Diagnosis**: Distinguish pathological from normal aging
- **Longitudinal Monitoring**: Track disease progression over time
- **Treatment Response**: Evaluate therapeutic efficacy
- **Radiological Reporting**: Reduce false positive alerts

### Research Applications

- **Baseline Comparisons**: Standardized evaluation framework
- **Method Development**: Foundation for advanced segmentation approaches
- **Multi-center Studies**: Protocol for broader validation
- **Reproducible Research**: Complete implementation available

## Limitations

- **Single Modality**: Trained on FLAIR MRI only
- **Scanner Specificity**: Primarily 1.5T TOSHIBA data
- **Disease Focus**: Optimized for MS patients
- **2D Segmentation**: Slice-by-slice processing (no 3D context)
- **Resolution**: Fixed 256×256 input size

## Model Card

### Intended Use

- **Primary**: Automated WMH segmentation for research and clinical decision support
- **Users**: Radiologists, neurologists, researchers, AI developers
- **Out-of-scope**: Not FDA/CE approved; not for standalone clinical diagnosis

### Ethical Considerations

- **Privacy**: All data anonymized per HIPAA/GDPR standards
- **Bias**: Limited scanner/protocol diversity may affect generalization
- **Clinical Validation**: Requires expert review before clinical use
- **Transparency**: Complete methodology and code openly available

### Model Card Authors

Mahdi Bashiri Bawil, Mousa Shamsi, Abolhassan Shakeri Bavil

## Citation

```bibtex
@article{bawil2026neuro,
  title={AI-Driven Multi-Parametric MS Lesion Analysis from T2-FLAIR Imaging: A Clinical Decision Support Framework for Neuroradiology},
  author={Bawil, Mahdi Bashiri and Shamsi, Mousa and Bavil, Abolhassan Shakeri},
  year={2026},
  note={Models: https://huggingface.co/Bawil/neuro-ai}
}
```

## License

MIT License - See [LICENSE](https://github.com/mri-ms-analyzer/neuro-ai/blob/main/LICENSE)

## Additional Resources

- **📄 Paper**: [Under Review]
- **💻 GitHub Repository**: [mri-ms-analyzer/neuro-ai](https://github.com/mri-ms-analyzer/neuro-ai)
- **📧 Contact**: m_bashiri99@sut.ac.ir | mehdi.bashiri.b@gmail.com
- **🏥 Institution**: Sahand University of Technology & Tabriz University of Medical Sciences

## Acknowledgments

- **Golgasht Medical Imaging Center**, Tabriz, Iran for providing clinical data
- Expert neuroradiologists for manual annotations
- Ethics Committee approval: IR.TBZMED.REC.1402.902

---

**Keywords**: Artificial intelligence (AI), multiple sclerosis, neuroradiology, clinical decision support, automated lesion analysis, deep learning, clinical AI