GCViT–PIO Medical Imaging

A hybrid Global Context Vision Transformer (GCViT) + Perceiver IO framework for multi-disease medical image classification — brain MRI, chest X-rays, and dermatology images in one pipeline.

Quickstart

git clone https://github.com/habib-analyst/GCViT-PIO-Medical-Imaging
cd GCViT-PIO-Medical-Imaging
pip install tensorflow keras numpy matplotlib scikit-learn

Open GCVIT_PIO_Habib_Ur_Rehman.ipynb and run it top to bottom. The notebook is self-contained: data preprocessing, the GCViT backbone, the Perceiver IO fusion block, training (early stopping + LR scheduling), and evaluation with confusion matrices and visualizations.

Input format: RGB images at 224×224, multi-class or binary labels, in train/val/test splits; normalization and augmentation are handled inside the notebook.

Results (as reported in the paper)

Test metrics from the published work (accuracy / precision / recall / F1):

Domain	Accuracy	Precision	Recall	F1
Neuro (stroke/Alzheimer's MRI)	1.00	1.00	1.00	1.00
Skin (melanoma/tinea)	0.95	0.95	0.95	0.95
Lung (pneumonia/lung cancer)	1.00	1.00	1.00	1.00

25,000 Kaggle images total (10k neuro MRI, 10k dermoscopic, 5k chest X-ray/CT), 70/15/15 split, standard config (224×224×3 input, batch 32, LR 1e-4, 5–20 epochs).

How it works

1. GCViT backbone — multi-scale hierarchical patch embedding with a hybrid CNN+Transformer stem extracts global-context features.
2. Perceiver IO fusion — latent cross-attention fuses the multi-scale features efficiently.
3. Classification head — fully connected layers with dropout and softmax for multi-class diagnosis.

Citation

@article{rehman2025gcvitpio,
  author  = {Habib Ur Rehman et al.},
  title   = {AI-Driven Multi-Disease Classification: {GCViT} \& {Perceiver IO} Synergistic Framework Revolutionizing Multi-Domain Medical Diagnosis Through Advanced Hybrid Neural Architecture},
  journal = {The Journal of Supercomputing},
  publisher = {Springer},
  year    = {2025},
  note    = {Accepted; DOI pending}
}

License

For research and educational use. © 2026 Habib Ur Rehman

---

Built by Habib Ur Rehman · ML/AI Engineer & healthcare-AI researcher · LinkedIn · habib.top