Lung Cancer Detection using EfficientNet-B0
A research-grade deep learning model for lung cancer CT slice classification using the IQ-OTH/NCCD dataset.
Overview
This model classifies CT slices into three categories:
The model was trained using transfer learning with EfficientNet-B0 and achieved strong performance on the IQ-OTH/NCCD benchmark dataset.
Performance
| Metric | Score |
|---|
| Accuracy | 97.59% |
| Weighted F1 Score | 97.59% |
| ROC-AUC | 99.63% |
Model Details
- Architecture: EfficientNet-B0
- Framework: PyTorch
- Training Strategy: Transfer Learning
- Input Size: 224 × 224
- Classes: 3
- Checkpoint Format: Full PyTorch checkpoint
The uploaded checkpoint contains:
- model_name
- state_dict
- class_names
- config
Features
- Lung cancer CT slice classification
- Transfer learning with ImageNet pretrained weights
- Explainability support through Grad-CAM
- Compatible with a Streamlit clinical-style dashboard
- Research-focused workflow and benchmarking pipeline
Dataset
Trained on the IQ-OTH/NCCD Lung Cancer Dataset.
Classes:
Limitations
- Slice-level classification only
- Uses JPG CT slices rather than DICOM images
- Not intended for clinical diagnosis
- Research and educational use only
License
Please follow the original dataset license and usage restrictions.
Citation
If you use this model, please cite the original IQ-OTH/NCCD dataset and associated research publications.
