Paper on intepretable and privacy preserved lung cancer detection published in Biomedical Signal Processing and Control!
Excited to share our latest research published in Biomedical Signal Processing and Control!
We introduced FVCM-Net, a novel deep learning framework designed to tackle one of the most pressing challenges in healthcare: early and accurate detection of lung cancer.
🔍 Why this matters
Lung cancer remains a leading cause of cancer-related deaths worldwide. Early detection is critical—but building robust, generalizable models is difficult due to privacy concerns and data scarcity across institutions.
💡 What we built
FVCM-Net combines the strengths of:
- ✅ VGG16 + CBAM for attention-guided, explainable lung cancer detection
- 🔐 Federated Learning to preserve patient privacy while enabling collaborative model training across institutions
- 🧠 Ensemble Learning to boost predictive accuracy and robustness
🧪 Key Results
- 📈 Federated + Ensemble learning: 97.37% accuracy, 97.37% F1-score
- 🏆 Ensemble learning alone: 98.26% accuracy, enhancing clinical decision support
🧬 Interpretability matters
We integrated XAI techniques like SHAP and HiResCAM to visualize and explain model decisions—empowering radiologists with transparent AI support.
📚 Datasets used
Our model was trained and validated on diverse lung CT scan datasets:
- LIDC-IDRI
- IQ-OTH/NCCD
- Kaggle public dataset
- Additional online sources
🌐 Impact
FVCM-Net demonstrates how privacy-preserving, explainable AI can revolutionize medical imaging—supporting clinicians with accurate, interpretable insights. Grateful to my co-authors, mentors, and collaborators who made this possible. Let’s keep pushing the boundaries of AI in healthcare! Check out the full paper and source code here: Paper Link: https://www.sciencedirect.com/science/article/pii/S1746809425012303?dgcid=coauthor
Source Code: