End-to-End-Chest-Cancer-Classification-using-MLflow-DVC

MLOps End-to-End Pipeline 🚀

An end-to-end Machine Learning pipeline for automated classification of Adenocarcinoma, a common and deadly form of lung cancer, using deep learning (VGG16 + CNN) integrated with a full MLOps workflow.

This project deploys a production-grade solution combining:

• 🧪 MLflow for tracking experiments
• 📦 DVC for dataset/model versioning
• ☁️ Dagshub for remote collaboration
• 🐳 Docker for deployment
• ⚙️ CI/CD pipeline for automation
• 🌐 HTML Web UI for real-time predictions

⚕️ Medical Impact: Adenocarcinoma is one of the deadliest lung cancers. Early detection via deep learning can drastically improve outcomes — especially in underserved healthcare systems.

---

💡 Why Adenocarcinoma?

• 📈 Accounts for 40% of all lung cancer cases
• 🧬 Often misdiagnosed in early stages
• ⏰ Early AI-powered detection improves survival by up to 50%
• 🌍 Designed for clinician support in resource-limited settings

---

🔍 Summary

🔬 Problem	🧠 Solution	💥 Impact
Adenocarcinoma detection is error-prone & delayed	Trained VGG16 + CNN on annotated X-ray datasets	Improves diagnostic accuracy & saves time
Reproducibility in AI pipelines is hard	Integrated MLflow, DVC & GitHub CI/CD	Ensures traceability & accountability
Clinician access to ML tools is limited	Built web UI & Dockerized deployment	Empowers healthcare workers globally

---

🧠 Deep Learning Models

✅ VGG16 – Transfer Learning

• Pre-trained on ImageNet
• Fine-tuned for chest X-ray Adenocarcinoma detection
• High performance with minimal compute needs

✅ Custom CNN

• Built from scratch
• Lightweight & fast
• Suitable for edge deployment (rural clinics, mobile devices)

---

📈 Model Performance

Metric	VGG16	Custom CNN
Accuracy	94.6%	91.2%
Precision	93.0%	89.5%
Recall (Sensitivity)	94.8%	90.1%
F1-Score	93.8%	89.8%

---

🛠️ Tech Stack

Tool	Role
Python, Keras	Model development
MLflow	Experiment tracking
DVC	Data/model versioning
Dagshub	ML dashboard & Git remote
Docker	App containerization
Flask + HTML	UI development
GitHub Actions	CI/CD automation

---

🌐 Web App Workflow

• 📤 Upload a chest X-ray image
• 🧠 AI predicts: Adenocarcinoma Positive or Normal
• 🖥️ Results shown on the browser with diagnosis label

Launch via Docker:

docker build -t adenocarcinoma-detector .
docker run -p 5000:5000 adenocarcinoma-detector