LungLens
Language
About LungLens

Built to make chest X-ray learning accessible

LungLens is a medical chest X-ray analysis and education tool designed to help people better understand imaging results in plain language. The focus is health literacy: helping users ask better questions, not replacing professional care.

Project Story
Research collaboration + independent product build

The machine learning model behind LungLens was developed as part of an MSc group project at CUHK Chinese University Hong Kong.

I then built this web application independently so the tool could be freely accessible to everyone in a clean, easy-to-use format.

Meet the Team

Engineers, researchers, and builders who trained the models and shaped the LungLens experience.

Charles Tsoi

Full-Stack Developer & Architect

Designed the system architecture and developed the frontend web application and backend API integration. Trained, evaluated, and deployed Model 3 (DenseNet-121) with integrated Grad-CAM visual interpretability.

Casper Lee

Vision AI Researcher

Developed and optimized two computer vision pipelines for the ensemble system: Model 1 (ResNet-50) for primary feature classification and Model 4 (Swin Transformer Tiny) for advanced pattern recognition.

Edward Choi

Vision AI Researcher & Clinical Logic

Developed and trained Model 2 (ResNet-152V2). Designed the conditional user intake questionnaire to capture patient context and supply structured clinical inputs to the LLM diagnostic module.

Jenna Tse

Systems Optimization & Presentation Specialist

Handled backend performance tuning, code quality improvements, and repository maintenance using AI-assisted engineering tools. Co-authored reporting templates and led the final presentation design.

Dicky Ng

Vision AI Researcher

Conducted data analysis on medical reporting standards and deploying Model 5 (DenseNet-121) to scale our diagnostic ensemble.

Medical Disclaimer

This tool is for educational and research purposes only.

It is NOT a substitute for professional medical diagnosis.

Always consult a qualified healthcare professional.

Tech Stack

Model: PyTorch, trained on [dataset name, e.g. NIH ChestX-ray14]

Frontend: Next.js, Tailwind CSS

Deployment: [Railway / Cloud Run / etc.]

Open Source / Contact
Interested in collaboration, feedback, or contributing?