Keywords
CNN, LLM, Mobile Application

Technologies
Python, PyTorch, React Native, MongoDB, Docker

This project aims to develop a multimodal AI system for the early detection of lung diseases, with a particular focus on conditions increasingly prevalent among young adults. The system is designed to identify potential lung abnormalities from medical imaging data and provide users with interpretable, personalised insights into their lung health through a conversational interface.

The implemented system integrates a convolutional neural network for image classification, trained using transfer learning on lung datasets. A large language model is employed to combine imaging results with patient-reported symptoms, delivering clinical explanations through a multi-turn conversational interface.

The frontend is built with React Native, enabling cross-platform access across web and mobile devices, supported by a FastAPI backend with MongoDB for storage of user records and conversation histories. This project demonstrates the feasibility of combining CNN models and large language models to make early lung disease screening more accessible to everyday users.