Lung Cancer Detection: Leveraging Transfer Learning & AI Chatbot for Doctor Prescription

This research focuses on developing end to end pipeline for the Lung Cancer Detection using Large Language Models (LLMs) & Generative AI. In the field of healthcare there is an advancement of tools and technology used for detection and diagnostics of disease. This project involves mainly two aspects of technology: MLOps and LLMOps. First, it creates a classification model that classifies the disease with the help of user's CT scan image to detect lung cancer using transfer learning model and then build an End-to-End CI/CDI CT pipeline for disease detection by handling new incoming data and version controlling. Afterward, this pipeline is integrated with a chatbot that incorporates a question-and-answer system by utilizing LLM models like GPT and Llama to help patients in receiving appropriate therapy like medicine recommendation and doctor consultancy. A dataset that used comprising of 81,584 images from TCIA is taken which consists of different types of lung cancer images. A pretrained model LLM was utilized to generate accurate doctor prescriptions and the model was fine-tuned using comprehensive medical literature. Four cutting-edge convolutional neural network (CNNs) architecture are used as transfer learning models which are DenseNET121, combination of Inception with Mobile NET, ResNET50, and VGG16 etc. The model, upon implementing, achieved an accuracy of 99.58%, with error rate of Type II being nearly zero by using the combination of Inception and Mobilenet. After identifying the disease, the patient can directly interact with the AI chatbot that is integrated with the pipeline. Both patients and healthcare professionals get the advantage for disease detection and getting diagnostics report related to disease respectively.