LUNG DISEASES DETECTION USING DEEP LEARNING

Abstract

The Artificial Intelligence (AI) techniques are nowadays a very effective element in the process of evolution of medicine. The ability of AI to make computers perform functions associated with the human mind like problem solving is well recognized these days. However, there is less recognition of how AI is employed in certain industries such as healthcare. Integrating AI into the healthcare ecosystem has many benefits including the ability to automate processes and analyse huge patient information, which helps deliver better patient treatment, faster and at lower cost. Any problem with the lungs that prevents them from working properly is considered as lung disease. Therefore, early diagnosis and prediction of lung diseases is an important part in improving the patient's life condition and survival. This study aims to assess the acceptable level of accuracy in the medical field by applying deep learning to a computed tomography (CT) data set and developing a platform (web application) that will allow users to submit images and receive the predictions. The result will be either cancerous, covid 19, pneumonia or normal when the patient's lung has none of these diseases. Any supervised project begins with data collection, especially data that will be used as a training dataset. In our case, we will have to acquire CT scans of the lungs from personal source, this medical imaging technique is used in radiology to provide images of the architecture and physiological processes of the body. Setting up this project will consist of going through the following steps: We will use a dataset containing CT scan images of the lungs, then tf dataset for data cleaning and preprocessing, ImageDataGenerator for real-time data augmentation, Convolutional Neural Network (CNN) for model building, then we will also build transfer learning models with VGG-16 and Inception V3 architectures. Once our model is saved, we iv will use FastAPI and TF serving to serve it, then React Js will be used to develop the frontend of our app where users will submit images and get the prediction result displayed on the screen. The next step of this project will be to deploy our model to a mobile application using Google Cloud Platform (GCP) and React Native, which is a hybrid framework for developing mobile applications