Honorable Mention

AI powered Mobile app and Web app to identify COVID-19 from Chest X-Ray

Solution Information:
Title : AI powered Mobile app and Web app to identify COVID-19 from Chest X-Ray
Description : We have developed a Mobile phone App and a Web app that can be used to detect COVID-19 induced Pneumonia X-Ray from other chest X-Ray. Researchers identified (Jianpeng Zhang et al) that COVID-19 induced Pneumonia has a unique Ground Glass Opacity pattern in lung that can be distinguishable from other lung diseases. This gives us the opportunity to uniquely identify COVID-19 X-Ray from other X-Ray images. However, given the nature of the virus as a novel one, radiologists are not trained to detect the virus infection from the X-Ray alone. Moreover, the difference of COVID-19 and ordinary Pneumonia induced lung infection are subtle in X-Ray image and hard to differentiate from X-Ray alone without the patient's other symptoms and case history by a radiologist. Here, AI comes into play the role of an expert assistant. It is much faster and efficient to train a machine over thousands of labelled training data to observe and detect subtle differences between various X-Ray images to train it’s Artificial Neural Network and classify them quickly which is otherwise not possible by a human eye. A Radiologist can use the app to primarily identify the X-Ray in question and combine it with his/her medical expertise along with patient's symptom/case history to determine if further tests like PCR/Antibody are needed or not. Saving precious resources of PCR and antibody test kits. The web app is performing 96% correct diagnosis with validation test data in a controlled setting. However, a rigorus test is needed to measure its actual performance with field data and proper legislative approval from Government Health agencies to implement it. As for the Mobile App, it is performing 85%-90% depending on Mobile hardware. We are limited by authentic sample data to train our Neural Network and a small margin of error can incur big difference when it comes to mobile inference.Quality of mobile device camera and processor plays a big part too. Good news is,The Mobile App does not require an internet connection for inference and can do real time on device inference and is intended to be used in a rural setting. Where PCR/Antibody kits are not readily available but a X-Ray machine is. It is assumed that all Upozela health complex has a working X-Ray machine in their disposal. This App could be a primary detection tool for Coronavirus for health workers. Along with detecting COVID-19 X-Ray, it is capable to detect common Pneumonia induced lung X-ray. It also has the capability to classify COVID-19 and common Pneumonia CT scans in a limited scope. It is an experimental project and as COVID-19 x-Ray data is in scarcity, we had limited training data in hand while building the App and can not guarantee to deliver a reliable result that you can solely depend on. It is not designed to replace a Radiologist rather to act as a decision making aid by a qualified Radiologist. It can only act as a Decision Support Expert System(DSES).
Category : Machine Learning
Media Links:
Video Link : https://youtu.be/e2AYONWcnWQ