Terenz’s AI Engine detects COVID-19 from Chest X-Rays with an Accuracy of 98.14%
Recently COVID-19 created havoc all over the world by infecting more than 200000 people with 8000 deaths and as per the prediction of WHO professionals as well as healthcare scientists, it will increase at a higher rate in the coming days. COVID-19 is highly contagious and its asymptomatic nature makes it more difficult to detect at an early stage. As the disease is spreading at a faster rate, detecting the patients quickly can help us to contain it within a short period and save the lives of people.
In recent times, AI has played a very important role in the field of health care diagnosis. High-end computing resources combined with advanced machine learning techniques help the healthcare professional to understand the problem deeply and find useful patterns that can detect infectious diseases and at the same time helps in the diagnostic practices. Looking at the current scenario the way the COVID-19 is creating panic, everybody around the world looking forward to AI-based technology that can help in the screening of the patients and prevent the spread as well as helps in the diagnostic process by providing the details about the progression of the disease.
Till now few countries have developed testing kits that can detect COVID-19 within few minutes, however, most of the countries are unable to test more people because the testing method is not efficient enough and takes longer duration. So, to overcome the difficulty, Terenz proposed an AI-based screening system that can detect the COVID-19 using chest X rays within a few seconds with an accuracy of 98.14%. This novel system not only detects COVID-19 quickly but also provides the details about the progression of the disease so that the disease can be monitored in a more quantitative manner depending on the urgency.
The X-ray data used for the development of the AI system has been acquired from primarily two public databases namely, SIRM — Società Italiana di Radiologia Medica e Interventistica and Kaggle. Predominantly the data was acquired from three different categories i.e. Normal/ Healthy People, Pneumonia and COVID-19. The X-ray data considered in the study were from posteroanterior view where the X-ray beam enters through the posterior (back) aspect of the chest and exits out of the anterior (front) aspect, where the beam is detected. For detecting COVID-19 and classifying the data into Normal, Pneumonia and COVID-19, an AI engine leveraging Convolutional Neural Network was developed. The CNN algorithm developed performed pretty well by plotting an overall accuracy of 98.141%. Further, it was derived from the outcomes that the recall for the COVID-19 X-Rays was 100% with an AUC — ROC of 0.99.
Terenz is proud of the novel system because this system will tend to ease the process of screening and save many lives. It will also allow a huge number of people to be screened within a short time that could prevent the community from spreading and contain the disease quickly.
Terenz Corp is a South Korean Startup co-founded by Indians and Koreans, works on AI-based decision support system for critical healthcare diseases to improve the quality of life of the people by detecting the high-risk diseases at an early stage and monitoring them with the Terenz platform. Terenz was awarded as the best innovative startup at Bounce 2019, a global startup conference organized at Busan, South Korea. The objective of the AI-based decision support system is to support the doctors and to make the decisions quickly at high precision and accuracy. Terenz’s vision is to provide an affordable solution to a larger community to improve their quality of life from high-risk diseases.
The Engine developed in the work is presently very preliminary and can be used for the investigational purpose only. As the dataset is very small therefore, a shallow network has been developed to avoid any overfitting of the model. But the model is retrained every day with the new sets of data that are arriving on the internet and will be continued until the team of Data Scientists at Terenz Corp are satisfied with the performance of the model. Moreover, the initial pipeline of the model has been uploaded on the GitHub Repository of the company and the retrained model will be committed to the repository every day.
