ABSTRACT
Health literacy is the ability of a person to read and understand medical text and to use that information to make informed healthcare decisions. Unfortunately, medical articles are difficult to comprehend by common people as they use complex language and domain-specific terms. Improving health literacy is important for empowering communities against emerging threats and the COVID-19 pandemic bears testimony to this statement. One way to improve health literacy is easing access to complex healthcare information by summarising medical texts and simplifying them lexically by translating specific medical terminology to laymen's terms. In this paper we propose a system that performs extractive summarization on the medical article given as input followed by named entity recognition for identifying medical terms. The meanings of identified medical entities are then found through web scraping and displayed to the user along with the summary. We have experimented with state-of-the-art summarization models and Albert (A lite BERT) has provided the best ROUGE-1 score of 0.3789 and ROUGE-L of 0.2084. © 2022 IEEE.
ABSTRACT
The COVID-19 outbreak has been world-shattering. Since the day it was discovered, it has challenged the world to develop and invent new approaches and methods to fight against it. With this being said, scientists and researchers are relentlessly working to make the situation better and easier for everybody. In this paper, we have utilised transfer learning models to learn and extract important feature vectors from a CT scan image which can prove to be beneficial in the determination of COVID-19. However, due to the limitation in the amount of CT images available publicly, it was problematic to achieve a high performance deep learning model. To overcome this, we have built a balanced dataset consisting of a total of 11,209 CT scan images combined from three different sources which helped in achieving a diverse set of images. Moreover, we have also developed our own convolutional neural network (CNN) which achieved an accuracy of 97.92% in the prediction of Covid-19. Extensive experiments demonstrate the ability and potential of our proposed approaches in achieving high performance models. VGG16 achieved a significant accuracy of 98.7% which is the highest among all the transfer learning models. © 2021 IEEE