Medical Report Generation and Chatbot for COVID_19 Diagnosis Using Open-AI

Abstract

Abstract The novel corona_virus (COVID_19) is an infectious disease have catastrophic impact on health and spread across the world. A crucial step in COVID-19 detection is to develop an automated and efficient classification system so that prompt treatment and medical care can be provided to the patients. However, most of the medical imaging systems just present the conditions of lung and scans are generated in large quantities that add a huge burdens to the workload of radiologists. Consequently, an intelligent system having capacity of lesions analysis in images and automatically creating a medical reports is of great significance for diagnosis of COVID_19. In this paper, we propose to use the fine tuned GPT3 and OPT350m models to automatically generate the medical text reports based on the segmented lesion regions of COVID_19 CT scan of patients. The proposed approach also provides the GPT3 based chat bot for the users to ask questions regarding COVID_19 identification. The custom trained chat bot responds to the user or practitioner queries based on the generated report by the fine tuned GPT3 and OPT model(350m). The experimental results showed that proposed models achieved beyond the state-of-the-art performances on medical report generation using COVID_19 CT scan data set. We conclude our research study by enumerating few future research directions in COVID_19 report generation.