RGen: Improving Multi-modal Deep Learning for Automated Radiology Report Generation

Abstract

"Radiology reporting has been a burden due to the extensive use of medical imaging in the clinical field. Thus, the different levels of expertise have led to instances where erroneous reports have gone unnoticed, resulting in profound inferences about patient well-being. Manual approaches have been proven to be laborious and expensive. With the advancements in medical imaging, radiology reporting has also attracted researchers’ interest in the development of novel approaches. Even though automation in radiology has made progress, most of the current research has focused only on conventional approaches for report generation. As a result, the author proposes a novel solution enhancing language fluency and clinical accuracy to address the gap and generate diverse radiology reports in order to assist clinicians in decision-making based on the limitations identified through the literature. This study contributes to the discussion of the approach, design, implementation, and evaluation phases throughout the research."