Automated Report Generation: A GRU Based Method for Chest X-Rays

Abstract

Radiology reports are the primary medium through which physicians communicate with patients and share diagnoses from medical scans. Examples include radiology reports for chest X-Rays and CT scans. Chest X-Ray images are frequently employed in clinical screening and diagnosis. However, writing medical reports for the X-Ray is tedious, error-prone, and time-consuming, even for experienced radiologists. The modern world of clinical practice demands that a radiologist with specialized training manually evaluate chest X-Ray and report the findings. Therefore, this paper explores the ability of artificial intelligence (AI) to automate diagnosing diseases through chest X-Rays and accurately generate radiology reports to alleviate the burdens of medical doctors. Automating this manual process could streamline a clinical workflow, and healthcare quality could be improved. The conventional AI-based abstract methods provide fluent but clinically incorrect radiology reports. The proposed Gated Recurrent Unit (GRU) based model provides both stan-dard language generation and clinical coherence. The model is evaluated on the Indiana University dataset with commonly-used metrics BLEU and ROUGE-L. Empirical evaluations illustrate that the proposed approach can make more precise diagnoses and generate more fluent and precise reports than existing baselines.