Leveraging GPT for Automated Radiology Reporting in Multimodal Medical Imaging

Automated radiology reporting represents a pivotal advancement in medical imaging diagnos- tics, offering a solution to the time-consuming process of manual report generation. In this study, we propose an innovative methodology leveraging OpenAI’s Generative Pre-trained Transformer (GPT) models for the automated generation of descriptive radiology reports in multimodal medical imaging. Our approach integrates state-of-the-art natural language processing techniques with im- age processing methodologies to synthesize comprehensive reports from diverse image modalities, including computed tomography (CT) scans and ultrasounds. Through rigorous experimenta- tion on a curated dataset of radiology images, encompassing various pathological conditions and anatomical structures, we demonstrate the efficacy of our methodology in producing clinically relevant and coherent reports. Our findings underscore the potential of GPT-powered systems in augmenting radiologists’ workflows and improving diagnostic efficiency in medical imaging in- terpretation. Additionally, we emphasize that the dataset used in this research is obtained from Kaggle, a renowned platform for sharing and accessing diverse datasets, ensuring the accessibility and reproducibility of our experiments.