Akademik Veri Yönetim Sistemi
2025 Innovations in Intelligent Systems and Applications Conference, ASYU 2025, Bursa, Türkiye, 10 - 12 Eylül 2025, (Tam Metin Bildiri)
The preparation of radiology reports from medical images is a time-consuming process and carries a risk of error, especially for inexperienced radiologists. In this study, a fusion process was performed to enable chest X-ray images to be reported more accurately and in greater detail using artificial intelligence supported models. In this context, reports generated by the high performance R2Gen and CvT2DistilGPT2 models were combined to obtain comprehensive reports. Necessary prompts were given to GPT-4 to improve the accuracy of these reports. The widely used IU X-Ray dataset was selected as the dataset. The reports were analyzed through expert evaluations and language generation metrics. Radiologist evaluations determined that the proposed model achieved a score of 2.65 out of 5, which is the closest performance to the reference score of 2.90. Additionally, the scores obtained with RadGraph_F1 and ChatGPT4o, which are important in terms of contextual and clinical accuracy, indicate that the proposed model is effective in improving the accuracy and coherence of the reports it generates.