Frontiers in Digital Health, Journal Year: 2025, Volume and Issue: 7
Published: Jan. 21, 2025
Radiology departments are under increasing pressure to meet the demand for timely and accurate diagnostics, especially with chest x-rays, a key modality pulmonary condition assessment. Producing comprehensive radiological reports is time-consuming process prone errors, particularly in high-volume clinical environments. Automated report generation plays crucial role alleviating radiologists' workload, improving diagnostic accuracy, ensuring consistency. This paper introduces ChestX-Transcribe , multimodal transformer model that combines Swin Transformer extracting high-resolution visual features DistilGPT generating clinically relevant, semantically rich medical reports. Trained on Indiana University Chest x-ray dataset, demonstrates state-of-the-art performance across BLEU, ROUGE, METEOR metrics, outperforming prior models producing meaningful However, reliance dataset potential limitations, including selection bias, as collected from specific hospitals within Network Patient Care. may result underrepresentation of certain demographics or conditions not prevalent those healthcare settings, potentially skewing predictions when applied more diverse populations different Additionally, ethical implications handling sensitive data, patient privacy data security, considered. Despite these challenges, shows promising enhancing real-world radiology workflows by automating creation reports, reducing efficiency. The findings highlight transformative transformers healthcare, future work focusing generalizability optimizing integration.
Language: Английский