@article{article_1571384, title={Extraction Of Clinical Entities from Chest Radiology Reports Using NLP Methods}, journal={Journal of Materials and Mechatronics: A}, volume={6}, pages={1–14}, year={2025}, DOI={10.55546/jmm.1571384}, author={Ergün, Uçman and Orcin, Sedanur and Barın, Sezin}, keywords={Derin Öğrenme, Doğal Dil İşleme, Adlandırılmış varlık tanıma, Radyolojik Rapor, BERT}, abstract={Radiology reports are essential for clinical decision-making and diagnosis, containing complex and detailed information. However, their unstructured nature makes efficient processing and analysis challenging, increasing the workload of healthcare professionals and slowing down clinical workflows. Natural Language Processing (NLP) techniques provide effective solutions by extracting meaningful information from such texts, reducing expert workload, and expediting decision-making processes. This study focuses on Named Entity Recognition (NER) in chest radiology reports using the RadGraph dataset, annotated with four tag types. The objective is to compare the performance of two NLP models—BERT (Bidirectional Encoder Representations from Transformers) and LSTM (Long Short-Term Memory) —to identify the most suitable approach for clinical data. Various training parameters, including learning rate, optimization algorithm, and input size, were optimized to enhance model performance. To address the class imbalance in the dataset, data augmentation techniques were applied, and both models were fine-tuned. The results revealed that BERT, leveraging its attention mechanism, demonstrated superior performance in identifying complex terms and entities, outperforming LSTM in accuracy, precision, recall, and F1 score. While LSTM effectively captured long-term dependencies, it required longer training times. This research highlights the potential of NLP in automating the extraction of clinical entities from radiology reports. It provides valuable insights for optimizing models and developing clinical decision support systems, ultimately aiming to enhance the efficiency of healthcare workflows.}, number={1}, publisher={Yusuf KAYALI}, organization={TÜBİTAK}