딥러닝 기반 의료 영상 판독 소견서 자동 생성 모델 비교 분석
Comparative Study of Deep Learning-based Automatic Generation of Radiology Report
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
딥러닝 기반 의료 영상 판독 소견서 생성 연구는 최근 몇 년간 의료 영상 분석 분야에서 중요한 발전을 이루었으며, 주로 비전 모델과 언어 모델의 융합에 초점을 맞춰왔다. 이러한 연구는 의료 전문가들의 업무 효율성을 높이고, 진단의 일관성을 향상시키며 오류를 줄이는 데 기여할 것으로 기대된다. 본 논문에서는 자연어 처리 기술과 의료 영상 분석 기술의 융합을 통해 흉부 방사선 영상에 대한 판독 소견서를 자동으로 생성하는 최근 연구들을 소개하고, 이를 위한 주요 딥러닝 모델과 공용 데이터셋인 MIMIC-CXR 및 IU-Xray를 활용하여 성능 비교를 수행하였다. 성능 분석 결과, MIMIC-CXR 데이터셋에서는 RGRG 모델이 BLEU-4에서 0.126의 점수로 우수한 성능을 보였으며, IU-Xray 데이터셋에서는 COMG 모델이 BLEU-4에서 0.206을 기록하였다. 연구 결과, 각 모델이 특정 지표에서 강점을 보이는 반면, 데이터 불균형 및 개인정보 보호 문제와 같은 한계가 존재함을 확인하였다. 이를 바탕으로 향후 연구의 발전 방향을 제시하며, 이러한 기술이 임상 현장에서 실질적으로 적용될 수 있는 가능성을 논의한다.
Abstract
Research on deep learning-based automatic generation of radiology reports has seen significant advancements in recent years, with a primary focus on the integration of vision and language models. These advancements are expected to enhance the efficiency of medical professionals, improve diagnostic consistency, and reduce errors. In this paper, we introduce recent studies that combine natural language processing and medical image analysis techniques to automatically generate radiology reports for chest X-rays. Using key deep learning models and public datasets, including MIMIC-CXR and IU-Xray, we conducted a comparative performance evaluation. The analysis shows that on the MIMIC-CXR dataset, the RGRG model achieved superior performance with a BLEU-4 score of 0.126, while on the IU-Xray dataset, the COMG model recorded a BLEU-4 score of 0.206. Our findings reveal that while each model excels in specific evaluation metrics, limitations such as data imbalance and privacy concerns persist. Based on these findings, we propose future research directions and discuss the potential for these technologies to be practically applied in clinical settings.
참고문헌 (21)
- Monshi MMA, "Deep learning in generating radiology reports : A survey", Artificial Intelligence Medicine, 106 (2020)[DOI]
- Johnson AEW, "MIMIC-CXR, a de-identified publicly available database of chest radiographs with free-text reports", Scientific Data, 6(317) (2019)[DOI]
- Demner-Fushman D, "Preparing a collection of radiology examinations for distribution and retrieval", Journal of American Medical Information Association, 23(2), pp.304~310 (2016)[DOI]
- Li M, "Dynamic Graph Enhanced Contrastive Learning for Chest X-Ray Report Generation", 2023 IEEE, 2023, 3334~3343
- Devlin J, "BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding", 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp.4171~4186 (2019)
- Tanida T, "Interactive and Explainable Region-guided Radiology Report Generation", 2023 IEEE, 2023, 7433~7442
- Radford A, "Language models are unsupervised multitask learners", OpenAI blog, 1(8), pp.9~ (2019)
- Gu T, "Complex Organ Mask Guided Radiology Report Generation", 2024 IEEE, 2024, 7980~7989
- He K, "Deep Residual Learning for Image Recognition", 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.770~778 (2016)
- Wu S, "MAKEN: Improving Medical Report Generation with Adapter Tuning and Knowledge Enhancement in Vision-Language Foundation Models", 2024 IEEE International Symposium on Biomedical Imaging (ISBI), pp.1~5 (2024)
- Li J, "Blip-2 : Bootstrapping language-image pre-training with frozen image encoders and large language models", International Conference on Machine Learning (ICML), pp.19730~19742 (2023)
- Huang Z, "KiUT : Knowledgeinjected U-Transformer for Radiology Report Generation", 2023 IEEE, 2023, 19809~19818
- Ronneberger O, "U-net:Convolutional networks for biomedical image segmentation", International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), pp.234~241 (2015)
- Wang Z, "METransformer:Radiology Report Generation by Transformer with Multiple Learnable Expert Tokens", 2023 IEEE, 2023, 11558~11567
- Vedantam R, "CIDEr : Consensus-based image description evaluation", 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.4566~4575 (2015)
- Papineni K, "BLEU : a method for automatic evaluation of machine translation", the 40th annual meeting of the Association for Computational Linguistics, pp.311~318 (2002)
- Satanjeev B, "METEOR: An automatic metric for MT evaluation with improved correlation with human judgments", ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and, 2005, 65~72
- Lin CW, "ROUGE: A package for automatic evaluation of summaries", Text Summarization Branches Out, pp.74~81 (2004)