Abstract
Free-form radiology reports associated with coronary computed tomography angiography (CCTA) include nuanced and complicated linguistics to report cardiovascular disease. Standardization and interpretation of such reports is crucial for clinical use of CCTA. Coronary Artery Disease Reporting and Data System (CAD-RADS) has been proposed to achieve such standardization by implementing a strict template-based report writing and assignment of a score between 0 and 5 indicating the severity of coronary artery lesions. Even after its introduction, free-form unstructured report writing remains popular among radiologists. In this work, we present our attempts at bridging the gap between structured and unstructured reporting by natural language processing. We present machine learning models that while being trained only on structured reports, can predict CAD-RADS scores by analysis of free-text of unstructured radiology reports. The best model achieves 98% accuracy on structured reports and 92% 1-margin accuracy (difference of \(\le\)1 in the predicted and the actual scores) for free-form unstructured reports. Our model also performs well under very difficult circumstances including nuanced and widely varying terminology used for reporting cardiovascular functions and diseases, scarcity of labeled data for training our model, and uneven class label distribution.
- [1] . 2019. The validity and applicability of CAD-RADS in the management of patients with coronary artery disease. Insights Imag. 10, 117 (2019).Google Scholar
- [2] . 2018. Inter-observer agreement of the Coronary Artery Disease Reporting and Data System (CAD-RADS TM) in patients with stable chest pain. Polish J. Radiol. (2018), 151–159. Google Scholar
- [3] . 2018. Inter observer agreement of the coronary artery disease reporting and data system (CAD-RADS TM) in patients with stable chest pain. Polish J. Radiol. 83, (2018), 151–159.Google Scholar
- [4] . 2017. Towards better understanding of gradient-based attribution methods for deep neural networks. arXiv:1711.06104 (2017).Google Scholar
- [5] . 2019. Comparative effectiveness of convolutional neural network (CNN) and recurrent neural network (RNN) architectures for radiology text report classification. Artif. Intell. Med. 97 (2019), 79–88.Google Scholar
Digital Library
- [6] . 2017. Intelligent word embeddings of free-text radiology reports. In AMIA Annual Symposium Proceedings, Vol. 2017. American Medical Informatics Association.Google Scholar
- [7] . 2019. Natural language processing of radiology reports in patients with hepatocellular carcinoma to predict radiology resource utilization. J. Amer. Coll. Radiol. 16, 6 (2019), 840–844.Google Scholar
- [8] . 2019. Use of machine learning to identify follow-up recommendations in radiology reports. J. Amer. Coll. Radiol. 16, 3 (2019), 336–343.Google Scholar
- [9] . 2011. Document-level classification of CT pulmonary angiography reports based on an extension of the ConText algorithm. J. Biomed. Inform. 44, 5 (2011), 728–737. Google Scholar
Digital Library
- [10] . 2009. Reproducibility of coronary artery plaque volume and composition quantification by 64-detector row coronary computed tomographic angiography: An intraobserver, interobserver, and interscan variability study. J. Cardiovasc. Comput. Tomog. 3, 5 (2009), 312–320.
DOI: https://doi.org/10.1016/j.jcct.2009.07.001Google Scholar - [11] . 2014. Guidelines on the management of stable coronary artery disease. Acta Cardiol. 69, 1 (2014), 51–52.
DOI: https://doi.org/10.2143/AC.69.1.3011345Google Scholar - [12] . 2016. Coronary Artery Disease—Reporting and Data System (CAD-RADS). JACC-Cardiovasc. Imag. 9, 9 (2016), 1099–1113.
DOI: https://doi.org/10.1016/j.jcmg.2016.05.005Google Scholar - [13] . 2016. CAD-RADSTM coronary artery disease—Reporting and data system.J. Cardiovasc. Comput. Tomog. 10, 4 (2016), 269–281.Google Scholar
- [14] . 2018. BERT: Pre-training of deep bidirectional transformers for language understanding. arXiv:1810.04805 (2018).Google Scholar
- [15] . 2005. Application of recently developed computer algorithm for automatic classification of unstructured radiology reports: Validation study. Radiology 234, 2 (2005), 323–329.Google Scholar
Cross Ref
- [16] . 2020. Attention-guided classification of abnormalities in semi-structured computed tomography reports. In Medical Imaging 2020: Computer-Aided Diagnosis, Vol. 11314. International Society for Optics and Photonics, 113141P.Google Scholar
- [17] . 2011. SCCT guidelines on radiation dose and dose-optimization strategies in cardiovascular CT. J. Cardiovasc. Comput. Tomog. 5 (2011), 198–224.
DOI: https://doi.org/10.1016/j.jcct.2011.06.001Google Scholar - [18] . 2016. Predicting high imaging utilization based on initial radiology reports: A feasibility study of machine learning. Acad. Radiol. 23, 1 (2016), 84–89.Google Scholar
Cross Ref
- [19] . 2017. Performance of a machine learning classifier of knee MRI reports in two large academic radiology practices: A tool to estimate diagnostic yield. Amer. J. Roentgen. 208, 4 (2017), 750–753.Google Scholar
- [20] . 2002. Use of natural language processing to translate clinical information from a database of 889,921 chest radiographic reports. Radiology 224, 1 (2002), 157–163.Google Scholar
- [21] . 2017. Google’s multilingual neural machine translation system: Enabling zero-shot translation. Trans. Assoc. Comput. Ling. 5 (2017), 339–351.Google Scholar
Cross Ref
- [22] . 2014. Convolutional neural networks for sentence classification. In Proceedings of Empirical Methods in Natural Language Processing
(EMNLP’14) . 1746–1751.Google ScholarCross Ref
- [23] . 2014. Distributed representations of sentences and documents. In Proceedings of the International Conference on Machine Learning. 1188–1196. Google Scholar
Digital Library
- [24] . 2020. BioBERT: A pre-trained biomedical language representation model for biomedical text mining. Bioinformatics 36, 4 (2020), 1234–1240.Google Scholar
Cross Ref
- [25] . 2014. ScienceDirect SCCT guidelines for the interpretation and reporting of coronary CT angiography: A report of the Society of Cardiovascular Computed Tomography Guidelines Committee. J. Cardiovasc. Comput. Tomog. 8, 5 (2014), 342–358.
DOI: https://doi.org/10.1016/j.jcct.2014.07.003Google Scholar - [26] . 2018. Coronary artery disease reporting and data system (CAD-RADS TM): Inter-observer agreement for assessment categories and modifiers?J. Cardiovasc. Comput. Tomog. 12, 2 (2018), 125–130.
DOI: https://doi.org/10.1016/j.jcct.2017.11.014Google Scholar - [27] . 2018. Extraction of BI-RADS findings from breast ultrasound reports in Chinese using deep learning approaches. Int. J. Med. Inform. 119 (2018), 17–21.Google Scholar
- [28] . 2013. Efficient estimation of word representations in vector space. arXiv:1301.3781 (2013).Google Scholar
- [29] . 2013. Distributed representations of words and phrases and their compositionality. In Proceedings of the International Conference on Advances in Neural Information Processing Systems. 3111–3119. Google Scholar
Digital Library
- [30] . 2013. 2013 ESC guidelines on the management of stable coronary artery disease. Eur. Heart J. 34, 38 (2013), 2949–3003.
DOI: https://doi.org/10.1093/eurheartj/eht296.Google Scholar - [31] . 2017. LSTM recurrent neural networks for short text and sentiment classification. In Proceedings of the International Conference on Artificial Intelligence and Soft Computing. Springer, 553–562.Google Scholar
Cross Ref
- [32] . 2010. Interobserver variations of plaque severity score and segment stenosis score in coronary arteries using 64 slice multidetector computed tomography: A substudy of the ACCURACY trial. J. Cardiovasc. Comput. Tomog. 4, 5 (2010), 312–318.
DOI: https://doi.org/10.1016/j.jcct.2010.05.018Google Scholar - [33] . 2019. 2018 ESC/EACTS Guidelines on myocardial revascularization. Eur. J. Cardio-thorac. Surg. 55, 1 (2019), 4–90.
DOI: https://doi.org/10.1093/ejcts/ezy289Google Scholar - [34] . 2019. How to fine-tune bert for text classification? In Proceedings of the China National Conference on Chinese Computational Linguistics. Springer, 194–206.Google Scholar
Digital Library
- [35] . 2015. Document modeling with gated recurrent neural network for sentiment classification. In Proceedings of the Conference on Empirical Methods in Natural Language Processing. 1422–1432.Google Scholar
Cross Ref
- [36] . 2010. ACCF/SCCT/ACR/AHA/ASE/ASNC/NASCI/SCAI/SCMR 2010 appropriate use criteria for cardiac computed tomography. J. Amer. Coll. Cardiol. 56, 22 (2010), 1864–1894.
DOI: https://doi.org/10.1016/j.jacc.2010.07.005Google Scholar - [37] . 2016. Attention-based LSTM for aspect-level sentiment classification. In Proceedings of the Conference on Empirical Methods in Natural Language Processing. 606–615.Google Scholar
Cross Ref
- [38] . 2019. A clinical text classification paradigm using weak supervision and deep representation. BMC Medical Informatics Decis. Mak. 19, 1 (2019), 1.Google Scholar
- [39] . 2017. World Health Organization (2017) Key facts Cardiovascular Diseases (CVDs). Retrieved from https://www.who.int/cardiovascular_diseases/en/.Google Scholar
- [40] . 2015. Show, attend and tell: Neural image caption generation with visual attention. In Proceedings of the International Conference on Machine Learning. 2048–2057.Google Scholar
Digital Library
Index Terms
Bridging the Gap between Structured and Free-form Radiology Reporting: A Case-study on Coronary CT Angiography
Recommendations
Cardiac Components Categorization and Coronary Artery Enhancement in CT Angiography
CASH '14: Proceedings of the 2014 International Conference on Computer Assisted System in HealthCoronary artery analysis is an important phase of each cardiac image processing system. Once physicians decide to check up coronary arteries and also other cardiac components in CT Angiography, it should be done precisely to give the correct information ...
Automatic detection and quantification of coronary calcium on 3D CT angiography data
Cardiac calcium scoring is an important step for the diagnosis of coronary heart diseases. Therefore, non-contrast enhanced cardiac computed tomography has been established as the de facto standard method for clinical risk assessment and contrast ...






Comments