BERT-based Transfer Learning in Sentence-level Anatomic Classification of Free-Text Radiology Reports.

Autor:	Nishigaki D; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Suzuki Y; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Wataya T; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Kita K; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Yamagata K; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Sato J; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Kido S; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan., Tomiyama N; Departments of Artificial Intelligence Diagnostic Radiology (D.N., Y.S., T.W., K.K., K.Y., J.S., S.K.) and Radiology (N.T.), Osaka University Graduate School of Medicine, 2-2 Yamadaoka, Suita, Osaka 565-0871, Japan.
Jazyk:	angličtina
Zdroj:	Radiology. Artificial intelligence [Radiol Artif Intell] 2023 Feb 15; Vol. 5 (2), pp. e220097. Date of Electronic Publication: 2023 Feb 15 (Print Publication: 2023).
DOI:	10.1148/ryai.220097
Abstrakt:	Purpose: To assess whether transfer learning with a bidirectional encoder representations from transformers (BERT) model, pretrained on a clinical corpus, can perform sentence-level anatomic classification of free-text radiology reports, even for anatomic classes with few positive examples. Materials and Methods: This retrospective study included radiology reports of patients who underwent whole-body PET/CT imaging from December 2005 to December 2020. Each sentence in these reports (6272 sentences) was labeled by two annotators according to body part ("brain," "head & neck," "chest," "abdomen," "limbs," "spine," or "others"). The BERT-based transfer learning approach was compared with two baseline machine learning approaches: bidirectional long short-term memory (BiLSTM) and the count-based method. Area under the precision-recall curve (AUPRC) and area under the receiver operating characteristic curve (AUC) were computed for each approach, and AUCs were compared using the DeLong test. Results: The BERT-based approach achieved a macro-averaged AUPRC of 0.88 for classification, outperforming the baselines. AUC results for BERT were significantly higher than those of BiLSTM for all classes and those of the count-based method for the "brain," "chest," "abdomen," and "others" classes ( P values < .025). AUPRC results for BERT were superior to those of baselines even for classes with few labeled training data (brain: BERT, 0.95, BiLSTM, 0.11, count based, 0.41; limbs: BERT, 0.74, BiLSTM, 0.28, count based, 0.46; spine: BERT, 0.82, BiLSTM, 0.53, count based, 0.69). Conclusion: The BERT-based transfer learning approach outperformed the BiLSTM and count-based approaches in sentence-level anatomic classification of free-text radiology reports, even for anatomic classes with few labeled training data. Keywords: Anatomy, Comparative Studies, Technology Assessment, Transfer Learning Supplemental material is available for this article. © RSNA, 2023. Competing Interests: Disclosures of conflicts of interest: D.N. No relevant relationships. Y.S. No relevant relationships. T.W. No relevant relationships. K.K. No relevant relationships. K.Y. No relevant relationships. J.S. No relevant relationships. S.K. No relevant relationships. N.T. No relevant relationships. (© 2023 by the Radiological Society of North America, Inc.)
Databáze:	MEDLINE
Externí odkaz:	Zobrazit plný text záznamu