AI-Assisted Summarization of Radiologic Reports: Evaluating GPT3davinci, BARTcnn, LongT5booksum, LEDbooksum, LEDlegal, and LEDclinical.

BACKGROUND AND PURPOSE: The review of clinical reports is an essential part of monitoring disease progression. Synthesizing multiple imaging reports is also important for clinical decisions. It is critical to aggregate information quickly and accurately. Machine learning natural language processing (NLP) models hold promise to address an unmet need for report summarization. MATERIALS AND METHODS: We evaluated NLP methods to summarize longitudinal aneurysm reports. A total of 137 clinical reports and 100 PubMed case reports were used in this study. Models were 1) compared against expert-generated summary using longitudinal imaging notes collected in our institute and 2) compared using publicly accessible PubMed case reports. Five AI models were used to summarize the clinical reports, and a sixth model, the online GPT3davinci NLP large language model (LLM), was added for the summarization of PubMed case reports. We assessed the summary quality through comparison with expert summaries using quantitative metrics and quality reviews by experts. RESULTS: In clinical summarization, BARTcnn had the best performance (BERTscore = 0.8371), followed by LongT5Booksum and LEDlegal. In the analysis using PubMed case reports, GPT3davinci demonstrated the best performance, followed by models BARTcnn and then LEDbooksum (BERTscore = 0.894, 0.872, and 0.867, respectively). CONCLUSIONS: AI NLP summarization models demonstrated great potential in summarizing longitudinal aneurysm reports, though none yet reached the level of quality for clinical usage. We found the online GPT LLM outperformed the others; however, the BARTcnn model is potentially more useful because it can be implemented on-site. Future work to improve summarization, address other types of neuroimaging reports, and develop structured reports may allow NLP models to ease clinical workflow.

Building deep learning models for evidence classification from the open access biomedical literature.

We investigate the application of deep learning to biocuration tasks that involve classification of text associated with biomedical evidence in primary research articles. We developed a large-scale corpus of molecular papers derived from PubMed and PubMed Central open access records and used it to train deep learning word embeddings under the GloVe, FastText and ELMo algorithms. We applied those models to a distant supervised method classification task based on text from figure captions or fragments surrounding references to figures in the main text using a variety or models and parameterizations. We then developed document classification (triage) methods for molecular interaction papers by using deep learning mechanisms of attention to aggregate classification-based decisions over selected paragraphs in the document. We were able to obtain triage performance with an accuracy of 0.82 using a combined convolutional neural network, bi-directional long short-term memory architecture augmented by attention to produce a single decision for triage. In this work, we hope to encourage biocuration systems developers to apply deep learning methods to their specialized tasks by repurposing large-scale word embedding to apply to their data.