A system for automatically extracting clinical events with temporal information

Abstract

Background The popularization of health and medical informatics yields huge amounts of data. Extracting clinical events on a temporal course is the foundation of enabling advanced applications and research. It is a structure of presenting information in chronological order. Manual extraction would be extremely challenging due to the quantity and complexity of the records. Methods We present an recurrent neural network- based architecture, which is able to automatically extract clinical event expressions along with each event’s temporal information. The system is built upon the attention-based and recursive neural networks and introduce a piecewise representation (we divide the input sentences into three pieces to better utilize the information in the sentences), incorporates semantic information by utilizing word representations obtained from BioASQ and Wikipedia. Results The system is evaluated on the THYME corpus, a set of manually annotated clinical records from Mayo Clinic. In order to further verify the effectiveness of the system, the system is also evaluated on the TimeBank _Dense corpus. The experiments demonstrate that the system outperforms the current state-of-the-art models. The system also supports domain adaptation, i.e., the system may be used in brain cancer data while its model is trained in colon cancer data. Conclusion Our system extracts temporal expressions, event expressions and link them according to actually occurring sequence, which may structure the key information from complicated unstructured clinical records. Furthermore, we demonstrate that combining the piecewise representation method with attention mechanism can capture more complete features. The system is flexible and can be extended to handle other document types.

Validated algorithms for identifying timing of second event of oropharyngeal squamous cell carcinoma using real‐world data

“…While the current body of literature on developing algorithms to establish second event of cancer in the head and neck population is sparse, lack of comparability extends beyond differences in location of the primary tumor. Studies that focus on algorithms that establish timing of second event of cancer also differ from our research in terms of method, population, region, data, and tumor type 7–28 . While our research is a novel and robust method for identifying second event OPSCC, it can also act as a useful complement to pre‐existing research.…”

“…While our research is a novel and robust method for identifying second event OPSCC, it can also act as a useful complement to pre‐existing research. Algorithms to identify timing of OPSCC second events are less common compared to algorithms designed to do the same for other tumor groups; however, the methods established by this research can be extrapolated to further contribute to other tumor groups 7–28 . In addition, the availability of a robust and accurate algorithm for identifying the timing of OPSCC second event can be used to answer various research questions important to progressing OPSCC detection and treatment.…”

“…While a number of studies explored the development of algorithms to identify second events of cancer, 6–28 very few explored developing algorithms to establish temporality of second events in the head and neck cancer population. Previous work by Xu et al developed and validated algorithms to determine second events of OPSCC 29 .…”

Temporal Relation Extraction in Clinical Texts

“…Among publications that addressed only within-sentence relations, the authors of [89] and [90] used tree-based Bi-LSTM-RNNs, the authors of [99] used BI-LSTM with self-training, the authors of [87] used CNNs, the authors of [101] used a hybrid approach based on a CNN and an SVM model, the authors of [104] used GRUs and attention, and the authors of [77] used RNNs, attention, and piecewise representation. Based on these results, we highlight [89] and [99].…”

“…We highlight some approaches for cross-sentence relations. The authors of [77] used a classifier for event-event and another for event-time. The authors of [11] and [72] used a classifier to detect co-references, but the authors of [11] used an additional classifier to detect the main events.…”