Awards
IMO Health algorithms are top-ranked for various clinical NLP challenges involving clinical documentation, biomedical literature, FDA drug labels, and patent documentation.
| International NLP challenges | Type | Rank | |
|---|---|---|---|
| 2020 | Chemical name detection from patent documents | Patent | #1 |
| 2018 | n2c2 medication and associated ADR | Clinical text | #1 |
| 2017 | TAC ADR extraction from drug labels | Drug label | #1 |
| 2016 | SemEval, temporal information extraction | Clinical text | #1 |
| 2016 | CEGS N-GRID, de-identification | Literature | #2 |
| 2015 | BioCREATIVE chemical-induced disease from literature | Clinical text | #1 |
| 2015 | SemEval disease-modifier extraction | Clinical text | #1 |
| 2014 | SemEval disease-modifier extraction | Clinical text | #1 |
| 2013 | SHARe/CLEF abbreciation recognition | Clinical text | #1 |
| 2012 | i2b2 temporal information extraction | Clinical text | #1 |
| 2010 | i2b2 problem, treatment, test extraction | Clinical text | #2 |
| 2009 | i2b2 medication information extraction | Clinical text | #2 |
Awards spotlight
Our team won 1st and 2nd place at the 2018 National NLP Clinical Challenges, with F1 scores of 93.45% for NER, 96.30% for RC, and 89.05% for end-to-end evaluation. Our approach to extracting medications and associated adverse drug events from clinical documents outperformed traditional machine learning algorithms, and our joint model for ADE recognition and relation extraction showed great promise. The challenge aimed to answer whether NLP systems can identify patients who meet selection criteria for clinical trials using narrative medical records.
Abstract: A study of deep learning approaches for medication and adverse drug event extraction from clinical text
TAC 2017 was a significant milestone in NLP and related applications. Our system ranked 1st place across all four ADR sub-tasks, showing that it is feasible to extract adverse drug reactions from drug labels using machine-learning methods with high performance.
Abstract: System for Adverse Drug Reaction Extraction from Drug Labels at TAC-ADR 2017
Our team at Melax Tech developed a top-performing end-to-end system for the medical domain, which achieved 1st place in the 2016 Clinical TempEval challenge. Using state-of-the-art techniques for entity recognition and temporal relation identification, the system excelled in all six sub-tasks and the TLINK: Contains identification task. Notably, it achieved the best performance in narrative container relation identification with gold standard annotations.
Abstract: UTHealth at SemEval-2016 Task 12: an End-to-End System for Temporal Information Extraction from Clinical Notes
