Active deep learning for the identification of concepts and relations in electroencephalography reports
Copyright © 2019 Elsevier Inc. All rights reserved..
The identification of medical concepts, their attributes and the relations between concepts in a large corpus of Electroencephalography (EEG) reports is a crucial step in the development of an EEG-specific patient cohort retrieval system. However, the recognition of multiple types of medical concepts, along with the many attributes characterizing them is challenging, and so is the recognition of the possible relations between them, especially when desiring to make use of active learning. To address these challenges, in this paper we present the Self-Attention Concept, Attribute and Relation (SACAR) identifier, which relies on a powerful encoding mechanism based on the recently introduced Transformer neural architecture (Dehghani et al., 2018). The SACAR identifier enabled us to consider a recently introduced framework for active learning which uses deep imitation learning for its selection policy. Our experimental results show that SACAR was able to identify medical concepts more precisely and exhibited enhanced recall, compared with previous methods. Moreover, SACAR achieves superior performance in attribute classification for attribute categories of interest, while identifying the relations between concepts with performance competitive with our previous techniques. As a multi-task network, SACAR achieves this performance on the three prediction tasks simultaneously, with a single, complex neural network. The learning curves obtained in the active learning process when using the novel Active Learning Policy Neural Network (ALPNN) show a significant increase in performance as the active learning progresses. These promising results enable the extraction of clinical knowledge available in a large collection of EEG reports.
| Medienart: |
E-Artikel |
|---|
| Erscheinungsjahr: |
2019 |
|---|---|
| Erschienen: |
2019 |
| Enthalten in: |
Zur Gesamtaufnahme - volume:98 |
|---|---|
| Enthalten in: |
Journal of biomedical informatics - 98(2019) vom: 01. Okt., Seite 103265 |
| Sprache: |
Englisch |
|---|
| Beteiligte Personen: |
Maldonado, Ramon [VerfasserIn] |
|---|
| Links: |
|---|
| Anmerkungen: |
Date Completed 14.10.2020 Date Revised 14.10.2020 published: Print-Electronic Citation Status MEDLINE |
|---|
| doi: |
10.1016/j.jbi.2019.103265 |
|---|
| funding: |
|
|---|---|
| Förderinstitution / Projekttitel: |
|
| PPN (Katalog-ID): |
NLM300794061 |
|---|
| LEADER | 01000naa a22002652 4500 | ||
|---|---|---|---|
| 001 | NLM300794061 | ||
| 003 | DE-627 | ||
| 005 | 20231225103305.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 231225s2019 xx |||||o 00| ||eng c | ||
| 024 | 7 | |a 10.1016/j.jbi.2019.103265 |2 doi | |
| 028 | 5 | 2 | |a pubmed24n1002.xml |
| 035 | |a (DE-627)NLM300794061 | ||
| 035 | |a (NLM)31470094 | ||
| 035 | |a (PII)S1532-0464(19)30184-4 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 100 | 1 | |a Maldonado, Ramon |e verfasserin |4 aut | |
| 245 | 1 | 0 | |a Active deep learning for the identification of concepts and relations in electroencephalography reports |
| 264 | 1 | |c 2019 | |
| 336 | |a Text |b txt |2 rdacontent | ||
| 337 | |a ƒaComputermedien |b c |2 rdamedia | ||
| 338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
| 500 | |a Date Completed 14.10.2020 | ||
| 500 | |a Date Revised 14.10.2020 | ||
| 500 | |a published: Print-Electronic | ||
| 500 | |a Citation Status MEDLINE | ||
| 520 | |a Copyright © 2019 Elsevier Inc. All rights reserved. | ||
| 520 | |a The identification of medical concepts, their attributes and the relations between concepts in a large corpus of Electroencephalography (EEG) reports is a crucial step in the development of an EEG-specific patient cohort retrieval system. However, the recognition of multiple types of medical concepts, along with the many attributes characterizing them is challenging, and so is the recognition of the possible relations between them, especially when desiring to make use of active learning. To address these challenges, in this paper we present the Self-Attention Concept, Attribute and Relation (SACAR) identifier, which relies on a powerful encoding mechanism based on the recently introduced Transformer neural architecture (Dehghani et al., 2018). The SACAR identifier enabled us to consider a recently introduced framework for active learning which uses deep imitation learning for its selection policy. Our experimental results show that SACAR was able to identify medical concepts more precisely and exhibited enhanced recall, compared with previous methods. Moreover, SACAR achieves superior performance in attribute classification for attribute categories of interest, while identifying the relations between concepts with performance competitive with our previous techniques. As a multi-task network, SACAR achieves this performance on the three prediction tasks simultaneously, with a single, complex neural network. The learning curves obtained in the active learning process when using the novel Active Learning Policy Neural Network (ALPNN) show a significant increase in performance as the active learning progresses. These promising results enable the extraction of clinical knowledge available in a large collection of EEG reports | ||
| 650 | 4 | |a Journal Article | |
| 650 | 4 | |a Research Support, N.I.H., Extramural | |
| 650 | 4 | |a Research Support, Non-U.S. Gov't | |
| 650 | 4 | |a Active learning | |
| 650 | 4 | |a Attribute classification | |
| 650 | 4 | |a Concept detection | |
| 650 | 4 | |a Deep learning | |
| 650 | 4 | |a Electroencephalography | |
| 650 | 4 | |a Long-distance relation identification | |
| 700 | 1 | |a Harabagiu, Sanda M |e verfasserin |4 aut | |
| 773 | 0 | 8 | |i Enthalten in |t Journal of biomedical informatics |d 2001 |g 98(2019) vom: 01. Okt., Seite 103265 |w (DE-627)NLM112821766 |x 1532-0480 |7 nnns |
| 773 | 1 | 8 | |g volume:98 |g year:2019 |g day:01 |g month:10 |g pages:103265 |
| 856 | 4 | 0 | |u http://dx.doi.org/10.1016/j.jbi.2019.103265 |3 Volltext |
| 912 | |a GBV_USEFLAG_A | ||
| 912 | |a GBV_NLM | ||
| 951 | |a AR | ||
| 952 | |d 98 |j 2019 |b 01 |c 10 |h 103265 | ||