Annotation of Entities and Relations in Spanish Radiology Reports

Abstract

Radiology reports express the results of a radiology study and contain information about anatomical entities, findings, measures and impressions of the medical doctor. The use of information extraction techniques can help physicians to access this information in order to understand data and to infer further knowledge.Supervised machine learning methods are very popular to address information extraction, but are usually domain and language dependent. To train new classification models, annotated data is required. Moreover, annotated data is also required as an evaluation resource of information extraction algorithms. However, one major drawback of processing clinical data is the low availability of annotated datasets. For this reason we performed a manual annotation of radiology reports written in Spanish. This paper presents the corpus, the annotation schema, the annotation guidelines and further insight of the data.

Spanish Named Entity Recognition in the Biomedical Domain

“…Also, for example, ovarian cyst should be annotated as Our CRF results outperform others obtained with the same feature set for French [14] (the original proposal of the feature set) and for German [23]. 11 Since all results are tested with different genre of data and in different languages it is not easy to draw a conclusion about the differences in the results. In Spanish and in French anatomical entities have a higher F1 than findings.…”

“…Table 2 describes the composition of the dataset and Table 3 shows the number of AEs and FIs and the number of abbreviations and acronyms found in the annotated dataset. For details about the process followed, schema and elaborated guidelines to annotate the dataset refer to [11]. Table 3.…”

The Chilean Waiting List Corpus: a new resource for clinical Named Entity Recognition in Spanish

“…For negation, there are the works of Cruz Diaz et al (2017) using anamnesis and radiology reports, Marimon et al (2017) using clinical reports from a hospital in Barcelona, and Lima et al (2020) who released a biomedical corpus annotated with negation and uncertainty. From Spanish-speaking countries besides Spain, and to the best of our knowledge, the only published work is by Cotik et al (2017) in Argentina for the annotation of clinical findings, body parts, negation, temporal terms, and abbreviations in radiology reports. Some of the work done on biomedical texts is also noteworthy; Moreno-Sandoval and Campillos-Llanos (2013) annotated Part-of-Speech in biomedical documents written in Spanish, Japanese, and Arabic, Krallinger et al (2015) annotated PubMed abstracts in Spanish with chemicals and drugs.…”

NUBES: A Corpus of Negation and Uncertainty in Spanish Clinical Texts

“…In Spanish, seven corpora descriptions have been published for negation. Out of those seven, five are from the clinical domain and only two of them factor in uncertainty: i) the IxaMed-GS corpus (Oronoz et al, 2015) is a medical texts corpus annotated at an entity-level, that is, some events are characterised as being negated, speculated, or neither; ii) the UHU-HUVR corpus (Cruz Díaz et al, 2017) and iii) the IULA Spanish Clinical Record Corpus (IULA-SCRC) (Marimon et al, 2017) include negation cues and their scopes; iv) Campillos Llanos et al (2017) report to be working on extracting negation cue patterns from a corpus of emergency admission notes; finally, v) Cotik et al (2017) present a corpus of radiology reports annotated with events and relations, including negation and uncertainty. The IxaMed-GS and the corpus by Cotik et al (2017) are the only two that annotate uncertainty.…”