Skip to main content

Table 1 Overview of selected clinical information extraction system evaluations; see Section ‘Related work

From: Fine-grained information extraction from German transthoracic echocardiography reports

Article

Year

Domain

Language

Test set

Concepts

Prec.

Rec.

F1

[28]

2005

Echo

English

408 doc.

10

.99

.78

.87

[29]

2012

Echo

English

475 doc.

4 a

.95

.89

.92

[7]

2009

Mammography

Polish

705 doc.

66

.996

.995

.996

[7]

2009

Diabetes

Polish

100 doc.

68

.993

.965

.979

[5]

2010

General

English

160 doc.

many c

.801 b

.645 b

.715 b

[24]

2004

General

English

150 sent.

many c

.89

.77

- d

[40]

2009

Metastatic Tumor

English

101 doc.

many c

.73

.58

.65

[40]

2009

Primary Tumor

English

101 doc.

many c

.80

.84

.82

[40]

2009

Anatomical Site

English

101 doc.

many c

.97

.98

.97

[14]

2013

Radiology

German

40 doc.

2 e

.54

.74

.63

  1. Year: year of publication. Domain: intended domain or the domain used for evaluation. Test Set: size of test set used for evaluation, i.e., number of documents/sentences. Concepts: number of classes, concepts or terminology used for reported results. aconcept level analysis, see related work for details. bnamed entity recognition results used as an upper estimate; see original work for more detailed figures. capplication uses standardized resources such as UMLS or ICD-O with a large number of concepts. domitted to reflect that precision and recall have been evaluated on different sets of sentences. eSentence-level classification of normal vs. pathological findings