Skip to main content
. Author manuscript; available in PMC: 2020 Sep 1.
Published in final edited form as: Int J Med Inform. 2019 May 23;129:13–19. doi: 10.1016/j.ijmedinf.2019.05.018

Table 5:

Fine-grained NLP application accuracy for eligibility criteria information extraction

Rule-based version SVM-based version
Criterion Recall (%) Precision (%) Recall (%) Precision (%)
ECOG 0 96.9 96.9 100.0 100.0
ECOG 1* 85.7 85.7 85.7 100.0
ER+ 61.7 91.0 91.3 86.8
ER− 32.7 93.8 82.6 95.0
PR+ 80.0 96.8 96.0 86.6
PR− 57.5 90.2 92.1 86.8
HER2+ 24.0 100.0 85.0 89.5
HER− 31.0 98.0 90.5 74.5
Postmenopausal 72.1 88.0 86.89 84.1
Not Postmenopausal** - - 46.2 100.0
T0 100.0 100.0 100.0 100.0
T1 89.1 87.5 96.4 93.0
T2 79.5 81.5 97.6 98.8
T3* 82.4 90.3 88.2 100.0
N0 79.5 85.3 91.8 97.1
N1 84.8 78.5 99.0 100.0
N2 77.8 58.3 100.0 90.0
N3 100.0 75.0 0.00 0.00
M0 100.0 98.0 100.0 96.2
M1 2.3 50.0 90.5 78.6
*

No positive instances occurred for ECOG 2–5 or T4

**

The rule-based system did not extract non-postmenopausal mentions.