Skip to main content

Advertisement

Table 3 Evaluation of semantic annotation performance

From: Using data-driven sublanguage pattern mining to induce knowledge models: application in medical image reports knowledge representation

Semantic Categories True Positive (TP) True Negative (TN) False Positive (FP) False Negative (FN) Precision Recall F-Score
Abnormality 16 1660 4 12 80.0% 57.1% 0.6667
Body Part 238 1438 38 26 86.2% 90.2% 0.8815
Classification 12 1664 0 6 100.0% 66.7% 0.8000
Functional Concept 90 1586 14 12 86.5% 88.2% 0.8738
Location 230 1446 54 58 81.0% 79.9% 0.8042
Medical Activity 22 1654 14 26 61.1% 45.8% 0.5238
Medical Device and Object 8 1668 14 0 36.4% 100.0% 0.5333
Observation 168 1508 16 62 91.3% 73.0% 0.8116
Pathology 202 1474 4 50 98.1% 80.2% 0.8821
Physiology 16 1660 4 4 80.0% 80.0% 0.8000
Qualitative Concept 172 1504 22 60 88.7% 74.1% 0.8075
Quantitative Concept 78 1598 4 4 95.1% 95.1% 0.9512
Substance 24 1652 12 24 66.7% 50.0% 0.5714
Temporal Concept 56 1620 2 0 96.6% 100.0% 0.9825
Overall 1332 22,132 202 344 86.8% 79.5% 0.8299