Natural language processing of radiology reports for the detection of thromboembolic diseases and clinically relevant incidental findings

Table 4 Performances of the Naïve Bayes versus the maximum entropy classifiers on the CTA/CTV and incidentaloma test sets

		Precision		Recall		F-measure
Features		NB	ME	NB	ME	NB	ME
Baseline (plain text)	PE	0.78	0.88	0.96	0.95	0.86	0.91
	DVT	0.42	0.84	0.89	0.89	0.57	0.86
	PE and/or DVT	0.77	0.90	0.85	0.96	0.81	0.93
	Incidentaloma	0.32	0.43	0.29	0.32	0.30	0.37
Baseline + annotations	PE	0.99	1.00	0.97	0.95	0.98	0.98
	DVT	0.73	1.00	0.89	1.00	0.80	1.00
	PE and/or DVT	0.92	1.00	0.85	0.96	0.89	0.98
	Incidentaloma	0.67	NC	0.50	NC	0.57	NC
Baseline + annotations + section typing	Incidentaloma	0.46	NC	0.63	NC	0.53	NC
Critical sections* + annotations	Incidentaloma	0.60	0.76	0.75	0.81	0.67	0.80

ISSN: 1471-2105