Recognition and Evaluation of Clinical Section Headings in Clinical Documents Using Token-Based Formulation with Conditional Random Fields
Table 2
Statistics of the section heading recognition corpus. Since the corpus only contained the topmost sections, several different concepts or representations may be included in each section heading category. For instance, “Personal Histories” included the occupation, daily activity amount, substance history, and allergies.
Section
Description
Number
Percentage
Chief Complaints
A statement describing the symptoms, problems, diagnoses, or other factors that are the reason of a medical encounter.
803
5.7%
Present Illness
Separated paragraphs summarizing chief complaints related history.
843
6.0%
Personal Histories
A merged concept of individual related histories, including past medical history, past surgical history, social history, and allergy.
2701
19%
Family Histories
The health status of parents, children, siblings, and spouse, whether dead or alive.
486
3.4%
Physical Examinations
The process by which a medical professional investigates the body of a patient for signs of disease.
1104
7.9%
Laboratory Examinations
Biochemical studies performed in clinical laboratory.
401
2.8%
Radiology Reports
Image studies. Some examples are X-ray, CT, MRI, and PET.
87
<1.0%
Data
A merged concept including laboratory examinations and radiology reports.
103
<1.0%
Impression
Medical diagnoses judged by doctors, also called assessments.
884
6.3%
Recommendations
Treatments toward impressions, also called plans.
468
3.3%
Others
Other section headings not included in the categories above, for example, patient ID, doctor ID, and hospital ID.
