- contents -- entities -- signals -- coreference -- relationships -- general guidelines -- recipe -- histopath -- radiology -

- previous -- next -

Relationships

Examples

The description of annotations in future sections make use of the following examples:

  1. "This patient has had a lymph node biopsy which shows melanoma in his right groin. Five out of ten nodes were involved. It is clearly secondaries from the melanoma on his right second toe. Although his PET scan is normal he does need a groin dissection. We will perform a CT scan to look at the left pelvic side wall and I will review him together with Dr. X next week."
  2. "I have discussed her with x. We agreed to treat with DTIC, and then consider radiotherapy."
  3. "This 56 year old woman was admitted to x ward on the date above, with increasing facial pain. This was initially relieved by co-codamol"
  4. "There was no evidence of extra pelvic secondaries"

The relationships

Relationships describe some interaction between entities. Like entities, relationships have a type. The CLEF relationships are given below. For each, its arguments are given. These are the entity or signals types that interact. Also a brief description and example.

Relationship type First argument type Second argument type Description Example
has_target Investigation, Intervention Locus Relates an intervention or an invetigation to the bodily locus at which it is targetted. There are several has_target relationships in example 1. lymph node is the target of the investigation biopsy, and groin is the target of the intervention dissection.
has_finding Investigation Condition, Result Relates a condition to an investigation that demonstrated its presence, or a result to the investigation that produced that result. In example 1, melanoma is a finding of the biopsy. normal is a finding of PET scan
has_indication Drug or device, Intervention, Investigation Condition Relates a condition to a drug, intervention, or investigation that is targetted at that condition In example 3, co-codamol is indicated by pain (which has two mentions, "pain" and "this".
has_location Condition Locus Relationship between a condition and a locus: describes the bodily location of a specific condition. has_location may also describe the location of malignant disease in lymph nodes, relating an involvement to a locus. There are three has_location relationships in example 1, melanoma is located in groin, and a second melanoma entity is located in second toe. The involvement "5 out of 10" is located in the lymph node entity. In example 3, pain is located in face.
Modifies Negation signal Condition Relates a condition to its negation or uncertainty about it "no evidence" in example 4 is a negation of "secondaries"
Modifies Laterality signal Locus, Intervention Relates a bodily locus or intervention to its sidedness: right, left, bilateral. In example 2 there are several laterality modifiers. For instance, the second toe has a laterality of right. For an example of an intervention laterality modifier, consider "right thoracotomy"
Modifies Sub-location signal Locus Relates a bodily locus to other information about the location: upper, lower, extra- etc. In example 4 there is a sub-location modifier. The pelvis has a sub-location of extra (as in external to).