|
|
|
|
An ontological probabilistic approach to the breast cancer problem in semantic medicine |
|
|
|
|
Bevilacqua V1, Cucci F1, Santarcangelo V1, Mastronardi G1*
Bevilacqua V et al.
Motivation
We show the application of semantic web and RDF (Resource Description
Framework) for organization of Breast Cancer patient for diagnostic.The analysis is
based on patient’s genomic map, information about risk factors and family history.
Methods
The RDF data model consists of resources, properties and values. Properties are
the relationships that bind together the resources and values, and are also
identified by URIs. The RDF data model allows to define a simple model to
describe the relationships between resources, in terms of properties identified by a
name and its values. However, RDF data model provides no mechanism for
declaring these properties, or to define relationships between these properties and
other resources. For this task there is RDF Schema. RDF graphs can be useful to
describe the patient as an URI with his properties and relations with other people
which shares a family history (characterized by his health problems). This could be
useful to have a complete clinical view of a person (this could be implemented
thanks to the family doctors) and to help doctors who haven’t information about a
patient. In particular in this work we show the application of this semantic medicine
approach is an experimental ontology for the Breast Cancer Diagnose. This
ontology model information from a world focused on real cases of patients
previously affected by breast cancer, whereas biological data of the patient (age,
menopause, etc..) and specific characteristics of disease (tumor size, lymph nodes
involved, grade, location, etc.).A short list of general risk factors crucial for breast
cancer is advantage age (65 age), late menopause (55 years or after), family
history (particularly in cases where one or more close relatives degree (mother,
sister and daughter) has been diagnosed with cancer at an early age); excessive
consumption of alcohol; have had first child after thirty years old; -have been
submitted recently to hormonal therapy based on only estrogen; early menarche
(first menstrual period before 12 years old); obesity after menopause; BRCA1 e
BRCA2 mutations. In our study we test our approach with Anna. Anna is a young
woman of twenty-three, his grandmother died of breast cancer, two other aunts
have died fifty years for ovarian cancer, and a third aunt of thirty-nine has breast
cancer. The gynecologist suggests that in the family of the patient there is a
mutation in one of two genes that control susceptibility to the development of
breast or ovarian cancer: the gene BRCA1 and BRCA2. To Anna veins discovered
a suspicious lump in her right breast and the doctor prepares a biopsy and asks
Anna to perform a DNA test, to detect a genetic mutation. It’s necessary to trace
the exact form of the gene mutated in a parent carrier of a change and identify it.
Anna learns that her aunt sick from a genetic test was a carrier of a rare mutation
in BRCA1 and BRCA2 genes. Analysis of the DNA of Anna turns out that she
carries the same mutation in the aunt. The RDF mapping can organize these
information and prevents the loss of data fundamental for the study of pathologies
in family. RDF, with the use of SKOS and OWL is used to model these scenarios
and organize important data. It can be interrogated by SPARQL queries.This
information need to be interpreted and elaborated by an algorithm. This algorithm
can be static (his rules are defined) or dynamic (his rules are obtained by a RDF
Knowledge Base). The output is a prediction about the probability to get Breast
Cancer.So far we have considered values defined for properties, but the diagnosis
of diseases: symptoms, causes and consequences are characterized by a degree
of uncertainty, which complicates the conceptualization of these sectors in formal
ontology and thus limits the ability of understanding at the level of machine. Then
it’s necessary to declare properties used to express probabilistic knowledge,
generic and concrete, of fundamental importance for the functioning of probabilistic
reasoning. Each factor of risk has an own range in which is defined and a
probability (pi) associated. Person considered can have some probability (pi)
related with its characteristics, in fact, it has a probability p1 associated at his age,
a probability p2 associated at the age of menarche if it is in the range [10-12], a p3
probability associated at his age of menopause (if it’s over 60), a p4 probability
associated to a positive outcome of a Last Echograpy, a p5 probability based on
the birthday of its first daughter. These information are used by the algorithm to
analyze the RDF Data of the Patient and produce a prediction about the probability
to get Breast Cancer.
Results
Possible results of this approach could be an implementation of a system that
through an user-friendly GUI can create, have access, modify and update the RDF
document associated to every patient. This system can be implemented integrating
the swi-prolog parser.
Scarica il documento originale. |
|
|
