=Paper=
{{Paper
|id=Vol-3603/Paper5
|storemode=property
|title=An Ontology for Mammography Screening Recommendation
|pdfUrl=https://ceur-ws.org/Vol-3603/Paper5.pdf
|volume=Vol-3603
|authors=Cindy Acuña,Yasmine Anchén,Edelweis Rohrer,Regina Motz
|dblpUrl=https://dblp.org/rec/conf/icbo/AcunaARM23
}}
==An Ontology for Mammography Screening Recommendation==
https://ceur-ws.org/Vol-3603/Paper5.pdf
An ontology for mammography screening
recommendation
Cindy Acuña1 , Yasmine Anchén2* , Edelweis Rohrer1* and Regina Motz1*
1
Facultad de Ingeniería, Universidad de la República, Julio Herrera y Reissig 565, 11300 Montevideo, Uruguay
2
Facultad de Medicina, Universidad de la República, Av. Gral. Flores 2125, 11800 Montevideo, Uruguay.
Abstract
This article presents the development of a simple Mammography Screening Ontology (MAMO-SCR-Onto)
designed to provide mammography application recommendations for the early detection of breast cancer
in patients at average risk of the disease. We highlight the simplicity of this ontology that allows, on
the one hand, to be a formal communication tool between doctor-patient; on the other, to be an open
educational resource for the training of preservice health professionals. The ontology is developed in
OWL and accessible with an open licenced.
Keywords
mammography screening recommendation, breast cancer early detection, patients with average risk
1. Introduction
Breast cancer is a significant global health concern, underscoring the vital need for effective early
detection methods. The American Cancer Society (ACS), a leading cancer-fighting organization,
has established comprehensive guidelines for mammography as a crucial test for the early
detection of breast cancer in women with an average risk of contracting breast cancer [1].
In spite of other recognized institutions’ recommendations, in the present work we consider
ACS’s recommendations as a case study without loss of generality. These recommendations are
about the frequency application of mammography as a test for early breast cancer detection
based on women’s values in risk factors. On the other hand, ontologies are widely adopted
and proven modelling artefacts for conceptualizing various domains, including education and
health. They provide a structured framework to describe vocabularies in terms of concepts
(or entity types), instances (or individuals), and roles (or relations), as well as assertions or
constraints about them [2]. Notably, the W3C standard ontology language OWL, with its formal
semantics based on description logics, facilitates the implementation of a comprehensive set
of assertions for a given vocabulary, particularly in the health domain [3]. OWL enables the
automatic validation of ontology constraints and facilitates the inference of assertions that
Proceedings International Conference on Biomedical Ontologies 2023, August 28-September 1, 2023, Brasilia, DF, Brazil
*
Corresponding authors.
†
These authors contributed equally: Yasmine Anchén, Regina Motz, Edelweis Rohrer
$ cindy.acuna@fing.edu.uy (C. Acuña); yasanmo@gmail.com (Y. Anchén); erohrer@fing.edu.uy (E. Rohrer);
rmotz@fing.edu.uy (R. Motz)
0009-0006-4254-4817 (C. Acuña); 0009-0003-8648-2687 (Y. Anchén); 0000-0002-8257-5291 (E. Rohrer);
0000-0002-1426-562X (R. Motz)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
48
may not have been explicitly declared. This capability empowers the ontology to go beyond
the explicitly stated information, uncovering implicit knowledge and facilitating more robust
reasoning. Reasoners, i.e. algorithmic implementations such as the tableau algorithm, play
a pivotal role in this process. By examining an ontology, reasoners ensure consistency by
verifying that the specified constraints are met while drawing logical inferences based on the
underlying description logic semantics [4].
The present work makes a significant contribution by modelling and implementing a simple
owl ontology designed to provide mammography screening application recommendations for
the early detection of breast cancer in women at average risk of the disease. The ontology
represents the conditions taken into account to consider a woman within the group "women
with an average risk of contracting breast cancer" (group of women to whom the guide is
directed), and it also represents specific recommendations that these women receive according
to their ages, such as the periodicity of the study. These conditions are modelled as constraints,
which the reasoner utilizes to infer the appropriate advice by the guidelines established by
the American Cancer Society. This ontology captures the intricate relationships between
various conditions that influence the decision to perform screening mammography, such as the
risk group to which the woman belongs (taking into account confirmed or suspected genetic
mutation, personal history, and exposure to chest radiation treatments) and the woman’s age. By
formalizing expert knowledge and integrating it into an ontology, we facilitate the development
of intelligent decision support systems that can aid healthcare professionals in making informed
and personalized screening recommendations. Moreover, this formal specification establishes a
standardized framework and is a valuable tool for visualization and reasoning on the conditions.
By leveraging the power of ontologies and their reasoning capabilities, we can optimize breast
cancer screening protocols and ultimately contribute to improving healthcare outcomes for
individuals at average risk.
This paper presents our approach to designing the mammography screening ontology, includ-
ing selecting relevant concepts, relationships, and axioms. We outline the process of integrating
domain-specific knowledge from authoritative sources, such as the ACS recommendations and
the NCIt thesaurus [5] into the ontology. Furthermore, we discuss the benefits and challenges
of using OWL and reasoning techniques to ensure the ontology’s consistency, coherence, and
inferential capabilities.
The remainder of this paper is organized as follows: Section 2 provides an overview of related
work in the field of ontology-based clinical decision support systems. Section 3 presents the
ontology’s key components and knowledge representation. Section 4 offers the evaluation
of the ontology. Finally, Section 5 concludes the paper with a summary of the contributions,
limitations, and future directions.
2. Related work
Over the past two decades, ontologies have emerged as crucial artefacts in supporting the study
and research of cancer. The adoption of ontologies as a powerful approach has been evident
through the development of numerous ontologies in the biomedical domain [6].
One prominent example of a comprehensive ontology in the field of cancer is the National
49
Cancer Institute Thesaurus and Ontology (NCIt). Designed as an extensive compilation of
terminologies, the NCIt covers various facets of cancer research and care [5]. As a fundamental
resource, the NCIt plays a pivotal role in facilitating standardized communication and promoting
knowledge sharing among researchers, clinicians, and other healthcare professionals in the field
of oncology. However, the significance of the NCIt and other ontologies in the cancer domain
extends beyond providing a common vocabulary. These ontologies establish a structured
framework that enables collaboration, interoperability, and the integration of diverse data
sources. They serve as a foundation for knowledge representation, helping researchers and
practitioners to capture and organize complex cancer-related information effectively.
In recent years, there has been increasing interest in using ontologies and knowledge graphs
in cancer research, as highlighted by the comprehensive review by Silva et al. [7]. This study
examined 141 open articles published between 2012 and 2021. Of the papers analyzed, 18
focus on breast cancer-related ontologies. We can observe that breast cancer ontologies serve
several purposes, each addressing different aspects of knowledge representation and integration
within the field or supporting decision-based systems. Some of these ontologies prioritize
the computational accessibility of breast cancer knowledge to represent the information in a
format that computer systems can efficiently process [8, 9, 10, 11]. On the other hand, specific
ontologies focus on integrating cancer data from various sources, consolidating the information
in a unified database [12, 13]. In some cases, the ontologies are designed to facilitate effective
communication between doctors and patients, harmonizing terminology and improving mutual
understanding [14] or designed to facilitate database user interfaces, where ontology class and
relationship labels allow text annotation within the breast cancer domain [15]. In addition,
some ontologies in the breast cancer domain claim to represent causal associations between
breast cancer incidence and various risk factors, providing valuable insight into the complex
relationships involved [16, 8, 17, 18]. Sherimon et al. [19] describe an architecture of six
ontologies (patient ontology, breast cancer, symptoms, risk, lab test and questionnaire) to
predict breast cancer. However, we cannot find any of these ontologies accessible. Another
application involves the semantic modelling of breast cancer-related drugs using ontological
approaches, allowing the inference of possible drug repositioning strategies [20]. However, the
most notable applications are those that produce clinical inferences [21, 22, 23].
Regarding the radiology domain, we find Sun et al. work presenting an ontology for training
breast radiologists [24]. Focusing on the mammography domain, the GIMI Project developed
a fundamental mammographic ontology and an ontology for computerized training in breast
radiology [25, 26, 27, 28, 29]. Other works on mammographic ontology are Bulubu et al. [30]
describing ontology-based mammography annotation and a case-based retrieval approach for
breast masses from the digital mammography archive, Ilyass et al. [31] which annotates images
in mammograms according to the concepts of a breast cancer ontology which produces RDF
metadata and the more recent ones [32, 33, 34, 35].
However, to the best of our knowledge, no open ontologies provide mammography application
recommendations for the early detection of breast cancer in women at average risk of the disease
as the simple one that we present in this work.
50
3. Modelling the mammography screening recommendation
ontology
ACS recommendations were based on the quality of the evidence obtained and judgment
on the balance of benefits and harms of conducting the study, classifying them into strong
recommendations (benefits of conducting the study outweigh the desirable effects) or qualified
recommendations (there is clear evidence of benefit but less certainty about the balance of
harm/benefit, or on the values and preferences of the patient, allows discussion between doctor
and patient to make the decision). Based on the age of women at average risk of breast cancer, the
ACS recommends Women at average risk of breast cancer should have screening mammography
starting at age 40, following the age range recommendations below.
∙ Women should begin annual screening between the ages of 40 and 44 as a qualified recom-
mendation.
∙ Women ages 45 to 54 should have screening mammograms annually as a strong recommen-
dation.
∙ Women age 55 and older should have screening mammograms biennially or have the
opportunity to continue screening annually as a qualified recommendation.
Given the relevance of the ACS recommendations, both for the doctor and the patient, the
contribution of the ontology described in this section is twofold: (i) conceptualizing the concrete
domain by describing those properties or attributes of women which are relevant inputs for
applying the recommendation as well as the rules behind it, and (ii) computing a customized
recommendation for a set of women instances, which represents a useful tool that aims most
doctors and patients quick access to the guide recommendation and to meet the ultimate goal
of the earliest prevention of the breast cancer.
Figure 1 depicts the mammography screening recommendation ontology, which has three
main classes: Woman, History and Recommendation.
Even though for the scenario described in the present paper, it is important to represent women
with an intermediate risk of contracting breast cancer, this group of women is represented
by the subclass With intermediate risk, of a more general class Woman, to facilitate a possible
extension of the ontology if maybe in the future recommendations for other groups of women
were conceptualized (for example, with high or low risk of breast cancer). Classes Age Range 40
to 44, Age Range 45 to 54, and Age gt or eq 55 represent subsets of women with an intermediate
risk that have corresponding ages.
As for the class Woman, the class History is modelled as a general class even though in the
present work, we focus on the subclass Personal Medical History that represents the set of
medical records of women. Properties hasHistory and doesNotHaveHistory connect each woman
to her medical records. The property hasHistory connects women to diseases, clinical diagnoses
or laboratory test results that they have. The property doesNotHaveHistory connects women to
those diseases, clinical diagnoses or important laboratory test results so that the doctor knows
that the patient does not have it. It is highly relevant that this be made explicit when evaluating
the type of breast cancer risk corresponding to the woman and to be able to make a correct
recommendation. The ACS guide clarifies that the recommendations are aimed at women with
an average risk of breast cancer; these are those without a personal history of breast cancer,
51
Figure 1: Mammography screening recommendation ontology model.
without suspicion or confirmation of a known genetic mutation that increases the risk of breast
cancer and without previous history of chest radiotherapy at a young age. These conditions are
represented in our ontology by individuals confirmed or suspected genetic mutation, personal
history of breast cancer, radiotherapy to chest, as instances of the class Personal Medical History.
The class Recommendation represents the set of recommendations provided by the guide.
Each particular recommendation has three characteristics or dimensions that describe it: (i) the
strength of the recommendation given by the level of evidence to issue the recommendation,
represented by the class Recommendation Type, which has instances Strong and Qualified (ii) the
recommended mammography study, represented by the class Mammography and the instance
Screening Mammography 1 , and (iii) the frequency with which mammography should be per-
formed, represented by the class Periodicity, with two instances: Annual and Annual or Biaannual.
Properties hasRecType, hasMammography and hasPeriodicity connect each recommendation
instance to the three dimensions.
Restrictions on classes and properties described above are represented in Table 1 as six groups
of Tbox and Rbox axioms. Group (1) represents that classes Woman, History, Recommendation,
Recommendation Type, Mammography and Periodicity cannot share instances with each other.
The axiom (2) defines the class With intermediate risk as the set of women that meet the three
conditions given by the guide to classify them in the category of women that have an average
risk of breast cancer. Group (3) defines subclasses Age Range 40 to 44, Age Range 45 to 54 and
Age gt or eq 55 as subsets of women with an average risk of breast cancer, based on the age
range, represented in the implemented OWL ontology by the data property age. This is why
1
Note that for the scenario described in this work, only screening mammographies are considered.
52
Table 1
Mammography screening recommendation domain Tbox and Rbox.
Axiom Description
𝑊 𝑜𝑚𝑎𝑛 ⊓ 𝐻𝑖𝑠𝑡𝑜𝑟𝑦 ⊑ ⊥
𝑊 𝑜𝑚𝑎𝑛 ⊓ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊑ ⊥
𝑊 𝑜𝑚𝑎𝑛 ⊓ 𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊑ ⊥
𝑊 𝑜𝑚𝑎𝑛 ⊓ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑇 𝑦𝑝𝑒 ⊑ ⊥
𝑊 𝑜𝑚𝑎𝑛 ⊓ 𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦 ⊑ ⊥
𝐻𝑖𝑠𝑡𝑜𝑟𝑦 ⊓ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊑ ⊥
(1) 𝐻𝑖𝑠𝑡𝑜𝑟𝑦 ⊓ 𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊑ ⊥ Pairwise class disjointness.
𝐻𝑖𝑠𝑡𝑜𝑟𝑦 ⊓ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑇 𝑦𝑝𝑒 ⊑ ⊥
𝐻𝑖𝑠𝑡𝑜𝑟𝑦 ⊓ 𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦 ⊑ ⊥
𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊓ 𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊑ ⊥
𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊓ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑇 𝑦𝑝𝑒 ⊑ ⊥
𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊓ 𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦 ⊑ ⊥
𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊓ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑇 𝑦𝑝𝑒 ⊑ ⊥
𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊓ 𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦 ⊑ ⊥
𝑊 𝑖𝑡ℎ 𝑖𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑡𝑒 𝑟𝑖𝑠𝑘 ≡ 𝑊 𝑜𝑚𝑎𝑛⊓
(2) ∃𝑑𝑜𝑒𝑠𝑁 𝑜𝑡𝐻𝑎𝑣𝑒𝐻𝑖𝑠𝑡𝑜𝑟𝑦.{𝐶𝑜𝑛𝑓 𝑖𝑟𝑚𝑒𝑑 𝑜𝑟 𝑠𝑢𝑠𝑝𝑒𝑐𝑡𝑒𝑑 𝑔𝑒𝑛𝑒𝑡𝑖𝑐 𝑚𝑢𝑡𝑎𝑡𝑖𝑜𝑛}⊓ Women with intermediate-risk are those
∃𝑑𝑜𝑒𝑠𝑁 𝑜𝑡𝐻𝑎𝑣𝑒𝐻𝑖𝑠𝑡𝑜𝑟𝑦.{𝑃 𝑒𝑟𝑠𝑜𝑛𝑎𝑙 ℎ𝑖𝑠𝑡𝑜𝑟𝑦 𝑜𝑓 𝑏𝑟𝑒𝑎𝑠𝑡 𝑐𝑎𝑛𝑐𝑒𝑟}⊓ do not meet a set of conditions.
∃𝑑𝑜𝑒𝑠𝑁 𝑜𝑡𝐻𝑎𝑣𝑒𝐻𝑖𝑠𝑡𝑜𝑟𝑦.{𝑅𝑎𝑑𝑖𝑜𝑡ℎ𝑒𝑟𝑎𝑝𝑦 𝑡𝑜 𝑡ℎ𝑒 𝑐ℎ𝑒𝑠𝑡}
𝐴𝑔𝑒 𝑔𝑡 𝑜𝑟 𝑒𝑞 55 ≡ 𝑊 𝑖𝑡ℎ 𝑖𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑡𝑒 𝑟𝑖𝑠𝑘 ⊓ ∃𝑎𝑔𝑒. ≥ 55
𝐴𝑔𝑒 𝑔𝑡 𝑜𝑟 𝑒𝑞 55 ⊑ ∃ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛.{55 𝐴𝑔𝑒 𝑟𝑎𝑛𝑔𝑒 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛}
(3) 𝐴𝑔𝑒 𝑅𝑎𝑛𝑔𝑒 40 𝑡𝑜 44 ≡ 𝑊 𝑖𝑡ℎ 𝑖𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑡𝑒 𝑟𝑖𝑠𝑘 ⊓ ∃𝑎𝑔𝑒. ≥ 40 ⊓∃𝑎𝑔𝑒. ≤ 44 Each age subclass has corresponding age
𝐴𝑔𝑒 𝑅𝑎𝑛𝑔𝑒 40 𝑡𝑜 44 ⊑ ∃ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛.{40 𝑡𝑜 44 𝐴𝑔𝑒 𝑟𝑎𝑛𝑔𝑒 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛} restriction and recommendation
𝐴𝑔𝑒 𝑅𝑎𝑛𝑔𝑒 45 𝑡𝑜 54 ≡ 𝑊 𝑖𝑡ℎ 𝑖𝑛𝑡𝑒𝑟𝑚𝑒𝑑𝑖𝑎𝑡𝑒 𝑟𝑖𝑠𝑘 ⊓ ∃𝑎𝑔𝑒. ≥ 45 ⊓∃𝑎𝑔𝑒. ≤ 54
𝐴𝑔𝑒 𝑅𝑎𝑛𝑔𝑒 45 𝑡𝑜 54 ⊑ ∃ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛.{45 𝑡𝑜 54 𝐴𝑔𝑒 𝑟𝑎𝑛𝑔𝑒 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛}
∃ℎ𝑎𝑠𝐻𝑖𝑠𝑡𝑜𝑟𝑦.⊤ ⊑ 𝑊 𝑜𝑚𝑎𝑛 ⊤ ⊑ ∀ℎ𝑎𝑠𝐻𝑖𝑠𝑡𝑜𝑟𝑦.𝐻𝑖𝑠𝑡𝑜𝑟𝑦
∃𝑑𝑜𝑒𝑠𝑁 𝑜𝑡𝐻𝑎𝑣𝑒𝐻𝑖𝑠𝑡𝑜𝑟𝑦.⊤ ⊑ 𝑊 𝑜𝑚𝑎𝑛 ⊤ ⊑ ∀𝑑𝑜𝑒𝑠𝑁 𝑜𝑡𝐻𝑎𝑣𝑒𝐻𝑖𝑠𝑡𝑜𝑟𝑦.𝐻𝑖𝑠𝑡𝑜𝑟𝑦
∃ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛.⊤ ⊑ 𝑊 𝑜𝑚𝑎𝑛 ⊤ ⊑ ∀ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛.𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛
(4) ∃ℎ𝑎𝑠𝑅𝑒𝑐𝑇 𝑦𝑝𝑒.⊤ ⊑ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊤ ⊑ ∀ℎ𝑎𝑠𝑅𝑒𝑐𝑇 𝑦𝑝𝑒.𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑇 𝑦𝑝𝑒 Properties domain and range
∃ℎ𝑎𝑠𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦.⊤ ⊑ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊤ ⊑ ∀ℎ𝑎𝑠𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦.𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦
∃ℎ𝑎𝑠𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦.⊤ ⊑ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 ⊤ ⊑ ∀ℎ𝑎𝑠𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦.𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦
∃𝑖𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑.⊤ ⊑ 𝑊 𝑜𝑚𝑎𝑛
⊤ ⊑ ∀𝑖𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑.(𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊔ 𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦 ⊔ 𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑇 𝑦𝑝𝑒)
(5) Dis(hasHistory, doesNotHaveHistory) A woman cannot have a clinical situation and
does not have it at the same time
ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 𝑜 ℎ𝑎𝑠𝑅𝑒𝑐𝑇 𝑦𝑝𝑒 ⊑ 𝑖𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 A woman is recommended a strong or
(6) ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 𝑜 ℎ𝑎𝑠𝑀 𝑎𝑚𝑚𝑜𝑔𝑟𝑎𝑝ℎ𝑦 ⊑ 𝑖𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 qualified recommendation, a kind of
ℎ𝑎𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛 𝑜 ℎ𝑎𝑠𝑃 𝑒𝑟𝑖𝑜𝑑𝑖𝑐𝑖𝑡𝑦 ⊑ 𝑖𝑠𝑅𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 mammography and a periodicity.
the guide sets a different recommendation for each group of women, which is represented in
the ontology as restrictions on instances of classes Age Range 40 to 44, Age Range 45 to 54 and
Age gt or eq 55 that must have the corresponding recommendation.
Axioms (2) and (3) are used by the reasoner to classify instances of the class Woman into
categories given by descriptions of classes With intermediate risk, Age Range 40 to 44, Age
Range 45 to 54 and Age gt or eq 55, entailing the corresponding recommendation. Group (4)
describes restrictions on domains and ranges of properties. (5) and (6) describe restrictions
on roles (Rbox axioms). The axiom (5) represents that a woman cannot be connected to the
same medical record by both properties hasHistory and doesNotHaveHistory. In (6), the property
isRecommended is defined as the superproperty of role chains hasRecommendation o hasTecType,
53
hasRecommendation o hasMammography and hasRecommendation o Periodicity. Besides inferring
if the recommendation is for 40 to 44, 45 to 54 or ≥ 55 age range, for each woman, the reasoner
directly entails if the recommendation is strong or qualified, the kind of mammography and its
periodicity.
4. Validation of the ontology.
The model and restrictions described in the previous section were implemented in the standard
W3C ontology language OWL2 by using the Protégé editor. Part of the vocabulary of the
implemented ontology was reused from the standard vocabulary of the NCI Thesaurus, e.g.
Woman, Mammography and History [36]. The ontology consistency and inferences were verified
by using the Hermit reasoner [37].
Figure 2: Reasoner entailments for a woman with intermediate risk and age range 40 to 44.
Figure 3: Reasoner entailments for a woman with intermediate risk and age range ≥ 55.
The ontology was tested with six different instances of the class Woman. Three of them meet
conditions to be classified as with intermediate risk of breast cancer, each one within a different
age range, one woman meets conditions to be classified as with intermediate risk but does not
fall into any of the three age range, and the other two women do not meet conditions to be
54
classified as with intermediate risk of breast cancer. Figures 2 to 7 show inferences made by the
reasoner for each Woman instance. The ontology is available in https://bioportal.bioontology.
org/ontologies/MAMO-SCR-ONTO and has the https://creativecommons.org/licenses/by-sa/4.0/
license.
Figure 4: Reasoner entailments for a woman that does not meet all conditions for intermediate risk.
Figure 5: Reasoner entailments for a woman that does not meet any condition for intermediate risk.
Figure 6: Recommendation entailed for a woman with intermediate risk and age range 45 to 54.
55
Figure 7: Reasoner entailments for a woman with intermediate risk without the age data.
5. Conclusions and future work
This paper presents a simple Mammography Screening Ontology (MAMO-SCR-Onto) designed
to provide mammography application recommendations for the early detection of breast cancer
in patients at average risk of the disease. Risk stratification allows for personalized recommen-
dations regarding the frequency, the recommendation strength, and the need for screening
mammography tailored to an individual’s risk profile. In this sense, MAMO-SCR-Onto is only
applied to patients with average disease risk, where the risk is defined based on three conditions
of the woman’s clinical history. However, the risk could be obtained by applying different evalu-
ation risk models outside the current ontology [38]. This is not necessarily seen as a limitation of
the MAMO-SCR-Onto ontology, in the sense that its design aimed to focus the recommendation
for exclusively medium-risk patients while maintaining its simplicity to be a helpful tool in
doctor-patient communication, and in the training of preservice health professionals.
As a main conclusion, from the point of view of final users, the advantage of using the MAMO-
SCR-Onto is that medical professionals and patients can access guideline recommendations in a
non-verbose presentation. Notably, the ontology makes it explicit whether these recommen-
dations are classified as strong or qualified, facilitating physician and patient to question the
benefits and harms of conducting the study and subsequently assisting decision-making by the
patient. Ultimately, this comprehensive ontology enables individuals to make well-informed
decisions regarding their healthcare attention.
From the point of view of ontology engineers, another conclusion of our work is regarding
ontology maintainability and reusability quality attributes. The MAMO-SCR-Onto represents
the conditions that women must meet to receive specific recommendations, established in the
ACS guidelines [39]. However, using the ACS guide is only a showcase proposal since it is no
difficulty to change ACS guide for any other one [40, 41]. The recommendations in general
terms established by the guidelines are concordant, and variations occur when setting the age
range in the recommendations by group and the frequency of screening, these constraints are
very well identified in the ontology, and they are straightforward the change.
As a future work, the ontology could be extended for women at high risk of breast cancer,
representing the recommendations provided for them [42]. More precise information about the
56
corresponding type of risk assessment could be included, such as being based on a statistical
model of breast cancer risk assessment, for example, the Tyrer-Cuzick , Clause and Gail models,
that the physician at the time of the consultation can assess the risk of breast cancer of the
woman with greater precision and in turn provide the relevant recommendations [43, 44, 45].
References
[1] American Cancer Society, https://www.cancer.org/, Last date accessed July 2023.
[2] S. Staab, R. Studer, Handbook on Ontologies, 2nd ed., Springer Publishing Company,
Incorporated, 2009.
[3] B. C. Grau, I. Horrocks, B. Motik, B. Parsia, P. F. Patel-Schneider, U. Sattler, OWL 2: The
next step for OWL, J. Web Sem. 6(4) (2008).
[4] P. Hitzler, M. Krötzsch, S. Rudolph, Foundations of Semantic Web Technologies, Chapman
& Hall/CRC, 2009.
[5] J. Golbeck, G. Fragoso, F. Hartel, J. Hendler, J. Oberthaler, B. Parsia, The National Cancer
Institute’s Thesaurus and Ontology, Web Semantics 1 (2003).
[6] P. L. Whetzel, N. F. Noy, N. H. Shah, P. R. Alexander, C. Nyulas, T. Tudorache, M. A. Musen,
BioPortal: enhanced functionality via new Web services from the National Center for
Biomedical Ontology to access and use ontologies in software applications, Nucleic Acids
Research 39 (2011).
[7] M. C. Silva, P. Eugénio, D. Faria, C. Pesquita, Ontologies and Knowledge Graphs in
Oncology Research, Cancers 14 (2022) 1906.
[8] O. N. Oyelade, A. E. Ezugwu, S. A. Adewuyi, Enhancing reasoning through reduction
of vagueness using fuzzy OWL-2 for representation of breast cancer ontologies, Neural
Computing and Applications 34 (2021).
[9] M. T. D. Melo, V. H. L. Gonçalves, H. D. R. Costa, D. S. Braga, L. B. Gomide, C. S. Alves, L. M.
Brasil, OntoMama: An Ontology Applied to Breast Cancer, Studies in Health Technology
and Informatics 216 (2015) 1104.
[10] J. Xi, L. Ye, Q. Huang, X. Li, Tolerating data missing in breast cancer diagnosis from
clinical ultrasound reports via knowledge graph inference, in: Proceedings of the 27th
ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Singapore, 2021.
[11] S. Bourougaa-Tria, H. Farah, Semantic ontology for representing breast cancer terminology,
in: M. Laouar, V. Balas, B. Lejdel, S. Eom, M. Boudia (Eds.), 12th International Conference
on Information Systems and Advanced Technologies "ICISAT 2022", volume 624 of Lecture
Notes in Networks and Systems, Springer, Cham, 2023.
[12] F. Jusoh, R. Ibrahim, M. S. Othman, N. Omar, Development of breast cancer ontology based
on hybrid approach, International Journal of Innovation in Computing 3 (2013) 1.
[13] O. Seneviratne, S. M. Rashid, S. Chari, J. P. McCusker, K. P. Bennett, J. A. Hendler, D. L.
McGuinness, Knowledge integration for disease characterization: A breast cancer example,
in: Proceedings of the International Semantic Web Conference, Springer, Monterey, CA,
USA, 2018.
[14] M. Tapi Nzali, J. Aze, S. Bringay, C. Lavergne, C. Mollevi, T. Optiz, Reconciliation of
patient/doctor vocabulary in a structured resource, Health Informatics Journal 25 (2019).
57
[15] K. Milian, R. Hoekstra, A. Bucur, A. Ten Teije, F. van Harmelen, J. Paulissen, Enhancing
reuse of structured eligibility criteria and supporting their relaxation, Journal of Biomedical
Informatics 56 (2015).
[16] A. Daowd, M. Barrett, S. Abidi, S. Abidi, Building a knowledge graph representing causal
associations between risk factors and incidence of breast cancer, in: Public Health and
Informatics, IOS Press, 2021.
[17] J. Bouaud, S. Pelayo, J.-B. Lamy, C. Prebet, C. Ngo, L. Teixeira, G. Guézennec, B. Séroussi,
Implementation of an ontological reasoning to support the guideline-based management of
primary breast cancer patients in the DESIREE project, Artificial Intelligence in Medicine
108 (2020) 101922.
[18] M. Gong, Z. Wang, Y. Liu, H. Zhou, F. Wang, Y. Wang, N. Hong, Toward early diagnosis
decision support for breast cancer: Ontology-based semantic interoperability, Journal of
Clinical Oncology 37 (2019).
[19] S. P. C, R. Krishnan, M. James, Mellrak: an ontology driven cdss for symptom assessment,
risk assessment and disease analysis of breast cancer, in: 2021 International Conference
on Software Engineering & Computer Systems and 4th International Conference on
Computational Science and Information Management (ICSECS-ICOCSIM), 2021.
[20] Q. Zhu, C. Tao, F. Shen, C. Chute, Exploring the pharmacogenomics knowledge base
PharmGKB for repositioning breast cancer drugs by leveraging Web ontology language
(OWL) and cheminformatics approaches, in: Pacific Symposium on Biocomputing, 2014.
[21] S. R. Abidi, Ontology-Based Modeling of Breast Cancer Follow-up Clinical Practice
Guideline for Providing Clinical Decision Support, in: Twentieth IEEE International
Symposium on Computer-Based Medical Systems (CBMS’07), 2007.
[22] S. Dasmahapatra, D. Dupplaw, B. Hu, P. Lewis, N. Shadbolt, Ontology-mediated distributed
decision support for breast cancer, in: Artificial Intelligence in Medicine: 10th Confer-
ence on Artificial Intelligence in Medicine, AIME 2005, Aberdeen, UK, July 23-27, 2005.
Proceedings 10, Springer Berlin Heidelberg, 2005.
[23] Implementation of an ontological reasoning to support the guideline-based management of
primary breast cancer patients in the DESIREE project, Artificial Intelligence in Medicine
108 (2020) 101922. Publisher: Elsevier.
[24] S. Sun, P. Taylor, L. Wilkinson, L. Khoo, An ontology for breast radiologist training, in:
Proceedings of the 10th IASTED International Conference on Computers and Advanced
Technology in Education (CATE 2007), Beijing, China, 2007.
[25] D. Qi, Development and evaluation of an ontology for a mammographic computer aided
diagnosis system, Ph.D. thesis, Aberystwyth, 2006.
[26] S. Sun, P. Taylor, L. Wilkinson, L. Khoo, An ontology to support adaptive training for
breast radiologists, in: E. A. Krupinski (Ed.), IWDM 2008, volume 5116 of LNCS, Springer,
Heidelberg, 2008.
[27] S. Sun, P. Taylor, L. Wilkinson, L. Khoo, Individualised training to address variability of
radiologists’ performance, in: Proceedings of the SPIE Symposium on Medical Imaging
(SPIE-MI 2008), SPIE, San Diego, 2008.
[28] P. Taylor, I. Toujilov, Mammographic knowledge representation in description logic, in:
KR4HC 2011, volume 6924 of Lecture Notes in Computer Science, Springer, 2012.
[29] GIMI Mammography Ontology, http://sourceforge.net/projects/gimimammography/, Last
58
date accessed July 2023.
[30] H. Bulu, A. Alpkocak, P. Balci, Ontology-based mammography annotation and case-based
retrieval of breast masses, Expert Systems with Applications 39 (2012).
[31] H. Ilyass, F. S. Mohamed, I. Diop, J. Tarik, Ontology-based mammography annotation
for breast cancer diagnosis, in: 2015 2nd World Symposium on Web Applications and
Networking (WSWAN), 2015.
[32] J. W. Pereira, M. X. Ribeiro, Semantic Annotation and Classification of Mammography
Images Using Ontologies, in: 2021 IEEE 34th International Symposium on Computer-Based
Medical Systems (CBMS), IEEE, 2021.
[33] A. Yagahara, Y. Yokooka, G. Jiang, S. Tsuji, A. Fukuda, N. Nishimoto, K. Kurowarabi,
K. Ogasawara, Construction of mammographic examination process ontology using
bottom–up hierarchical task analysis, Radiological Physics and Tech. 11 (2018).
[34] Y. B. Salem, R. Idoudi, K. Saheb Ettabaa, K. Hamrouni, B. Soleiman, Mammographie image
based possibilistic ontological representation, in: 2018 4th International Conference on
Advanced Technologies for Signal and Image Processing (ATSIP), IEEE, 2018.
[35] Y. B. Salem, R. Idoudi, K. Saheb Ettabaa, K. Hamrouni, B. Solaiman, High level mam-
mographic information fusion for real world ontology population, Journal of Digital
Information Management 15 (2017).
[36] NCI Thesaurus., https://ncit.nci.nih.gov/, Last date accessed June 2023.
[37] I. Horrocks, B. Motik, Z. Wang, The HermiT OWL Reasoner, in: 1st International Workshop
on OWL Reasoner Evaluation (ORE-2012), CEUR-WS.org, 2012.
[38] E. M. Ozanne, B. Drohan, P. Bosinoff, A. Semine, M. Jellinek, C. Cronin, F. Millham,
D. Dowd, T. Rourke, C. Block, K. S. Hughes, Which risk model to use? clinical implications
of the acs mri screening guidelines, Cancer Epidemiol Biomarkers Prev 22 (2013).
[39] Oeffinger KC, Fontham ETH, E. et al., Breast Cancer Screening for Women at
Average Risk:2015 Guideline Update From the American Cancer Society, JAMA.
2015;314(15):1599–1614. doi:10.1001/jama.2015.1278, 2015.
[40] Canadian Task Force Recommendation 2011, https://pubmed.ncbi.nlm.nih.gov/30530611/,
Last date accessed July 2023.
[41] Guía práctica clínica de detección temprana del cáncer de mama - Tamizaje y diag-
nóstico precoz (uruguayan guide 2015), https://www.gub.uy/ministerio-salud-publica/
comunicacion/publicaciones/guia-practica-clinica-deteccion-cancer-mama/, Last date ac-
cessed July 2023.
[42] D. Saslow, C. Boetes, W. Burke, S. Harms, M. O. Leach, C. D. Lehman, E. Morris, E. Pisano,
M. Schnall, S. Sener, R. A. Smith, E. Warner, M. Yaffe, K. S. Andrews, C. A. Russell, American
cancer society guidelines for breast screening with mri as an adjunct to mammography,
CA: a cancer journal for clinicians 57 (2007).
[43] J. Tyrer, S. W. Duffy, J. Cuzick, A breast cancer prediction model incorporating familial
and personal risk factors, Statistics in Medicine 23 (2004).
[44] E. B. Claus, N. Risch, W. D. Thompson, Autosomal dominant inheritance of early-onset
breast cancer. implications for risk prediction, Cancer 73 (1994).
[45] M. H. Gail, L. A. Brinton, D. P. Byar, D. K. Corle, S. B. Green, C. Schairer, J. J. Mulvihill,
Projecting individualized probabilities of developing breast cancer for white females who
are being examined annually, J Natl Cancer Inst 81 (1989).
59