This project aims at applying ontology-based data management models, methods and technologies to gather, integrate and analyze data about health-related quality of life of cancer patients. Our specific goals are: 1) the construction of an ontology about health-related quality of life of cancer patients; 2) the application of such an ontology in the development of an application for gathering, integrating and analyzing information about the quality of life of thyroid cancer patients. The adoption of the ontology-based data management methods and techniques allows for both flexible and extendible solutions and for the tight integration with existing data and meta-data resources and standards (vocabularies, ontologies) in the field of health-related quality of life of cancer patients. Moreover, we plan to extend the current OBDM technology to take into account the privacy and data protection issues of the above application domain.
We describe the progress and innovation provided by the four main project contributions:
(1) Development of an OWL ontology on health-related quality of life of thyroid cancer patients. In the medical field there is an increasing interest in quality of life (QoL) of cancer patients. However, it is reported that physicians underestimate the amount of physical symptoms associated with thyroid cancer treatments [3] and data on QoL are not included in the primary endpoints of randomized controlled trials. Clinical practice guideline recommendations should be graded on the basis of evidence about effectiveness, safety, quality of life, patients¿ compliance and costs. While the effectiveness and safety of specific therapies can be easily derived from randomized controlled trials, data on QoL, patients¿ compliance and costs are lacking or do not reflect the real life practice. To the best of our knowledge, no published study attempted to apply an ontology-based approach in order to integrate and correlate real-life QoL data with efficacy and safety of the treatment. On the other hand, several ontologies have been developed for providing an integrated vocabulary and a semantic description of the cancer domain [12,13]. We thus intend to analyze such proposals, extract from them a unified ontology, and extend it to model aspects related to the QoL of thyroid cancer patients. The outcome of our work will be a novel, rich ontology, which might become a reference ontology for applications that manipulate data on the QoL of this class of patients.
(2) Integration of the new ontology with popular biomedical ontologies and existing standards. To understand the formal nature of the various available medical vocabularies and ontologies and the existing initiatives for their integration, such as UMLS, we will pursue a logic-based approach, which aims at characterizing the relationship of such proposals with standard ontology languages (OWL), in the line with previous work [14,7]; also, we will investigate the application of ontology matching and alignment methods and tools [2] towards this goal.
(3) Definition of a mechanism to specify and ensure data privacy and protection rules at the conceptual level. The distributed nature of the context considered in the present project challenges most of the approaches developed in the literature of data privacy [9]. Among various proposals, we will extend the work on controlled query evaluation, which considers a framework where protection policies are declaratively specified in terms of queries that the system must not answer to preserve privacy. In this approach, the challenge is to provide answers to non-protected queries guaranteeing the crucial properties of correctness (the data protection policy is never violated by the query answers returned) and completeness (the maximum amount of query answers that are compatible with the data protection policy are returned). This problem has been studied only recently for OWL ontologies, for which preliminary results have been provided [3]. We aim at extending these studies to a full-fledged OBDM scenario.
(4) Development of an OBDM system for integration and analysis of data about thyroid cancer patients. The application of OBDM to biomedical data is still quite unexplored. In particular, biomedical ontologies have been rarely connected to the data to a large scale, mainly due to the variety of formats in which both ontologies and data are represented in this context, and also for the typically large size of ontologies and information to manipulate. As for the first problem, as already mentioned (cf. (2)), we will investigate the relationships between the most important proposals for biomedical ontologies and OWL, and will rely on a unified, formal representation of the domain we are going to consider (QoL of thyroid cancer patients). As for the second issue, foundational theoretical studies of the last 15 years have clearly established the kind of languages to use when connecting ontologies to data [15,1], and recent advancements have obtained crucial optimizations to make OBDA effective in practice, so that we can now rely on advanced tools for OBDM, such as MASTRO [16].
[12] J. G. Golbeck et al.:. The national cancer institute's thesaurus and ontology. J. Web Sem. 2003
[13] M. Brochhausen et al. The ACGT Master Ontology and its applications: towards an ontology-driven cancer research and management system. J. Biomedical Informatics. 2011
[14] V. Kashyap, A. Borgida: Representing the UMLS Semantic Network Using OWL. ISWC 2003
[15] A. Poggi et al. Linking Data to Ontologies. J. Data Semantics. 2008
[16] F. Di Pinto et al. Optimizing query rewriting in ontology-based data access. EDBT 2013