Developing the BOUNCE Psychological Ontology1 Haridimos Kondylakis1,2, Efthymios Alekos2, Kostas Marias1,2, Manolis Tsiknakis1,2, Nikos Papadakis2 1 FORTH-ICS, N. Plastira 100, Heraklion, Crete, Greece 2 HMU-ECE, Heraklion, Crete, Greece kondylak@ics.forth.gr, alekefth@gmail.com, tsiknaki@ics.forth.gr, kmarias@ics.forth.gr, npapadak@hmu.gr Abstract. It is well known that the mental and emotional state of cancer patients plays an important role in the treatment of their disease. As such, for building prediction tools, patient’s psychology should also be considered, along with med- ical, clinical, biological and lifestyle data. However, for modelling patients psy- chological status, only a limited set of terms is available in existing ontologies. The BOUNCE Psychological Ontology (BPO) is an attempt to model all relevant psychological constructs, for cancer patients, effectively capturing patients' emo- tional and mental disposition in order to further study methods for coping with and recovering from the disease. Keywords: psychological; ontology; health Introduction Coping with breast cancer has increasingly become a major socio-economic challenge, among others, due to its constantly increasing incidence in the developing world. The BOUNCE EU project (https://www.bounce-project.eu/), takes into consideration clini- cal, cancer-related biological, lifestyle, and psychosocial parameters in order to predict individual resilience trajectories throughout the cancer continuum. Eventually the tar- get is to increase resilience in breast cancer survivors and help them remain in the work- force and enjoy a better quality of life. In order for such prediction tools to be implemented, a unified view over all available data sources is essential, effectively integrating medical, clinical, biological, lifestyle and psychosocial parameters, collected from four clinical sites across Europe, effec- tively enabling the informed secondary use of patient’s personal data [1, 2]. Although multiple ontologies are already available, integrating and modelling clinical, biological and lifestyle data [3], when coming into psychological constructs, the available ontol- ogies (e.g. Psychological Ontology for Breast Cancer Patients, Mental Functioning On- tology, Emotion Ontology [4]) offer really limited relevant terms, covering only skin deep the aforementioned domain. However, within the BOUNCE project, more than 25 1 Copyright © 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). 2 psychological questionnaires are used, (e.g. the Ten item Personality measure, PTSD Checklist, Connor Davidson Resilience Scale, Family Resilience Questionnaire, NCCN Distress Thermometer), measuring more than 100 psychological parameters, dictating a more detailed and extensive model. The model should be able not only to model con- structs captured by these questionnaires, but also the interrelations between those con- structs, requiring and capturing knowledge available mostly to domain experts. To this direction, in this paper, we present the process we followed to develop the BOUNCE Psychological Ontology (BPO) and present a glimpse of the current status of the developed ontology. The ontology is an OWL ontology developed using Protégé and a first version is already available online2. Methodology There are several methodologies for developing an ontology. These methodologies pro- vide a guideline on how to carry out the activities specified in the ontology development process and the type of techniques that are most appropriate for each activity. In our case we relied on a methodology similar to [5] and [6] including purpose & scope spec- ification, knowledge acquisition, conceptualization, implementation, evaluation and documentation as shown in Fig. 1 Purpose & Scope Knowledge Conceptualization Implementation Evaluation Documentation Specification Acquisition Fig. 1. The methodology adopted for developing the BOUNCE Semantic Model. After defining the purpose and the scope of the developed ontology, in the knowledge acquisition step we a) identified other relevant ontologies including psychological con- structs; b) studied the already available retrospective datasets from the participating clinical centers; and c) collected the relevant patient reported questionnaires used dur- ing the BOUNCE prospective study. Unfortunately, no existing ontology was able to cover the plethora and the diversity of the required psychological constructs (see [4] for a detailed review of relevant ontologies). In the next phase of conceptualization, we collected, analyzed and conceptualized the necessary psychological terms. Then interrelations between those were captured after constant interaction with clinical psychologists, exploiting their domain knowledge, based on relevant publications. Next, we implemented the ontology as an OWL ontology using Protégé. At the mo- ment, the ontology contains 310 classes, 106 object properties and 10 data properties. BPO is a module of the IMC Semantic Core Ontology, which also covers medical, clinical, biological, lifestyle constructs. As such, BPO adopts at the upper level the basic formal ontology (BFO), the upper level ontology upon which OBO Foundry on- 2 https://cbml-gitlab.ics.forth.gr/kondylak/the-bounce-psychological-ontology 3 tologies are built. The high-level classes of the BPO ontology, their main interconnec- tions with BFO and some examples, are shown in Fig. 2. For the development of the BPO it was crucial to be able to represent both the clinical reality and the various kinds of questionnaires of the clinical reality in the domain of our research. To achieve this goal our ontology includes a questionnaire class (BPO:Questionnaire), modelling the various questionnaires used in the project. For example, the BR23 and C30 EORTC quality of life questionnaires, require in- stantiation in some paper or electronic bearer (e.g., a printed questionnaire or a pdf file), but they are not particularly important for the existence of the questionnaire for which a particular bearer can instantiate it. Fig. 2. The Questionnaire class as subclass of the Information Object. As shown in the example, for assessing the perceived quality of life of a patient, there are multiple scales (three of them are shown in the example) that can be measured using BR23 and C30 questionnaires developed for cancer patients. The introduced classes and properties have been documented adding relevant useful information using multiple annotation properties, providing also the links to relevant publications. To evaluate an ontology there have been proposed several methods. Independent of the specific methods employed, two main lines of evaluation process are usually adopted, usually seen as complementing each other: The “glass box” or “component” evaluation and the “black box” or “task-based” evaluation, each one evaluating differ- ent ontology qualities. We evaluated the BPO ontology using the following glass box methods: Logical consistency: Logical soundness assesses the ontology for logical con- sistency, detecting contradictory statements. As the ontology has been developed using the Protégé tool, logical consistency, subsumption and satisfiability is automatically and constantly checked using the Pellet and the Hermit reasoners. 4 Common pitfalls in ontology development: Next, we employed an automated web- based tool, namely OOPS! [7], to automatically identify common errors in ontology development that could lead to modelling errors. We used the tool to evaluate the struc- tural, functional, and usability-profiling dimensions of our ontology as well as to eval- uate its consistency, completeness and conciseness. The results were very good with only some minor pitfalls noted, due to the fact that is not yet used by others, besides the creators. Application domain coverage & task orientation: Finally, validating domain cover- age is crucial to ensuring the usability of an ontology. The ontology is used by technol- ogy experts to generate (via mappings [8, 9]) a unified data access layer, effectively integrating all external, retrospective and the prospective data collected throughout the lifetime of the project. Then it is used by psychologists to query the available data and visualize them through an advanced data analysis tool [10] and by modelers to generate AI models [11, 12] on top effectively querying the homogenized, integrated data. As such, the ontology was able to answer all clinical and modeling questions involving such data, and to generate effective recommendation [13, 14] for cancer patients based on the predictions and their correlations with the available data. Conclusions In this paper, we present the BOUNCE psychological ontology, explaining the meth- odology followed for developing, presenting in short the modelling choices made, and providing a glimpse of the developed model. We also described the process followed for its glass box evaluation. BPO cannot be seen as a complete domain ontology, but rather an application ontology tailored to the needs of the BOUNCE platform. How- ever, it succeeds in modeling all psychological parameters used today for cancer pa- tients. Nevertheless, we have to note that ontologies are living artefacts and subject to con- tinuous change [15, 16]. As such, although ontology development within the BOUNCE project is complete, actually it will continue till the end of the project and beyond that, as long as there are people using it, continuously extending and adapting the model to fit their needs. We expect that, as we understand more on the psychological concepts under study, we will be able to refine classes and terms included in the ontology and to improve the mapping to the data sources. Acknowledgement This work has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 777167 (BOUNCE). 5 References 1. Crico, C., et al.: mHealth and telemedicine apps: in search of a common regulation. ecancer- medicalscience 12 (2018). 2. Kondylakis, H., et al.: Donor’s support tool: Enabling informed secondary use of patient’s bio- material and personal data. International journal of medical informatics 97: 282-292 (2017). 3. Kondylakis, H., Bucur, A.I., et al.: Patient empowerment for cancer patients through a novel ICT infrastructure. J. Biomed. Informatics 101: 103342 (2020). 4. The BOUNCE Consortium, D3.2 Initial Semantic Model (2017). 5. Uschold, M., Gruninger, M.: Ontologies: Principles, methods and applications. The Knowledge Engineering Review, 11(2) (1996). 6. Fernandez-Lopez M., Gomez-Perez A., Juristo N.: METHONTOLOGY: from Ontological Art towards Ontological Engineering, AAAI, pp. 33 – 40 (1997). 7. Poveda-Villalón, M., Gómez-Pérez, A., Carmen Suárez-Figueroa, M.: OOPS! (OntOlogy Pit- fall Scanner!): An On-line Tool for Ontology Evaluation. Int. J. Semantic Web Inf. Syst. 10(2): 7-34 (2014). 8. Minadakis, N., Marketakis, Y., Kondylakis, H. et al.: X3ML Framework: An Effective Suite for Supporting Data Mappings. EMF-CRM@TPDL, 1-12 (2015). 9. Kondylakis, H., Plexousakis, D.: Ontology Evolution: Assisting Query Migration. ER, 331-344 (2012). 10. Koumakis, L., et al.: A content-aware analytics framework for open health data. International Conference on Biomedical and Health Informatics. Springer, Singapore, (2017). 11. Manikis, G.C., Pat-Horenczyk, R., Fotiadis, D.I. et al.: Computational Modeling of Psycholog- ical Resilience Trajectories During Breast Cancer Treatment. BIBE, 423-427 (2019). 12. Kondylakis, H., Koumakis, L., Katehakis, D.G., et al.: Developing a Data Infrastructure for Enabling Breast Cancer Women to BOUNCE Back. CBMS, 652-657 (2019). 13. Stratigi, M., Kondylakis, H., Stefanidis, K.: Fairgrecs: Fair group recommendations by exploit- ing personal health information. International Conference on Database and Expert Systems Ap- plications. Springer, Cham, 2018. 14. Stratigi, M., Kondylakis, H., Stefanidis, K.: Fairness in group recommendations in the health domain. 2017 IEEE 33rd International Conference on Data Engineering (ICDE). IEEE, (2017). 15. Kondylakis, H., Flouris, G., Plexousakis, D.: Ontology and Schema Evolution in Data Integra- tion: Review and Assessment. OTM Conferences (2), 932-947, (2019). 16. Kondylakis, H., Plexousakis, D.: Ontology Evolution in Data Integration: Query Rewriting to the Rescue. ER, 393-401 (2011).