According to "Global Tuberculosis report 2021" by World Health Organisation (WHO) [16], Tuberculosis (TB) is one of the leading causes of death around the globe and can be communicated easily. Statistics [16] say, that there are estimated 5.8 million people affected by TB all over the world in year 2020 that were reported, out of approximately 10 million people who developed TB in 2020 and 1.5 million deaths were reported in the same year due to TB.
India has been worst affected by TB in recent year as per WHO. In 2020, 2.59 million people were affected by TB in India and half a million people died because of it which accounts to 34% of the global value[16]. The main cause of TB epidemic is illiteracy, malnutrition and unavailability of proper medication. The TB cases in children have also increased. Around 146,335 children died of TB in 2020. Also during the year 2020, there were approximately 49,679 multi-drug resistant TB patients among the notified pulmonary TB cases. According to a WHO report, about 66 million lives of TB patients were saved between 2000-2020 through appropriate TB diagnosis and proper treatment.
This research paper focuses on use of “Probabilistic Web Ontology Language (PR -OWL)” which is an extension of "Web ontology language (OWL)" to create a decision support system for TB treatment.
The word Ontology is defined as the study of what exists (the entities in reality and their interrelationships). Ontology provides a clear, shared understanding of a domain of interest with all necessary information in a formal and machine interpretable model.
"An ontology is a formal, explicit specification of a shared conceptualization" Tom Gruber [15].
The representation of a knowledgebase in the form of unambiguous terminology definitions in a machine readable form have a certain degree of computational semantics which is globally accepted. This shared conceptualization makes a shared understanding of a domain of interest. However, ontology lacks in uncertainty reasoning. And many real world applications are structured in such a way that there is always a chance of having uncertain knowledge. Probabilistic Ontology is an extension to Web Ontology Language (OWL) incorporating statistical data for uncertainty representation.
"Probabilistic ontology is used for describing knowledge about a domain and the uncertainty associated with that knowledge in a principled, structured, and sharable way [7]."
Probabilistic ontology is an upper ontology which is used by various probabilistic systems and is capable of representing very complex models. Multi Entity Bayesian Network (MEBN)[9] implements probabilistic ontologies by using First Order Bayesian Logic to perform efficient reasoning with the uncertain knowledge. Probabilistic ontology combines expressive power of First Order Logic (FOL) and reasoning ability of Bayesian inference. The exibility and representation power of probabilistic ontology are its greatest advantages. This work proposes a probabilistic ontology model for perceiving the early research and clinical trials in real time for the treatment of disease Tuberculosis (TB) in patients, which is a leading source of death around the world. As per the "Global Tuberculosis report 2017"[1], the estimated Tuberculosis incidence in India was around 2.8 million which is about a quarter of the world’s tuberculosis cases. The highest cases of both TB and MDR TB are found in India. Also India stands second highest in number of estimated HIV associated TB cases.This work presents a knowledge enabled TB management and it's control with quantitative reasoning capabilities in order to improve TB incident rates, proper patient and programme management. The proposed work uses Multi Entity Bayesian Network (MEBN) [13] to design a Decision Support System (DSS) [11] .Using probabilistic ontology enables standardization of tuberculosis treatment process and greatly increases its reusability. Also, a systematic analysis of knowledge can be done using the ontological framework and computerised processing and decision making becomes easier. Incorporating the probabilistic analysis tool provides reasoning with uncertainty associated with the domain knowledge. UsingMulti Entity Bayesian Network facilitates addition of new knowledge and data,and thus increases the applicability of the model.
The rest of the paper is organized as follows. Section 2 gives a brief description of the problem domain, that is, diagnosis and treatment of Tuberculosis based on RNTCP guidelines implemented using Multi Entity Bayesian Network with the use of UnbBayes reasoning tool. Section 3 describes about past research / work done in "Multi Entity Bayesian Network based DSS". Section 4 describes the proposed methodology for designing a decision support system for TB Control and management program. Section 5 describes the implementation of Tuberculosis treatment regime using Multi Entity Bayesian Network. Section 6 discusses the work presented in this dissertation along with the limitations and future work of the research followed by conclusion in Section 7.