arch:knowledge-based-modeling-in-healthcare

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

arch:knowledge-based-modeling-in-healthcare [2026/08/29 07:53] – Bulk sync migration userarch:knowledge-based-modeling-in-healthcare [2026/08/29 07:59] (current) – external edit 127.0.0.1
Line 1: Line 1:
 +====== Practical knowledge-based modeling in healthcare ======
  
 +By Lyudmila Rezunik (lrezunik@edu.hse.ru)
 +
 +===== Introduction =====
 +
 +The healthcare field is vast and contains big volumes of domain knowledge that needs to be used effectively to improve the clinical decision-making as well as enhance patient care quality. Practical knowledge-based modeling is a one way to approach this problem. 
 +
 +What exactly is knowledge-based modeling and how one can create a knowledge model of their own? Basically, knowledge modelling is a process of gathering domain-specific data and translating it into data that is computer interpretable. The capturing and structuring of knowledge are important steps in creation of knowledge-based systems, the examples of which are Knowledge Management Systems, Knowledge Graphs, Machine Learning. 
 +
 +Knowledge can be represented in a variety of ways, depending on its type and thus different modeling techniques are applied for each case [1]. Antony Rhem, author of the book “UML for Developing Knowledge Management Systems” states that the knowledge can be classified in such a way:
 +
 +  * Declarative knowledge – knowledge of facts
 +  * Procedural Knowledge – knowledge of how to do things
 +  * Process Knowledge – knowledge contained in processes
 +  * Concept Knowledge – knowledge contained in concepts
 +  * Tacit Knowledge – knowledge contained within humans (expert knowledge) that cannot be easily articulated
 +  * Explicit Knowledge – knowledge contained in documents, computer programs, databases, etc., which can be easily articulated.
 +
 +Depending on the knowledge type a corresponding modeling technique can be used. Several of the most common ways to describe knowledge are ontologies, decision trees, concept maps.
 +
 +Knowledge-based modeling is already in use in different areas, some of the examples being: processes optimization, semantic parsing and question answering (in NLP), risk assessment (in finances). This essay will cover benefits and limitations of knowledge-based modeling, specifics regarding knowledge modeling in healthcare along with some examples of application of such models in the domain and discussion about possible future developments. 
 +
 +===== Benefits of knowledge-based modeling =====
 +
 +Since the healthcare sector started adopted technology into the working process, the volumes of data were significantly increasing, and the data has become more complex. Thus, data management and analytics tools appeared to be in demand, improving the workflow, accuracy, and efficiency. An example would be the use of machine learning, which is applied by various medical organizations in image processing, segmentation, computer diagnosis. The mentioned use cases use the data which can be easily handled: images and signals. 
 +
 +One of the benefits of knowledge-based modeling is that it provides an opportunity to work with unstructured data such as electronic health records. There have been studies that stated that knowledge graphs (which are developed using ontologies) are most suitable in this case, because of their capability to express heterogeneous knowledge in various domains [2]. However, knowledge graph may not be suitable for all use cases, one of them being the straightforward task of image classification. It would be impractical to represent the pixels of MRI or ultrasound images as separate entities within the knowledge base, which proves the point that different task requires a different method.
 +
 +Knowledge-based models can be applied (and are already applied) in many subdomains of healthcare. This will be covered later in the text, but there is need to mention that these models can consolidate vast amounts of data of different kind (medical guidelines, expert knowledge, patient-specific data, etc.) which is beneficial for decision-making systems. 
 +
 +===== Challenges and Limitations =====
 +
 +Data modeling comes with some challenges, especially if we are talking about health-related domains. One of the most common problems is data availability, which is connected to privacy and security concerns. Healthcare data has become more digitized, distributed. This and the use of smartphones and smart devices made it more affected to privacy breaches [3]. 
 +
 +Besides, the quality of this data also needs to be taken into consideration, because it might be incomplete, erroneous, which will obviously pose challenges for model accuracy. This can also be connected to interoperability issues, that is a complex task itself, because in the domain of healthcare there are different medical devices, health record systems that may operate different data formats. This problem can be partially avoided by the usage of DICOM, which is a standard for data transmission between medical devices.
 +
 +There are also not so obvious challenges such as ethical considerations and bias. Knowledge-based models are developed with use of historical data, and the quality of predictions or decisions depends on whether it contains biases or inequities. There we proved cases of algorithms that predict the health’s risks and identify need in care to be racially biased, underestimating the level for black patients [4]. Thus, data used for modeling should be transparent and eliminate all bias.
 +
 +===== Applications in Healthcare =====
 +
 +Knowledge-based modeling can be applied in different sub-domains of medicine and healthcare. In this section will be given a few different examples of such usage.
 +
 +Knowledge-based models can be created using ontologies, via such development frameworks as OWL API [5]. Ontologies can help define concepts, relationships, and axioms, enabling reasoning in a model. A use case for this kind of models can be decision-making systems. Such systems can be useful in case of improving efficiency and accuracy of prescription of treatment, establishing of diagnosis. Influence diagrams and decision trees can be used along as well [6].
 +
 +Knowledge graph is another popular example of knowledge-based model, which uses ontology as a framework. It can be applied in different areas, being firstly developed for a search engine (Google) know it is also applied in medicine. Figure 1 demonstrates an example of such a graph. Knowledge graphs have proven to be an effective solution for semantic integration of data acquired from different sources in different formats, ensuring flexibility [7].
 +
 +[[arch:_detail:arch:kg.png|]]
 +
 +Figure 1 – A sample knowledge graph in healthcare domain (Abu-Salih et al.) 
 +
 +===== Conclusion =====
 +
 +Practical knowledge-based modeling in the domain of healthcare has been experiencing a lot of developments in the past few years. With the growing popularity of applying AI to problem-solving, it should find more applications. Knowledge-based modeling has its own limitations but fits perfectly into healthcare domain with its heterogenous data of differently standardized formats. At the moment knowledge-based models are applied in a variety of subdomains: treatment prescription, diagnosis establishment, clinical workflow enhancement and medical expert systems. Healthcare industry has obviously benefited from this data modeling method. 
 +
 +Considering the number of research conducted know, it is likely that knowledge-based modeling will remain of interest to researchers in the domain. It is believed that knowledge graphs are an especially promising area of research, providing recommendations for practicing doctors and giving an understanding of diseases by finding dependencies between symptoms.
 +
 +===== References =====
 +
 +  - T. Rhem. Knowledge modeling Concepts, 2023. Available: [[https://knowledgemanagementdepot.com/2022/01/31/knowledge-modeling-concepts/|https://knowledgemanagementdepot.com/2022/01/31/knowledge-modeling-concepts/]]
 +  - G. G. Abdullayeva, N. O. Alishzade. Modeling the knowledge from healthcare systems for machine learning applications. ICP, 2020, pp. 69-79.
 +  - A. H. Seh, M. Zarour, M. Alenezi, A. K. Sarkar, A. Agrawal, R. Kumar, R. A. Khan. Healthcare Data Breaches: Insights and Implications. Healthcare (Basel), 2020, 8(2):133.
 +  - Z. Obermeyer, B. Powers, C. Vogeli, S. Mullainathan. Dissecting racial bias in an algorithm used to manage the health of populations. Science 366, 447-453 (2019).
 +  - OWL API. Owlcs. Available: [[http://owlcs.github.io/owlapi/|http://owlcs.github.io/owlapi/]]
 +  - A. K. Das, B. A. Ahmed, Y. Garten, J. I. Robin, M. K. Goldstein. Knowledge-Based Method for Building Patient Decision-Analytic Tools. AMIA Annu Symp Proc. 2006, 175-179.
 +  - B. Abu-Salih, M. AL-Qurishi, M. Alweshah, M. AL-Smadi, R. Alfayez, H. Saadeh. Healthcare knowledge graph construction: A systematic review of the state-of-the-art, open issues, and opportunities. Journal of Big Data, 2023, 10:81.
 +  - B. Abu-Salih et al. 2023, A sample healthcare KG, digital image, accessed 22 October 2023, [[https://www.researchgate.net/publication/371123523_Healthcare_knowledge_graph_construction_A_systematic_review_of_the_state-of-the-art_open_issues_and_opportunities|https://www.researchgate.net/publication/371123523_Healthcare_knowledge_graph_construction_A_systematic_review_of_the_state-of-the-art_open_issues_and_opportunities]]