Abstract:
The Healthcare domain is a very sensitive domain where the direct stakeholders are patients
in the public, which makes a healthcare system directly dealing with patients’ lives. There
could be heard of several cases in a year where it leads to critical damage or even loss of life,
because of misdiagnosis or delays in the diagnosis process or the delay or ignorance of new
treatment methods. A clinical decision support system facilitating support for diagnosis and
therapeutic decisions could greatly help healthcare professionals when identifying diseases by
going through patients’ biometrics, life cycle, and symptoms, and when deciding on necessary
clinical tests to arrive at confirmation of the diagnosis and decide on treatment methods. The
main focus of this study is mapping the real-world diagnosis process to a digitalized system.
Senior clinicians use their own experience to derive medical diagnoses accurately. This study
proposes an architecture for an evidence-based clinical decision support system, where the
system infers knowledge from past knowledge, using machine learning algorithms, and use
for future predictions, which could infer and use the medical incidents of the past for future
diagnosis, just like experienced doctors. In a practical scenario, diagnosis of disease happens
step by step, going through several stages starting from an initial level and digging deeper. To
incorporate this behavior, a layered knowledge modeling system is proposed with an ensemble
classifier of Random Forest classifier, Support Vector Machine, and Naïve Bayes classifier,
and organized into a tree structure based on disease classification hierarchy. Additionally, the
proposed system provides feedback and suggestions for clinical tests using feature selection,
and a rationale for the diagnosis derived by incorporating explainable machine learning
concepts.
Citation:
Dayathilake, K.C.V. (2022). Effective architecture for clinical decision support system [Master's theses, University of Moratuwa]. Institutional Repository University of Moratuwa. http://dl.lib.uom.lk/handle/123/21863