Bayesian Kernel Methods: Applications in Medical Diagnosis Decision-Making Processes (A Case Study)

Introduction

Many people all around the world die due to error in healthcare systems. In healthcare industry, several strategies have been proposed by various professionals like IT adoption, collaboration among various disciplines etc. to design medical decision support systems, which help clinicians for medical decision making. From the last three decades, the applications of Bayesian approaches have grown at an exponential pace, but research in this domain has developed very slowly in first one hand half decade. The major reason for that is to generate Bayesian networks for practical and analytical purposes are quite difficult. Therefore, generating Bayesian networks was a challenge for researchers and this made inaccessible to vast community of scientist for their applicability. Now, Bayesian and its native forms are generated by the help of computers, which address the vast community of researchers to became a tool and have wings spread over many disciplines such as computer science, logic, information theory, probability theory, statistics, machine learning etc. and can be utilized in almost all the disciplines for the purpose of application. Bayesian based models are mainly used to answer the queries for the variables, their relationship to compute the evidence for the variables in complex situations and are considered as graphical probabilistic models. Decisions making under the state of uncertainty are the typical most problems in medical decision making and Bayesian kernel methods are very promising in tackling such situations problems and becomes more rational as compared to the conventional models, the assumptions made are more obvious and thus making the decision making easy and defendable. The analytical approach of the theory became a boon for the researchers, which flavoured with machine learning algorithms to grow beyond their domain. The traditional statistical techniques such as regressions have explanatory purposes and have applications in simulation and optimization, epidemiology and social sciences. The foundation of Bayesian approach lies from the rule for updating probabilities, given by Rev. Thomas Bayes (1702-1761) as Bayes’ theorem. He addresses both discrete and continuous probability distributions by means of conditional and marginal probabilities. Bayes’ theorem is used for the calculation of inverse probability and is based on the apriori and posterior probability measures and provides uncontroversial result in the field of probability but its applications fall under controversy for more than two decades. This chapter is an introduction to the modernized Bayesian approach which leads to algorithms within the frame work of risk minimization and giving us a new insight into kernel algorithms. There are many ways we could take to motivate the readers for using Bayesian kernel methods in medical diagnosis, which is based on the assumption that all the quantities of interest are governed by probability distributions. Optimized results can be obtained in Bayesian theory by evaluating these probabilities together and combines prior knowledge with the observed data and is the sole theory in medical diagnosis which accommodates hypothesis.