Generally, Bayesian networks are constructed either from the available information or starting from a naïve Bayes. In the medical domain, some systems refine Bayesian network manually created by domain experts. However, existing techniques verify the relation of a node with every other node in the network. In our previous work, we define a Refinement algorithm that verifies the relation of a node only with the set of its independent nodes using Markov Assumption. In this work, we did propose Extension of Refinement Algorithm that uses both Markov Blanket and Markov Assumption to find the list of independent nodes and adhere to the property of considering minimal updates to the original network and proves that less number of comparisons is needed to find the best network structure.