Quantification of visual diagnostic heuristics during simulated pathology diagnosis

摘要

We introduce a method to quantify visual diagnostic clues that are considered during simulated pathology diagnosis. We extend Apriori method of association rule mining with custom Perl extensions to collect sets of diagnostic clues that show an improvement in diagnostic accuracy. We developed an information gain measure based on Kullback-Leibler divergence in an attempt to gauge the impact of additional factors in diagnostic process.