Statistical two-dimensional correlation coefficient mapping of simulated tissue phantom data: Boundary determination in tissue classification for cancer diagnosis

Statistical correlation coefficient mapping has proven to be a useful technique in tissue classification for cancer diagnosis. The classification is achieved by comparing the correlation coefficients for an unknown to a set of selected tissue samples with known pathological conditions. Currently, the correlation coefficient threshold in the classification is empirically determined. In this paper, boundaries of statistical significance between different tissue pathological conditions are established through Bayesian analysis on the Fisher's z-transformed Pearson's correlation coefficients between tissue samples. Moreover, probability values are provided in assigning a tissue sample to a specific tissue clinical condition, which is more appreciable in clinical practices. The methodology is examined with a simulated tissue-phantom data set, yielding satisfactory diagnostic results.