In-vivo tissue identification using ambient ionization mass spectrometry-comparison of different multivariate classification workflows

摘要

Linear algorithms perform better in the classification of healthy and canceroustissue compared to non-linear algorithms. OPLS-DA and PCA+LDA had the highest classification rate, while K star and IBK (modified K nearest neighbour) are not suitible for this analysis. PCA+LDA is the only algorithm performing always above 90%!