A comparison on performance of data mining algorithms in classification of whewellite and weddellite urinary human calculi images color using different descriptors

摘要

The aim of this paper is to investigate the best correct rate of data mining supervised algorithms such as k-nearest neighbor (KNN), support vector machines (SVM), artificial neural networks (ANN), and PLS-DA. The significant information from the image is extracted using Zernike-, Gist-, and Centrist descriptor models in order to classify the two most frequent constituents of urinary stones using their colored images, calcium oxalate monohydrate and calcium oxalate dihydrate. The first class that contains images of calcium oxalate monohydrate (CaC2O4, H2O: whewellite (Wh)) crystal type. The second class contains the images of calcium oxalate dihydrate (CaC2O4, 2H2O weddellite (Wd)) crystal type. The results showed that the PLS-DA model coupled to Gist descriptor was superior to SVM-, kNN and KNN models in prediction using different descriptors. Using PLS-DA, SVM- KNN and ANN model the correct prediction rate reached 100%, 94.74%-, and 80.22% and 94.15% respectively.