Skip to main content
. 2016 Jan 13;6:19256. doi: 10.1038/srep19256

Figure 8. Venn diagram for the results obtained by different Feature Selection methods (from the R library VennDiagram).

Figure 8

Comparing FSMKL, SVM-RFE, GA-SVM and PSO-SVM methods by checking the overlaps of their selected textural features (Top features are in bold character). Labels for the Histogram features: Perc. = percentile derived from the image histogram, Mean and Kurtosis. Labels for the absolute gradient features: Gr. = absolute gradient parameters (kurtosis, mean, skewness, variance and and percentage of pixels with nonzero gradient). Labels for the run-length matrix features: GLevNonU = grey-level non-uniformity, LongREmph = long-run emphasis. Calculated for vertical, horizontal, 45-degree and 135-degree directions. Labels for the co-occurrence matrix features: values in parenthesis represent coordinates, containing information about distance and direction between pixels (InvDfMom = inverse difference moment, AngScMom = angular second moment, DifEntrp = difference entropy, SumEntrp = sum entropy, Entropy, Correlat = correlation, SumVarnc = sum variance, DifVarnc = difference variance, SumOfSqs = sum of squares, SumAverg = sum average). Labels for the Autoregressive model features: Theta and Sigma. Labels for the Wavelet features: WavEnLH = energy of wavelet coefficients in subband LH, WavEnLL = energy of wavelet coefficients in subband LL.