Skip to main content

Table 1 Significant features on BCDR database

From: A machine learning approach on multiscale texture analysis for breast microcalcification diagnosis

Normal/AbnormalBenign/Malignant
Embedded M.freq (%)Filter M.freq (%)Embedded M.freq (%)Filter M.freq (%)
(k ≤2) (k ≤6) (k ≤10) (k ≤26) 
# Interest Points100# Interest Points100Variance _LL2100Variance _LL1100
# Interest Corners100Kurtosis _HL299.80# Interest Corners100Skewness _LL1100
  # Interest Corners99.10Variance _LL199.90Entropy _LL1100
  Kurtosis _HL197.80RelSmoothness _LL299.90RelSmoothness _LL1100
  Kurtosis _LH176.40RelSmoothness _LL199.60Entropy _HL1100
  Kurtosis _LH261.90# Interest Points91.30Entropy _HH1100
  Variance _LH224.80Variance _HH177.70Kurtosis _HH1100
  RelSmoothness _LH221.90RelSmoothness _HH177.40Variance _LL2100
    Entropy _HH158.90Skewness _LL2100
    Entropy _HL144.80Entropy _LL2100
    Mean _HH141.20RelSmoothness _LL2100
      Kurtosis _LH2100
      Kurtosis _HL2100
      Kurtosis _HH2100
      # Interest Points100
      # Interest Corners100
      Entropy _LH199.20
      Entropy _LH298.60
      Entropy _HH297.80
      Kurtosis _HL197.10
      RelSmoothness _HH196.10
      Variance _HH188.80
      Skewness _HL276.30
      Mean _LL159.00
  1. The features whose occurrence in the first k positions of the rankings defined by the filter and embedded methods is significantly different from the case (p-value null model test ≤0.05) are reported. k is the number of features that maximizes the accuracy of normal vs abnormal and benign vs malignant classification problems