We then ran the hill climbing feature selection algorithm [10] to investigate if further classification improvements were possible. The algorithm began with the double support symmetry, which had the highest ROC area i.e., 0.727 and the remaining features were sequentially combined to obtain the maximum LOO accuracy (Figure 1). It was found that only two features were required to produce the maximum LOO accuracy of 97.1 % (Table 2) after which no further improvement in accuracy was obtained. These two features were the double and single support symmetry indexes, suggesting that symmetry measures had sufficient discriminative power to accurately classify the two groups.