Zm. Huo et al., Effect of dominant features on neural network performance in the classification of mammographic lesions, PHYS MED BI, 44(10), 1999, pp. 2579-2595
Two different classifiers, an artificial neural network (ANN) and a hybrid
system tone step rule-based method followed by an artificial neural network
) have been investigated to merge computer-extracted features in the task o
f differentiating between malignant and benign masses. A database consistin
g of 65 cases (38 malignant and 26 benign) was used in the study. A total o
f four computer-extracted features-spiculation, margin sharpness and two de
nsity-related measures-was used to characterize these masses. Results from
our previous study showed that the hybrid system performed better than the
ANN classifier. In our current study, to understand the difference between
the two classifiers, we investigated their learning and decision-making pro
cesses by studying the relationships between the input features and the out
puts. A correlation study showed that the outputs from the ANN-alone method
correlated strongly with one of the input features (spiculation), yielding
a correlation coefficient of 0.91, whereas the correlation coefficients (a
bsolute value) for the other features ranged from 0.19 to 0.40. This strong
correlation between the ANN output and spiculation measure indicates that
the learning and decision-making processes of the ANN-alone method were dom
inated by the spiculation measure. Three-dimensional plots of the computer
output as functions of the input features demonstrate that the ANN-alone me
thod did not learn as effectively as the hybrid system in differentiating n
on-spiculated malignant masses from benign masses, thus resulting in an inf
erior performance at the high sensitivity levels. We found that with a limi
ted database it is detrimental for an ANN to learn the significance of othe
r features in the presence of a dominant feature. The hybrid system, which
initially applied a rule concerning the value of the spiculation measure pr
ior to employing an ANN, prevents over-learning from the dominant feature a
nd performed better than the ANN-alone method in merging the computer-extra
cted features into a correct diagnosis regarding the malignancy of the mass
es.