The purpose of this study was to evaluate the performance of neural network
model self-organizing maps (SOM) in the classification of benign and malig
nant sonographic breast lesions. A total of 243 breast tumors (82 malignant
and 161 benign) were retrospectively evaluated. When a sonogram was perfor
med, the analog video signal was captured to obtain a digitized sonographic
image. The physician selected the region of interest in the sonography, An
SOM model using 24 autocorrelation texture features classified the tumor a
s benign or malignant. In the experiment, cases were sampled with k-fold cr
oss-validation (k = 10) to evaluate the performance using receiver operatin
g characteristic (ROC) curves,The ROC area index for the proposed SOM syste
m is 0.9357 +/- 0.0152, the accuracy is 85.6%, the sensitivity is 97.6%, th
e specificity is 79.5%, the positive predictive value is 70.8%, and the neg
ative predictive value is 98.5%. This computer-aided diagnosis system can p
rovide a useful tool and its high negative predictive value could potential
ly help avert benign biopsies. (C) 2000 World Federation for Ultrasound in
Medicine & Biology.