search railabs.duhs.duke.edu.

Papers Published

1. Lo, J.Y. and Baker, J.A. and Koruguth, P.J. and Floyd, C.E., Jr., Computer-aided diagnosis of mammography: artificial neural networks for optimized merging of standardized BIRADS features, WCNN '95. World Congress on Neural Networks. 1995 International Neural Network Society Annual Meeting, vol. vol.2 (1995), pp. 885 - 8 .
   (last updated on 2007/04/15)

   Abstract:
   An artificial neural network was developed for computer aided diagnosis in mammography, using an optimally minimized number of inputs from a standardized lexicon for mammographic features. A three layer backpropagation neural network merged eleven inputs (ten radiographic findings extracted by radiologists plus age) to predict biopsy outcome as its output. Each input feature was ranked by importance, as determined by the reduction of Az when that feature was excluded and the network retrained. Once ranked, the input features were discarded in order from least to most important until performance was significantly degraded, resulting in an optimized subset of features. The neural network trained on all eleven input features performed with an Az of 0.84±0.03, compared to experienced radiologists' Az of 0.85±0.03. The difference in Az was not statistically significant (p=0.54). The network continued to perform well given as few as two inputs: age and mass margin

   Keywords:
   backpropagation;diagnostic radiography;feature extraction;medical image processing;neural nets;