Initial results of training neural networks to detect breast cancer using evolutionary programming

摘要

Artificaial neural networks are applied to the problem of detecting breast cancer from radiographic features and patient age. Evolutionary programming is used to train neural networks based on sigmoid or Gaussian kernel functions. Preliminary results on 96 biopsy-proven cases (62 malignant, 34 benign) indicate that a reasonable probability of detecting malignancies can be achieved using simple neural architectures. The features appear to be more amenable to discrimination by partitioning functions than to clus- tering functions, although final analysis remains for larger sample sizes.