Ten lectures on wavelets
Neural Computation
Bayesian regularization and pruning using a Laplace prior
Neural Computation
| Hi-index | 0.00 |
Optical scattering spectra obtained in the clinical trials of breast cancer diagnostic system with a minimal invasive probe were analyzed for the purpose to detect in the dataflow the segments corresponding to malignant tissues. Large amount of information acquired in each procedure, fuzziness in criteria of 'cancer' family membership and data noisiness make neural networks to be an attractive tool for analysis of these data. To define the dividing rule between 'cancer' and 'non-cancer' spectral families a three-layer perceptron was applied. In the process of perceptron learning the back propagation method was used to minimize the learning error. Regularization was done using the Bayesian approach. The learning sample was formed by the experts. Much attention was paid on the spectra of the tissues with high blood content. Another perceptron was learnt exceptionally on such spectra. Independent processing of two collecting channels and including into the model their voting allowed improving the prediction ability. The model was tested over the sample of 29 'cancer' and 29 'non-cancer' cases and demonstrated total separation. Comparing with the traditional breast cancer diagnostics the suggested method has advantages of automated on-line diagnosing and minimal tissue destruction. For further development of the mathematical model more data are needed to prove its reliability.