Regresar

Automatic breast tumor detection using cellular characteristics, multi-resolution analysis and neural networks

Abstract:

In this work a new method to detect breast tumor detection automatically from medical images has been proposed. The method is based on the information obtained form a multi-resolution analysis (MRA) in the HSV color space together with those obtained from a cell texture analysis. A noise reduction algorithm has also been developed, based on a monochromatic multi-resolution analysis using Mallat and Zhong's wavelet. The subsequent classification has been performed on a polynomial-fitting basis. The detection of breast tumors in tissues samples is accomplished by means of an artificial neural network, whose parameters come from a cellular characteristics analysis and those coming form the color multi-resolution analysis. This is part of a software tool for tumor detection and diagnosis of neoplasic diseases.