Mammogram analysis based on wavelets and fractals

Although improvements in mammogram resolution and film contrast have occurred in past decade, the correct diagnosis of mammograms is not easy. The main reason is the minor difference in X-ray attenuation between normal glandular tissues and malignant diseases. Considering the texture difference of breast tissues, they can be distinguished by texture property of mammograms. Being random and statistically self-similar, fractional Brownian motion (fBm) can be appropriately viewed as a model of the mammogram. The index H of fBm that only depends on fractal dimension D, is an identification of image complexity or roughness. Both wavelets and fractals take scale as their main property-they must have an inner relationship. Here, the authors describe an orthonormal wavelet transform to estimate index H that closely relates with the texture complexity of breast tissues. Some experiments described given, and the experimental results show that there exists an obvious fractal difference between the malignant tumor and the normal tissue