A evolução temporal dos expoentes de Hurst em imagens interferométricas de um carcinoma mamário

Abstract

This research aimed to estimate and verify whether the behavior of directional Hurst exponents over time with the use of undetermined wavelet transform, in 128 interferometric images of a canine anaplastic breast carcinoma, obtained at regular time intervals, through the dynamic biospekle method, may differentiate sub-images of cancerous tissue from healthy ones. To achieve this purpose, sub-images that depict only cancerous tissues and sub-images that depict only healthy tissues were extracted applying the multi-resolution analysis methodology with four levels of resolution and use of the Daubechies wavelet with eight null moments, as proposed in a previous article. To estimate the Hurst exponents, the mean of the squares of the wavelet coefficients, Mid-energies and the median of the square of the wavelet coefficients (MEDL) were used and their effects were verified in the time series of Hurst exponents. It was concluded, with the prospective study, that the temporal analysis of the Hurst exponents can be an alternative tool to differentiate images of healthy and cancerous tissue. There was a variation in the directional Hurst exponents, over time, around an average value, both for cancerous regions and for healthy regions. The data suggest that the estimate of Hurst’s directional exponent in the horizontal direction, obtained via MEDL, may constitute an alternative to differentiate sub-images of cancerous tissue from healthy ones, as they present less variance and better discrimination.