Proposta de metodologia para realce de contraste em imagens de mamas densas utilizando decomposição multiescala com transformada discreta wavelet

Abstract

Breast cancer is the second most frequent type of cancer in women, thus, it is considered a global issue. Breast density is directly related to the probability of developing breast cancer, mainly due to the difficulty detecting lesions in dense breasts. Mammography is the main exam to screen breast cancer allowing early detection and improving the chances of curing the disease. In the past years, a novel imaging technique has emerged aiming at improving early detection of breast cancer: the digital breast tomosynthesis technique. However, this technology is relatively unknown and not of easy access. Digital image processing, on the other hand, is a strong asset for the attempt of improving image quality. Therefore, the aim of this thesis, besides being a proposal of new methodology for contrast enhancement in dense mammographic images, is creating and implementing a new global metric to calculate contrast. This new methodology consisted of applying the Contrast-limited adaptive histogram equalization (CLAHE) technique to the sub-image generated from decomposing the approximation coefficients of the discrete wavelet transform. The technique was quantitatively and qualitatively validated. When testing real mammographic images of dense breasts, measures of contrast such as: variance, entropy, and measure of enhancement (EME), were calculated and reflected an increase when processed with the proposed methodology, compared with their corresponding values in original images. For the Carneiro Contrast Index (CCI, in Portuguese, Índice Carneiro de Contraste) created, applying CLAHE with a 15x15 window for the approximation coefficients has shown the best results, a 47% improvement in contrast, compared with original images. Hence, it was possible to propose a technique which is both optimized and simple to apply, especially for dense breasts, allowing contrast enhancement in breast structures and contributing to the early detection of breast cancer.