ABSTRACT
Mammography is a radiographic technique used for the detection of breast lesions. The analysis of the digital image normally requires a previous application of filters as a preprocessing step to reduce the noise level of the image, while preserving important details to carry out a suitable diagnostic. In the literature, there are a large amount of denoising techniques applied to different medical images. In this work we have studied the performance of a diffusive filter with a stopping condition based on the statistical concept of negentropy, applied to denoise mammographic images. The negentropy has been succesfully prove with other denoising methods as independent component analysis by the authors in [1]. We have evaluated the final image quality obtained, measuring a root mean squared error between the noise-free initial image and the final restored image and compared the results obtained by this diffusive filter with those obtained by an adaptative non-linear Wiener filter.
Subject(s)
Mammography/instrumentation , Radiographic Image Interpretation, Computer-Assisted/methods , Algorithms , Artifacts , Breast Diseases/diagnosis , Breast Diseases/diagnostic imaging , Breast Neoplasms/diagnosis , Breast Neoplasms/diagnostic imaging , Diffusion , Entropy , Humans , Mammography/methods , Models, Statistical , Models, Theoretical , Normal Distribution , Radiographic Image Enhancement/instrumentation , Radiographic Image Interpretation, Computer-Assisted/instrumentation , Reproducibility of Results , Subtraction TechniqueABSTRACT
Diffusive iterative filtering may be used to restore mammographic noisy images. Assuming that images are corrupted by additive gaussian noise, the negentropy as a statistical measure of the nongaussianity, can be used to estimate the amount of noise removed from noisy images. In particular, the evolution of the negentropy with the iterations is used to introduce a stopping criterion to end the diffusive filtering process. Because of computational complexity of the negentropy function, it is estimated by using an approximation of the negentropy introduced by Hyvärinen in the context of independent component analysis.