Avaliação de ruído em perfil de imagens mamográficas

Abstract

The success of breast cancer treatment depends on early detection, which can make the medical treatment more successful and less aggressive. The exam used to peform the early detection of this disease is the mammography. To have a correct diagnosis is necessary that the mammographic image has a good accuracy so that the health professional can identify structures that can evolve into a future breast cancer. These structures are identified through contrast difference in the image. During image acquisition occours the interaction between the x-ray beam and material, such interaction causes absorption and scattering phenomena that will contribute to the increase of image noise. The noise increases the amount of grayscale in the image, and can collaborate to reduce contrast, sharpness and quality. Noise cannot be completely remove from the image due to their random feature, but it can be attenuated. By smoothing out noise by filtering the signal, using low-pass filters, blurring and distortion can occur in the resulting images. By filtering a set of phantom PMMA (Polymethylmethacrylate) images, it was possible to trace the image profile and use different low-pass filters in the space domain to analyze which one results in a better noise reduction while maintaining good image quality. The image quality was evaluated through evaluation metrics such as SNR (Signal-to-Noise Ratio), PSNR (Peak Signal-to-Noise Ratio) and MSE (Mean Squared Ratio). To read the image, to draw the profile and to calculate the metrics was used an algorithm developed in Matlab. It was necessary to analyze the results qualitatively and quantitatively, because images that presented a good qualitative result, due to the average filter, presented worse quantitative results and this incoherence of the results comes from the fact that the Phanton has uniform density.