Tree-structured nonlinear filter and wavelet transform for microcalcification segmentation in mammography

摘要: The development of an extensive array algorithms for both image enhancement and feature extraction microcalcification cluster detection is reported. Specific emphasis placed on detail preservation automatic or operator independent methods to enhance the sensitivity specificity that should allow standardization breast screening procedures. Image include novel tree structured non-linear filters with fixed parameters adaptive order statistic designed further improve preservation. Novel developed two channel wavelet transform three quadrature mirror filter banks multiresolution decomposition reconstruction specifically tailored extract MCC's. These were evaluated using fifteen representative digitized mammograms where similar (true positive (TP) rate 100%) (0.1 - 0.2 average false (FP) MCC's/image) was observed but varying degrees important characterization step proved be very critical minimize noise associated FP rates MCC's individual microcalcifications.