Abstract:
Magnetic Resonance Mammography (MRM), accepted by for use as asupplemental tool to mammography in 1991, provides detailed information about very small lesions that X-ray mammography and ultrasound often cannot detect. Women who are at increased risk for developing cancer, or those who have completed breast conserving "lumpectomy", young women with dense breasts or those with a greatamount of DCIS (ductal carcinoma in situ) are good candidates for MRM. Resolution of the breast imaging is important for improving differentiation between benign and malignant lesions and for refining treatment strategy. Inhomogeneity of the static magnetic field or secondary magnetic field and nonuniformity of the receiver coil have adverse effects on resolution. A number of methods have been proposed to minimize these effects. In this thesis work we present anovel improved homomorphic filtering method to minimize artifacts caused by these inhomogeneities. Unlike other homomorphic filtering methods, we apply a tissue maskto eliminate filter artifacts, and then apply low-pass filtering to estimate the bias field.Restored image is obtained by the difference of the original image and the estimatedbias field. A frequency range is defined and a number of bias fields and restored imagesare estimated for each image. Entropy minimization is used to define an optimum cutoff frequency of the low-pass filter. This results in a fast, user independent, nonparametricalgorithm. The method is demonstrated on various breast images from different patients. A performance evaluation method is also defined for quantitative measurement.|Keywords: Magnetic resonance mammography, field inhomogeneity; homomorphic filtering, entropy minimization.
