Matched Reassignment of the Ultrasound Data of Breast Lesions

Abstract: The goal of this thesis was to, from several randomly selected patients with diagnosed malignant and benign tumors, record optimal lambdas and respective Renyi entropies for each lambda, run a basic statistical analysis of the results and see if there is any significant difference between lambdas/Renyi entropies of malignant and benign lesions.\\ The results showed no significant difference.\\ The reassignment technique is a method that moves the signal energy to the center of gravity, giving higher energy concentration at the instantaneous frequency of the signal. The novel matched reassigned spectrogram method (MRS) was recently invented, with the goal to localize and classify transient functions of arbitrary shape. It was tailored for very short oscillating transients, which, theoretically, gives perfect reassignment localization to one single point in the time-frequency (TF) plane. Renyi entropy was used as concentration measure, where lambda is its minimizing parameter. MATLAB algorithm for matched reassigned spectrogram was run on the OASBUD dataset that contains the raw radio-frequency echoes of breast lesions, recorded in two scan planes, along with their respective regions of interest (ROI), and classification of malignancy of the lesion for each patient.