TUMOR-DETECTION IN NONSTATIONARY BACKGROUNDS

We introduce two detectors which we use to locate simulated tumors of fixed size in clinical gamma-ray images. The first method was conceive d when it was observed that small tumors possess an identifiable signa ture in curvature feature space, where ''curvature'' is the local curv ature of the image data when viewed as a relief map. Computed curvatur e values are mapped to a normalized significance space using a windowe d t-statistic. The resulting test statistic is thresholded at a chosen level of significance to give a positive detection. Nonuniform anatom ic background activity is effectively suppressed. The second detector is an adaptive prewhitening matched filter, which uses a form of prepr ocessing known as statistical scaling to adaptively prewhiten the back ground. Tests are performed using simulated Gaussian-shaped tumors sup erimposed on twelve clinical gamma-ray images. When the tumors to be d etected are small-less than 3 pixels in diameter - the curvature detec tor out-performs the matched filter in true positive/false positive te sts. A mean true positive rate of 95% at one false positive per image is achieved when the local signal-to-noise ratio of the tumor-backgrou nd is greater-than-or-equal-to 2. At larger tumor sizes the best perfo rmance is displayed by a different form of matched filter, namely the statistical correlation function proposed by Pratt.