Automated detection of prostate cancer nuclei

In this work, a cross-disciplinary approach was presented for the solution of a biomedicai pattern recognition problem: cancer diagnosis. Specifically, prostate specimen images were analyzed. Images were acquired by a microscope, video camera and a digitizing board and stored on a computer. The software required to indicate the areas of interest of the tissue image to the pathologist, was developed on the object-oriented basis. User interface was designed for the required functions. Discriminatory power of shape feature were evaluated statistically. The sensitivity and specificity ratios for different threshold levels of the feature set were calculated.