Supervised localization of cell nuclei on TMA images

We consider the problem of localizing renal cancer cell nuclei in Tissue Micro Array (TMA) images. We address this problem in three steps. An initial image processing-based procedure finds potential candidate nuclei, while the subsequent phase employs a trained classifier to prune the candidate cell nuclei found in the first. A third phase is then used to perform a clustering of the positive classified blobs. In this work, we study cases when the second step is attained by extracting fixed size patches centred on the candidates, and representing these images with pixel-intensity histograms or related pair-wise distances (dissimilarities). Our results, based on a Parzen classifier in the histogram feature space, show that the proposed procedure attains an optimal Fl-measure 0/0.9152 in localizing cell nuclei, providing state-of-the-art performance.