MICCAI 2016 Daily - Wednesday

Mattia Gentil presented this paper at the Interactive Medical Image Computing (IMIC) workshop held on Monday. Mattia presented on behalf of Danna Gurari , from the Department of Computer Science at the University of Texas at Austin. The presented work focused on providing accurate and inexpensive segmentation of live cells in microscopy images. The proposed method, that was named SAVE , involves both algorithms and crowd workers in order to provide the best performance possible on the produced annotations. In more detail, instead of collecting a single annotation from either an algorithm or a crowd worker, a pipeline is realized in order to output the desired annotation. First, given the raw grayscale image of a cell, 5 annotations are collected from either 5 different algorithms or from 5 crowd workers. Then either the computer or another set of 5 human annotators is asked to pick the most accurate segmentation from the ones obtained in the first step. Finally majority voting is applied in order to get a single binary image as the output. The results on their diverse and challenging datasets show that SAVE is able to outperform the standalone segmentation algorithms and that it managed to consistently provide expert-quality annotations . From the IMIC workshop 7 MICCAI Daily: Wednesday Mixing Crowd and Algorithm Efforts to Segment Objects in Biomedical Images