International Journal of Computer Vision
Construction of neuroanatomical shape complex atlas from 3D brain MRI
MICCAI'10 Proceedings of the 13th international conference on Medical image computing and computer-assisted intervention: Part III
Logarithm odds maps for shape representation
MICCAI'06 Proceedings of the 9th international conference on Medical Image Computing and Computer-Assisted Intervention - Volume Part II
Thoracic abnormality detection with data adaptive structure estimation
MICCAI'12 Proceedings of the 15th international conference on Medical Image Computing and Computer-Assisted Intervention - Volume Part I
Hi-index | 0.01 |
This paper presents a novel segmentation algorithm which automatically learns the combination of weak segmenters and builds a strong one based on the assumption that the locally weighted combination varies w.r.t. both the weak segmenters and the training images. We learn the weighted combination during the training stage using a discriminative spatial regularization which depends on training set labels. A closed form solution to the cost function is derived for this approach. In the testing stage, a sparse regularization scheme is imposed to avoid overfitting. To the best of our knowledge, such a segmentation technique has never been reported in literature and we empirically show that it significantly improves on the performances of the weak segmenters. After showcasing the performance of the algorithm in the context of atlas-based segmentation, we present comparisons to the existing weak segmenter combination strategies on a hippocampal data set.