Fuzzy connectedness image co-segmentation for hybrid PET/MRI and PET/CT scans
Lecture Notes in Computational Vision and Biomechanics, volume 22, pages 15–24; Workshop of Computational Methods for Molecular Imaging, MICCAI 2014
Abstract
An abstract is not available in the lab archive. View the paper at its source.
This paper presents a three-dimensional co-segmentation method for identifying and delineating lesions in hybrid PET/MRI and PET/CT scans. It combines modality-specific visibility weighting with fuzzy connectedness and compares the segmentations with radiologist-provided reference labels.
Summary
The study describes a three-dimensional computer-aided method for co-segmenting tumors or lesions across PET/MRI and PET/CT images. It uses high-uptake regions in PET images to select seed points, then applies a fuzzy connectedness (FC) segmentation framework with modality-specific visibility weighting and an affinity function intended to use complementary information from the functional and anatomical images. The authors also use iterative relative fuzzy connectedness with automatically identified background seeds.
The source record reports comparison against radiologist-provided reference labels and a mean Dice similarity coefficient (DSC) of 85%. This result is reported for the study’s experiments; it does not establish prospective clinical validation or clinical readiness. The source record identifies the work as published in 2015 in Lecture Notes in Computational Vision and Biomechanics, volume 22, pages 15–24, and associates it with the MICCAI 2014 Workshop of Computational Methods for Molecular Imaging.
Cite this work
Ziyue Xu, Ulas Bagci, Jayaram K. Udupa, Daniel J. Mollura (2015). Fuzzy connectedness image co-segmentation for hybrid PET/MRI and PET/CT scans. Lecture Notes in Computational Vision and Biomechanics, volume 22, pages 15–24; Workshop of Computational Methods for Molecular Imaging, MICCAI 2014 https://doi.org/10.1007/978-3-319-18431-9_2
BibTeX