A Generative Model for Image Segmentation Based on Label Fusion

By Mert Rory Sabuncu, B. T. T. Yeo, K. V. Leemput, Bruce Fischl, Polina Golland. IEEE Transactions on Medical Imaging, 2010

517 citations recorded by the source corpus.

We propose a nonparametric, probabilistic model for the automatic segmentation of medical images, given a training set of images and corresponding label maps. The resulting inference algorithms rely on pairwise registrations between the test image and individual training images. The training labels are then transferred to the test image and fused to compute the final segmentation of the test subject. Such label fusion methods have been shown to yield accurate segmentation, since the use of multiple registrations captures greater inter-subject anatomical variability and improves robustness against occasional registration failures. To the best of our knowledge, this manuscript presents the first comprehensive probabilistic framework that rigorously motivates label fusion as a segmentation approach. The proposed framework allows us to compare different label fusion algorithms theoretically and practically. In particular, recent label fusion or multiatlas segmentation algorithms are interpreted as special cases of our framework. We conduct two sets of experiments to validate the proposed methods. In the first set of experiments, we use 39 brain MRI scans - with manually segmented white matter, cerebral cortex, ventricles and subcortical structures - to compare different label fusion algorithms and the widely-used FreeSurfer whole-brain segmentation tool. Our results indicate that the proposed framework yields more accurate segmentation than FreeSurfer and previous label fusion algorithms. In a second experiment, we use brain MRI scans of 282 subjects to demonstrate that the proposed segmentation tool is sufficiently sensitive to robustly detect hippocampal volume changes in a study of aging and Alzheimer's Disease.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/093096bd4694ef1d3d87ebf34fa7b3aa368ffa25

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Data Augmentation Using Learned Transformations for One-Shot Medical Image Segmentation (2019, background): “When multiple atlases are available, they are each aligned to a target volume, and the warped atlas labels are fused [36, 41, 68, 78].”
Attenuation Correction Synthesis for Hybrid PET-MR Scanners: Application to Brain Studies (2014, methodology): “This morphological similarity, normally interpreted as an image similarity measure, is used to enforce the fact that the most morphologically similar atlases should carry more weight during the fusion process [22].”
Within-subject template estimation for unbiased longitudinal image analysis (2012, methodology): “By assuming that the current image I is generated with equal probability from the {I i } and dropping 1/N we obtain: ˜ L(x) = arg max l p(L(x) = l, I(x); {L i , I i }) where we further assumed I and L to be conditionally independent (not meaning independent, see Sabuncu et al. (2010)).”
Local Temperature Scaling for Probability Calibration (2020, methodology): “IEEE transactions on medical imaging 18(8), 712–721 (1999) [51] Sabuncu, M.”
FreeSurfer (2012, background): “, 2010a, 2010b); and finally, last but certainly not least is Mert Sabuncu, who was Polina's postdoc and is currently faculty at MGH whose work involves cutting edge segmentation using probabilistic label fusion (Sabuncu et al., 2010), and is more recently working on imaging genetics, multivariate p”
Deformable MR Prostate Segmentation via Deep Feature Learning and Sparse Patch Matching (2015, methodology): “Multi-atlas-based methods are widely used in medical imaging [10]–[12].”
Multi-Atlas Segmentation of Biomedical Images: A Survey (2014, methodology): “Since early MAS methods (Heckemann et al., 2006), one of the most popular strategies has been to utilize “nearest neighbor interpolation,” where each atlas transfers a single label to each novel image voxel, e.g., (Artaechevarria et al., 2009; Langerak et al., 2010, 2013; Sabuncu et al., 2010).”
A supervised patch-based approach for human brain labeling (2011, unknown): “The labels are then mapped and fused using classification rules to produce a new label map of the input image [30], [15], [3], [2], [16], [8], [33], [35], [26], [38], [34], [25], [19].”
Multi-Atlas Segmentation with Joint Label Fusion (2012, background): “We hypothesize that this strategy will improve segmentation accuracy over existing techniques that consider atlas-target similarity independently [3], [33].”
MUSE: MUlti-atlas region Segmentation utilizing Ensembles of registration algorithms and parameters, and locally optimal atlas selection (2015, unknown): “…with particular emphasis on atlas selection and robust and accurate fusion of the warped atlas labels (Aljabar et al., 2009; Lötjönen et al., 2010; Sabuncu et al., 2010; Landman et al., 2011; Leung et al., 2011; Asman and Landman, 2013; Cardoso et al., 2013; Zikic et al., 2014; Wu et al., 2015).”
DeepNAT: Deep Convolutional Neural Network for Segmenting Neuroanatomy (2017, unknown): “23, 81369 München, Germany; Email: christian.wachinger@med.unimuenchen.de Pohl et al., 2006; Heckemann et al., 2006; Rohlfing et al., 2004, 2005; Svarer et al., 2005; Sabuncu et al., 2010; Asman and Landman, 2012; Wang et al., 2013b; Wachinger and Golland, 2014).”
Automated Volumetry and Regional Thickness Analysis of Hippocampal Subfields and Medial Temporal Cortical Structures in Mild Cognitive Impairment (2014, methodology): “The unique feature of JLF compared to earlier multi-atlas segmentation methods that use weighted voting [Aljabar et al., 2009; Artaechevarria et al., 2009; Heckemann et al., 2010; Landman and Warfield, 2012; Sabuncu et al., 2010] is that when atlas weights are computed, an attempt is made to estimat”