Learning Task-Optimal Registration Cost Functions for Localizing Cytoarchitecture and Function in the Cerebral Cortex

Computer Science and Artificial Intelligence Laboratory, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, Cambridge, MA 02139, USA.
IEEE transactions on medical imaging 07/2010; 29(7):1424-41. DOI: 10.1109/TMI.2010.2049497
Source: PubMed


Image registration is typically formulated as an optimization problem with multiple tunable, manually set parameters. We present a principled framework for learning thousands of parameters of registration cost functions, such as a spatially-varying tradeoff between the image dissimilarity and regularization terms. Our approach belongs to the classic machine learning framework of model selection by optimization of cross-validation error. This second layer of optimization of cross-validation error over and above registration selects parameters in the registration cost function that result in good registration as measured by the performance of the specific application in a training data set. Much research effort has been devoted to developing generic registration algorithms, which are then specialized to particular imaging modalities, particular imaging targets and particular postregistration analyses. Our framework allows for a systematic adaptation of generic registration cost functions to specific applications by learning the "free" parameters in the cost functions. Here, we consider the application of localizing underlying cytoarchitecture and functional regions in the cerebral cortex by alignment of cortical folding. Most previous work assumes that perfectly registering the macro-anatomy also perfectly aligns the underlying cortical function even though macro-anatomy does not completely predict brain function. In contrast, we learn 1) optimal weights on different cortical folds or 2) optimal cortical folding template in the generic weighted sum of squared differences dissimilarity measure for the localization task. We demonstrate state-of-the-art localization results in both histological and functional magnetic resonance imaging data sets.


Available from: B.T. Thomas Yeo
  • Source
    • "There are a wide range of registration algorithms, which employ different objective functions, deformation models, and optimization strategies (Sotiras et al., 2013). The optimal choice of algorithm specifics largely depend on the biomedical application, its goal, and operational constraints , such as available computational resources, desired accuracy, and restrictions on time (Yeo et al., 2010). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Multi-atlas segmentation (MAS), first introduced and popularized by the pioneering work of Rohlfing, et al. (2004), Klein, et al. (2005), and Heckemann, et al. (2006), is becoming one of the most widely-used and successful image segmentation techniques in biomedical applications. By manipulating and utilizing the entire dataset of "atlases" (training images that have been previously labeled, e.g., manually by an expert), rather than some model-based average representation, MAS has the flexibility to better capture anatomical variation, thus offering superior segmentation accuracy. This benefit, however, typically comes at a high computational cost. Recent advancements in computer hardware and image processing software have been instrumental in addressing this challenge and facilitated the wide adoption of MAS. Today, MAS has come a long way and the approach includes a wide array of sophisticated algorithms that employ ideas from machine learning, probabilistic modeling, optimization, and computer vision, among other fields. This paper presents a survey of published MAS algorithms and studies that have applied these methods to various biomedical problems. In writing this survey, we have three distinct aims. Our primary goal is to document how MAS was originally conceived, later evolved, and now relates to alternative methods. Second, this paper is intended to be a detailed reference of past research activity in MAS, which now spans over a decade (2003-2014) and entails novel methodological developments and application-specific solutions. Finally, our goal is to also present a perspective on the future of MAS, which, we believe, will be one of the dominant approaches in biomedical image segmentation. Copyright © 2015 Elsevier B.V. All rights reserved.
    Medical image analysis 12/2014; 24(1). DOI:10.1016/j.media.2015.06.012 · 3.65 Impact Factor
  • Source
    • "Usually, such parameters are adjusted manually by observing the registration results, which does not always guarantee that the best combination is achieved. A solution to overcome this limitation is proposed in (Yeo et al., 2010b). "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper presents a review of automated image registration methodologies that have been used in the medical field. The aim of this paper is to be an introduction to the field, provide knowledge on the work that has been developed and to be a suitable reference for those who are looking for registration methods for a specific application. The registration methodologies under review are classified into intensity or feature based. The main steps of these methodologies, the common geometric transformations, the similarity measures and accuracy assessment techniques are introduced and described.
    Computer Methods in Biomechanics and Biomedical Engineering 01/2014; 17(2):73-93. DOI:10.1080/10255842.2012.670855 · 1.77 Impact Factor
  • Source
    • "In calculating group average surface areas, we corrected for multiple sources of distortion, resulting in estimated sizes that were very similar to those of a typical individual subject. Further progress might be obtained by using more detailed functional information to drive the MSM registration, such as the signed distance transform of areas (Yeo et al., 2010) or by using more detailed retinotopic information (e.g., horizontal and vertical meridia, or even continuously varying eccentricity and polar angle maps). We also addressed an underappreciated problem of systematic drift that can occur during iterative methods of registration to a template. "
    [Show abstract] [Hide abstract]
    ABSTRACT: We generated probabilistic area maps and maximum probability maps (MPMs) for a set of 18 retinotopic areas previously mapped in individual subjects (Georgieva et al., 2009 and Kolster et al., 2010) using four different inter-subject registration methods. The best results were obtained using a recently developed multimodal surface matching method. The best set of MPMs had relatively smooth borders between visual areas and group average area sizes that matched the typical size in individual subjects. Comparisons between retinotopic areas and maps of estimated cortical myelin content revealed the following correspondences: (i) areas V1, V2, and V3 are heavily myelinated; (ii) the MT + cluster is heavily myelinated, with a peak near the MT/pMSTv border; (iii) a dorsal myelin density peak corresponds to area V3D; (iv) the phPIT cluster is lightly myelinated; and (v) myelin density differs across the four areas of the V3A complex. Comparison of the retinotopic MPM with cytoarchitectonic areas, including those previously mapped to the fs_LR cortical surface atlas, revealed a correspondence between areas V1-3 and hOc1-3, respectively, but little correspondence beyond V3. These results indicate that architectonic and retinotopic areal boundaries are in agreement in some regions, and that retinotopy provides a finer-grained parcellation in other regions. The atlas datasets from this analysis are freely available as a resource for other studies that will benefit from retinotopic and myelin density map landmarks in human visual cortex.
    NeuroImage 01/2014; · 6.36 Impact Factor
Show more

Similar Publications