A Hybrid Geometric–Statistical Deformable Model for Automated 3-D Segmentation in Brain MRI

Dept. of Electr. & Comput. Eng., Univ. of British Columbia, Vancouver, BC
IEEE Transactions on Biomedical Engineering (Impact Factor: 2.35). 08/2009; 56(7):1838 - 1848. DOI: 10.1109/TBME.2009.2017509
Source: PubMed


We present a novel 3-D deformable model-based approach for accurate, robust, and automated tissue segmentation of brain MRI data of single as well as multiple magnetic resonance sequences. The main contribution of this study is that we employ an edge-based geodesic active contour for the segmentation task by integrating both image edge geometry and voxel statistical homogeneity into a novel hybrid geometric-statistical feature to regularize contour convergence and extract complex anatomical structures. We validate the accuracy of the segmentation results on simulated brain MRI scans of both single T1-weighted and multiple T1/T2/PD-weighted sequences. We also demonstrate the robustness of the proposed method when applied to clinical brain MRI scans. When compared to a current state-of-the-art region-based level-set segmentation formulation, our white matter and gray matter segmentation resulted in significantly higher accuracy levels with a mean improvement in Dice similarity indexes of 8.55% (p<0.0001) and 10.18% (p<0.0001), respectively.

Download full-text


Available from: Albert Huang, Oct 10, 2015
18 Reads
  • [Show abstract] [Hide abstract]
    ABSTRACT: Summary form only given. We demonstrate a carrier suppressed source that reduces the problem of relative drift between the carrier and filter wavelengths. The source consists of a ring laser cavity with an intracavity optical modulator and a frequency selective output coupler (FSOC). The FSOC transmits light at a frequency ω<sub>0</sub> and couples light out of the cavity for frequencies outside of the narrow passband centered around ω<sub>0</sub>. When this optical signal is demodulated by a photodetector, the DC electrical power is reduced more than the RF power. Hence, optical amplification to the maximum photodetector DC power results in a net RF gain as compared to the case with the same DC power without carrier suppression. An advantage of this system is that the carrier field is passively locked to the frequency that is maximally transmitted through the FSOC.
    Optical Fiber Communications, 1996. OFC '96; 01/1996
  • [Show abstract] [Hide abstract]
    ABSTRACT: Diffuse optical tomography is an emerging, non-invasive technique with unique portability and hemodynamic contrast capabilities. A major challenge of diffuse optical tomography is the requirement for a prior head anatomy for constructing three dimensional images of brain activation. In this paper, we use different characteristics of T1, T2 weighted magnetic resonance imaging to segment head into three parts (scalp-skull, CSF, and brain) based on tissue optical properties. Similarity index successfully demonstrated segmentation results with acceptable accuracy for diffuse optical tomography construction.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Segmentation of brain MRI is the core part in plenty of medical image processing methods. Due to some properties of MR images such as intensity inhomogeneity of tissues, partial volume effect, noise and some other imaging artifacts, segmentation of brain MRI based on voxel gray values is prone to error. Hence involving problem specific information and expert knowledge in designing segmentation algorithms seems to be useful. A two-fold fuzzy segmentation algorithm based on Bayesian method is proposed in this paper. The Bayesian part uses the gray value of voxels in segmenting images and the segmented image is used as the input to fuzzy classifier to improve the misclassified voxels especially in borders between tissues. Similarity index is used to compare our algorithm with the well known method of Ashburner which has been implemented by Statistical Parametric Mapping (SPM) software. Two different brain MRI datasets are used to evaluate the algorithm. Brainweb as a simulated brain MRI dataset and ADNI as real brain MRI dataset are practiced images. Results show that our algorithm performs well in comparison with the one implemented in SPM. It can be concluded that incorporating expert knowledge and problem specific information in segmentation process improve segmentation result. The major advantage of proposed method is that one can update the knowledge base and incorporate new information into segmentation process by adding new fuzzy rules.
Show more