Related Experiment Video
Updated: Jun 8, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
Gaussian mixtures on tensor fields for segmentation: applications to medical imaging
Rodrigo de Luis-García1, Carl-Fredrik Westin, Carlos Alberola-López
1Laboratory of Mathematics in Imaging, Brigham and Women's Hospital, 1249 Boylston St., Boston, MA 02215, USA. rluigar@tel.uva.es
Summary
This study introduces a flexible tensor field segmentation method using Gaussian mixture models within Geodesic Active Regions. The approach successfully segments medical images like DT-MRI and hand radiographs, outperforming existing techniques.
Area of Science:
- Medical Imaging
- Computer Vision
- Statistical Modeling
Background:
- Tensor field segmentation is crucial for analyzing complex medical data.
- Existing methods using single Gaussian distributions lack flexibility for intricate data patterns.
Purpose of the Study:
- To develop a novel tensor field segmentation approach using Gaussian mixture models.
- To enhance adaptability to complex data and improve segmentation accuracy in medical imaging.
Main Methods:
- Utilized a statistical model based on mixtures of Gaussians on tensors.
- Integrated the model within the Geodesic Active Regions segmentation framework.
- Applied to both tensor-valued and scalar textured images, including a parallel processing scheme.
Main Results:
- Demonstrated successful segmentation of the corpus callosum in Diffusion Tensor Magnetic Resonance Imaging (DT-MRI) data.
- Achieved accurate bone segmentation from hand radiographs using an automatic-semiautomatic approach with anatomical priors.
- Showcased favorable comparisons with existing segmentation methods in the literature.
Conclusions:
- The proposed Gaussian mixture model offers a more flexible and adaptable tensor field segmentation than single Gaussian approaches.
- The method is suitable for diverse medical imaging segmentation tasks, including DT-MRI and radiographic analysis.
- The approach provides accurate and robust segmentation results, advancing the field of medical image analysis.
