image segmentation system for brain MRI images by integrating and adapting existing segmentation techniques

Image segmentation system for brain MRI images by integrating and adapting existing segmentation techniques

✔️ Overview

✔️ Background and Context

Developments in finite element method (FEM) simulation and additive manufacturing (3D

printing) stand to revolutionize neuromechanics and neurosurgery. FEM simulation allows

for efficient simulation of mechanical deformation of brain tissue, which is useful in analyzing brain injury and defects [3]. Similarly, additive manufacturing opens up the possibility of creating physically-accurate brain analogues for analysis and neurosurgery training. However, the primary bottleneck in this workflow is efficiently segmenting MRI images for conversion to a finite element mesh or printable STL file. While there are commercial medical image processing software packages available, these are often expensive and still require the user to manually segment much of the image.

✔️ Project Goals

The goal of this project is to develop an image segmentation system for brain MRI images

by integrating and adapting existing segmentation techniques. The project will combine

principles from image processing, neuromechanics, and stochastic inverse modeling to de-

velop a robust algorithm for edge detection and extraction in brain MRI. The segmentation

algorithm will be separated into three main steps. First, the image will be pre-processed

using unsupervised segmentation, such as fuzzy c-means clustering (FCM) and small region

removal, to eliminate structures such as bone and background noise which are not part of

the brain. FCM has very limited success for fully segmenting complex features, but it is able

to effectively remove features that are well-separated from the relevant features of an image

[1]. The algorithm will then iterate between steps two and three. Step two is an image

processing routine using wavelet-based edge detection, since this method has been shown

to be successful at resolving multiscale edges, and in particular those found in brain MRI

image [5]. Step three will be edge optimization via minimization of an energy functional of

the surface folding, a method demonstrated to be effective in Yoon et al. [4]. The energy

functional for cortical folding will be based on exiting cortical folding mechanics models,

such as those shown in Budday et al. [2]. On each iteration, the scale of the wavelet-based

edge detection will shrink, allowing for segmentation of progressively less-apparent edges,

and eventual convergence to the final brain image. Overall, the hope is that this project can

efficiently segment the boundaries of the main brain tissue accurately enough such that the

resulting surface is useable in 3D printing and FEM simulation applications.

✔️ References

[1] M. A. Balafar, A. R. Ramli, M. I. Saripan, and S. Mashohor. Review of brain MRI image segmentation methods. Artificial Intelligence Review, 33(3):261–274, January 2010.

[2] Silvia Budday, Paul Steinmann, and Ellen Kuhl. The role of mechanics during brain development. Journal of the Mechanics and Physics of Solids, 72:75–92, December 2014.

[3] Matthew B. Panzer, Barry S. Myers, Bruce P. Capehart, and Cameron R. Bass. Development of a finite element model for blast brain injury and the effects of CSF cavitation.Annals of Biomedical Engineering, 40(7):1530–1544, February 2012.

[4] Sung Won Yoon, Hang Sik Shin, Se Dong Min, and Myoungho Lee. Medical endo-scopic image segmentation with multi-resolution deformation. In 2007 9th InternationalConference on e-Health Networking, Application and Services. Institute of Electrical &Electronics Engineers (IEEE), June 2007.

[5] Yudong Zhang, Zhengchao Dong, Lenan Wu, Shuihua Wang, and Zhenyu Zhou. Fea-

ture extraction of brain MRI by stationary wavelet transform. In 2010 International

Conference on Biomedical Engineering and Computer Science. Institute of Electrical &

Electronics Engineers (IEEE), April 2010.

SOURCE CODE CLICK HERE