Semi-automatic active contour-based segmentation to remove eyes, meninges, and skull from MRI

Abstract

Purpose

Brain imaging acquisition can present different issues, such as noisy images which can result in a problematic diagnosis. Image preparation such as skull stripping and region segmentation is a fundamental step in order to support a better medical diagnosis outcome. Therefore, this study presented a segmentation technique based on the active contour model to perform skull stripping.

Methods

The method is applied on the neuroimaging database available by the OASIS neuroimaging dataset. The method proposed here uses active contour model followed by k-means clustering technique in order to converge to a locally minimal energy value which can be equivalent to the brain tissue area aiming to avoid loss of image quality and brain structures. Statistical analysis was also performed in order to determine how image texture characteristics were affected.

Results

The active contour method achieved results within the ones presented on the state-of-art values segmentation methods with 96.4% of sensitivity and 96% of specificity using only 4 k-means clusters. Image texture characteristics such as entropy and correlation presented values of 1.8804 and 0.96, respectively.

Conclusion

The high entropy value accentuated the gray-level contrast and highlighted anatomical structures for brain visualization. These high evaluation scores demonstrate that the semi-automatic contour-based segmentation algorithm is a powerful tool for segmentation and skull-stripping decreasing loss of image quality and brain structures.