An efficient automatic segmentation of spinal cord in MRI images using interactive random walker (RW) with artificial bee colony (ABC) algorithm

• Published in: Multimedia tools and applications Vol. 79; no. 5-6; pp. 3623 - 3644
• Main Authors: Brindha, D., Nagarajan, N.
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-02-2020; Springer Nature B.V
• Subjects: Algorithms; Atrophy; Central nervous system; Computer Communication Networks; Computer Science; Contours; Data Structures and Information Theory; Experimentation; Image segmentation; Machine learning; Magnetic resonance imaging; Medical imaging; Multimedia Information Systems; Multiple sclerosis; Neurological diseases; Polybutylene terephthalates; Propagation; Shape; Solvers; Special Purpose and Application-Based Systems; Spinal cord; Support vector machines; Swarm intelligence; Topology; Support vector machine (SVM); Random-walk solvers (RW); Probabilistic boosting tree (PBT); Spinal cord segmentation; Magnetic resonance images (MRI); Artificial bee Colony (ABC); Central nervous system (CNS)
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-018-6331-8

Spinal cord Magnetic Resonance Images (MRI) have a remarkable role to play in the learning of neurological diseases such as Multiple Sclerosis (MS) affecting the Central Nervous System (CNS), in which spinal cord atrophy can help in the measurement of disease advancement and the changes in shape. Spinal cord segmentation plays a significant part in analyzing the neurological disease. In this paper here, an approach on the basis of the automatic spinal cord segmentation is proposed. This automatic technique presented performs the segmentation of the spinal cord with the help of MRI datasets. This new segmentation follows the interactive Random-Walk solvers (RW) along with Artificial Bee Colony (ABC) optimization algorithm in order to be an entirely automatic flow pipeline. The initialization of the automatic segmentation pipeline is then done with a reliable voxel-wise classification employing features similar to Haar and supervised machine learning technique i.e. Probabilistic Boosting Tree (PBT) along with Support Vector Machine (SVM) so named as PBTSVM. Thereafter, the extraction of the refinement topology of the spinal cord is then done from the temporary segmentation and it is fine-tuned for the further next random-walk solver with ABC. The refined topology results in the spinal cord’s boundary conditions from the MRI that permits the following random-walk solver with ABC for improving the segmentation result. The experimental outcomes of the novel segmentation approach depending on the MRI images indicate that the system proposed PBT-SVM algorithm provides better accuracy when compared to the other existing Active Contour Model, Multi-Resolution Propagation algorithms. Experimentation results of the proposed PBT-SVM algorithm produces higher accuracy results of 93% which is 2.5 and 3.233% higher when compared to Active Contour Model and Multi-Resolution Propagation methods respectively.