Cerebrum Tumor Segmentation of High Resolution Magnetic Resonance Images Using 2D-Convolutional Network with Skull Strip
- PDF / 2,224,296 Bytes
- 14 Pages / 439.37 x 666.142 pts Page_size
- 106 Downloads / 201 Views
Cerebrum Tumor Segmentation of High Resolution Magnetic Resonance Images Using 2D-Convolutional Network with Skull Stripping R. Pitchai1 · Ch Madhu Babu1 · P. Supraja2 · Mahesh Kumar Challa1 Accepted: 8 October 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The automatic segmentation of the tumor region from Magnetic Resonance cerebrum imageries is a difficult task in medical image analysis. Numerous techniques have been created with the goal of improving the segmentation effectiveness of the automated framework. As of late, Convolutional Neural Networks have accomplished better performance in various recognition tasks. In this paper, 2D-ConvNet with skull stripping (SS-2D ConvNet) based brain tumor segmentation technique have been presented. In the proposed method, initially, the input MRI images are preprocessed to reduce noise and skull stripped to correct the contrast and non-uniformity. It is further processed through the 2D-ConvNet for the segmentation of brain tumor. In particular, the proposed method has been compared with other existing methods, and it achieves better performances and yield precise segmentation with dice scores of 91%, accuracy of 89%, specificity of 98%, and sensitivity of 87%. Keywords Brain tumor · Segmentation · Skull stripping · Wiener filter · 2D-ConvNet
1 Introduction Brain tumors influence the individual severely due to the irregular development of cells inside the cerebrum. It can disturb the appropriate cerebrum operation and be dangerous.
B
R. Pitchai [email protected] Ch Madhu Babu [email protected] P. Supraja [email protected] Mahesh Kumar Challa [email protected]
1
Department of Computer Science and Engineering, B.V. Raju Institute of Technology, Narsapur, Medak Dist, Telangana, India
2
Department of Information Technology, School of Computing, SRM Institute of Science and Technology, Kattankulathur, Chennai, Tamilnadu, India
123
R. Pitchai et al.
Two sorts of brain tumors have been distinguished as benign tumors and malignant tumors. Benign tumors are less destructive than malignant tumors as malignant are fast developing and harmful when compared with benign tumors. Even though there are many medical imaging techniques such as MRI, CT scan, Ultrasound, X-ray, PET, and more are used to capture the brain images, MRI gives more prominent differentiation images. Therefore, brain tumor identification is performed by using MRI imageries. This paper focuses on brain tumor segmentation using image processing techniques. Segmenting brain tumors inside the MRI imageries have a significant challenge in computer vision. Physical separation of tumor is a difficult and tedious errand. In addition, there is irregularity among the segmented results created by human specialists. A precise automated segmentation strategy helps in recovering treatment, and observation of tumor development by absolutely measuring the size of the tumor. The tumor tissues in MRI imageries profoundly cover with different pathologies or
Data Loading...
