Brain Tumor Segmentation from 3D Brain MRI Using 3D Convolutional Neural Network
Hiren Patel1, Mehul C Parikh2, Mahesh Pipalia3
1Hiren Patel, Department of Information Technology, L D College of Engineering, Ahmedabad, (Gujarat) India.
2Mehul C Parikh, Department of Information Technology, L D College of Engineering, Ahmedabad (Gujarat) India.
3Mahesh Pipalia, Department of CEO and Founder, Script All DNA Technology, Ahmedabad (Gujarat) India.
Manuscript received on 01 May 2019 | Revised Manuscript received on 15 May 2019 | Manuscript published on 30 May 2019 | PP: 2611-2617 | Volume-8 Issue-7, May 2019 | Retrieval Number: G5431058719/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Magnetic Resonance Imaging (MRI) is one of the medical imaging modality that widely used by the doctors to represent the internal brain structure information digitally. There are plenty of methods available for the classification and segmentation of tumor from the brain MRI study. Brain tumor detection in the initial stage is very important for accurate and better treatment. Patient survival chances can be increased if the accurate tumor is segmented from the brain which can help doctors to treat the patient accordingly. There are a number of existing emerging machine learning algorithms contributed to this problem area. Convolutional Neural Network (CNN) is a widely used method for this type of image segmentation problems. 3D CNN is already achieving better results in this work but it takes lots of data and the time to train such very high accurate model. In this research, 3D CNN is used along with the biological structural information of the brain i.e. the brain has a symmetric structure which can be divided into two nearly equal half, information from each half can help CNN model to differentiate the abnormal tissues from the normal tissues. Using the biological information the results are improved by 10.27%. The fractal search algorithm is implemented to reduce the time complexity of the tumor segmentation process. The segmentation processing time has been reduced by 41.75% on GPU while 23.69% on CPU and improves the segmentation result by 2.76%.
Keyword: Brain Tumor Segmentation, CNN, Fractal Search.
Scope of the Article: Neural Information Processing