International Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-9 Issue-1, October 2019 7226 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: E7484068519/2019©BEIESP DOI: 10.35940/ijeat.E7484.109119 Active Contour Model for Brain MR Tumor Segmentation and Volume Estimation G. Anand Kumar, P. V. Sridevi Abstract: Brain MR tumor segmentation and estimation of volume is a critical task in medical applications. Brain tumors are analyzed by the common test method known as magnetic resonance imaging (MRI) which provides a detail image of brain. The proposed work involves detection of tumor in brain using deep learning based active contour model. Segmentation is the main objective of the proposed work for achieving detailed information about the tumor and accurate volume estimation to detect the size of the tumor. The Euclidean similarity factor (ESF) is used for considering the spatial distances and intensity differences of the region there by preserving all the fine details of the image. 3D convolutional neural network (3DCNN) is used for extracting the features and segmentation to identify the tumor location in the brain. Finally, shoelace method is used to estimate the volume of the tumor, and it provides treatment planning, surgical methods, estimation of dose, etc. The simulation results in this suggested approach could attain effective performance as compared with the existing approaches. Keywords: Brain tumor, Magnetic resonance imaging, Euclidean similarity factor, Convolutional neural network. I. INTRODUCTION In medical image processing, the segmentation of brain tumor and analysis of volume estimation is an essential process focused in research area [1]. Segmentation of tumor and volume estimation also focused on diagnosis and treatment planning. From the past decades, cancer is one of the main diseases that frights the people more. Brain disease is an important challenging malignant tumor to cure [2]. So the technology gives more importance to the estimation of various tumors in brain by oncologist, neurosurgeons and all medical team, they needed to identify the entire information and images of the brain tumors. Moreover the technology associated with more number of images, can‟t be easily findable by Surgeons or oncologists. Therefore, there is need for segmentation. Revised Manuscript Received on December 22, 2018. G. Anand Kumar, Assistant Professor, Department of ECE, Gayatri Vidya Parishad College of Engineering(Autonomous), Visakhapatnam, Andhra Pradesh, India.. Dr. P. V. Sridevi, Professor, Department of ECE, Andhra University College of Engineering (Autonomous), Visakhapatnam , Andhra Pradesh, India. The segmentation change the characteristics and estimate the tumor [3]. Solid or active cancer, necrosis and edema are the different tumor matters which is separated, then it segments the standard and abnormal tissues. The typical tissue consist of gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF)[4]. In brain tumor segmentation the normal tissue can detect and separate easily while the abnormal tissue cannot be able to detect easily. The maximum size of tumor detected in boundary box is pixel 1-5035, in which the minimum size is 1-1190 [5]. Among past years, the segmentation are taken through manually, in which physicians and oncologist handle the segmentation. This manual segmentation face more problems such as time consuming and inter-intra rater errors [6]. So automatic or semi-automatic segmentations are used. In automatic segmentation, the tumor tissues automatically segmented without the manual method. The atlas is estimated in the segmentation with different shapes and locations of tissues. Voxels are used through Markov Random Fields (MRF) for smooth segmentation [7]. It also used for the segmentation of the super voxels. Moreover the voxels may be mistakenly segment wrong class and locations. To overcome these Conditional Random Field, and classifier, are used [8-9]. The automatic brain tumor image segmentation is classified as edge based segmentation, regions manipulation segmentation and segmentation by pixel manipulation. The edge based segmentation consist of edge detection and active contours, then the region based is divided as merge/split and graph cut, and the pixel based is subdivide as thresholding and clustering, along it divided as global, adaptive, k-means and fuzzy-c means segmentation respectively [10]. The various techniques and their algorithm are used for the automatic brain tumor segmentations. They are given as; Histogram based method, it‟s an efficient segmentation method compared to other techniques [11]. Peaks and valleys is used to locate the clusters in histogram. It also used for the measurement of dimensions of image pixels. But it have a limitation as, it is more challenging to find the insignificant of peaks of segmentation and valleys in the certain images. Edge based segmentation, detects and identifies the region of tumor. The edge based segmentation is a common method and it segment the boundaries of images [12]. In edge based segmentation the gray level and color images are used.