Open AccessMRI BRAIN IMAGE SEGMENTATION BY USING A DEEP SPECTRUM IMAGE TRANSLATION NETWORK
Author(s) -
Srinivasarao Gajula,
V. Rajesh
Publication year - 2021
Publication title -
journal of medical pharmaceutical and allied sciences
Language(s) - English
Resource type - Journals
ISSN - 2320-7418
DOI - 10.22270/jmpas.v10i4.1103
Subject(s) - computer science , segmentation , artificial intelligence , deep learning , transfer of learning , pixel , translation (biology) , process (computing) , pattern recognition (psychology) , image segmentation , flexibility (engineering) , identification (biology) , neuroimaging , ground truth , task (project management) , computer vision , machine learning , medicine , psychiatry , biology , messenger rna , economics , gene , operating system , biochemistry , chemistry , statistics , botany , mathematics , management
Now a days medical image processing is challenging task. Because of its structure, flexibility in place, and irregular borders, manual identification and segmentation of brain tumours is difficult. The proposed work uses the super pixel technique to identify and segment brain tumours based on transfer learning. This process is called as dense prediction because we are predicting for each pixel in the image. It is important to identify these tumours early to provide better treatment to patients. Early detection improves the patient's chances of survival. The primary goal of this study is to use deep learning to segment brain tumours in MRI images. The suggested technique is tested using data from Kaggle data sets for Brain Tumour Segmentation. In the first step we are pre-processing the required data sets, after getting required manner we are applying the data to VGG-19 transfer learning network to identify the disorder of the brain tumours. And then we are using UNet model for tumour detection process. Due to these processes, we are getting better improvement in terms of quality metrics.