International Research Journal of Engineering and Technology (IRJET) Volume: 09 Issue: 05 | May 2022 www.irjet.net
e-ISSN: 2395-0056 p-ISSN: 2395-0072
Brain Tumor Detection From MRI Image Using Deep Learning Divya Pathak1, Dadi Ganesh2, Gosu Yogeswar3, Dodda Pavan Chandra4, P. Karunakar5 1,2,3,4 UG 5Assistant
Student Dept. of EEE PVPSIT, Vijayawada, Andhra Pradesh, India- 520 007 Professor, Dept. of EEE PVPSIT, Vijayawada, Andhra Pradesh, India- 520 007
-------------------------------------------------------------------------------------------------------------***------------------------------------------------------------------------------------------------------------
Abstract— Identification of brain tumors is one of the most important and challenging tasks in medical imaging because manual human-assisted classification can lead to inaccurate predictions and diagnoses. Medicine requires fast, automated and reliable technology for tumor detection. Applying a deep learning approach in the context of improving health diagnosis provides an effective solution. This experimental work was performed with a dataset consisting of magnetic resonance imaging (MRI) images of tumors of various shapes, sizes, textures and locations. MRI scans reveal information about abnormal tissue growth in the brain. A self defined Convolution Neural Network (CNN) deep learning architecture is used for the classification of image as tumor or non-tumor. In our work, CNN model without transfer learning technique gained an accuracy of 81.42% and with transfer learning the accuracy reached to 98.8% which is very compelling.
and benign brain tumors and other central nervous system (CNS) tumors will be diagnosed in the United States in 2019 [2]. Hence early detection of brain tumor is very important. In this paper, we proposed a Convolution Neural Network architecture with and without transfer learning and achieved an accuracy of 98.8%. II. LITERATURE REVIEW In medical diagnosis, the robustness and accuracy of prediction algorithms are important, as the results are crucial to the treatment of the patient. There are many popular classification and clustering algorithms used for prediction. The purpose of medical image clustering is to simplify the representation of an image into a meaningful one and make it easier to analyze. Several clustering and classification algorithms aim to improve the predictive accuracy of the diagnostic process in anomaly detection.
Keywords—Convolution Neural Network, Magnetic Resonance Imaging, Transfer Learning. I.
INTRODUCTION
Devkota et al. [3] established the whole segmentation process based on the mathematical morphological operations and the spatial FCM algorithm to improve the computation time, but the proposed solution has not been tested until the evaluation stage. prices and results because it detects cancer with 92% and the classifier has an accuracy of 86.6%. Yantao et al. [4] is similar to the histogram-based segmentation technique. Treat the brain tumor segmentation task as a three-class classification problem (tumor including necrotic tumor and tumor, edema and normal tissue) involving two modalities FLAIR and T1. Abnormal regions were detected using a regionbased active contour model on the FLAIR method. Edema and tumor tissues were differentiated into abnormal regions based on T1 contrast enhancement by kmeans method and obtained a Dice coefficient and sensitivity of 73.6% and 90.3%, respectively.
Medical imaging is a method and process for generating visual representations of the functions of specific organs or tissues as well as visual representations of the interior of the body for clinical analysis and medical intervention. Medical imaging aims to diagnose and treat disease by revealing the internal structures hidden in the skin and bones. Medical imaging also creates a database of normal anatomy and physiology that enables anomaly detection. Image processing technology is the manipulation of digital images using computers. This method has many advantages such as elasticity, adaptability, data storage and communication. An increasing number of different image resizing methods allow you to store images efficiently. This method has a large set of rules for synchronously running the image. Multidimensional processing of 2D and 3D images is possible.
Yu et al. [5] states that image segmentation is used to extract important objects from an image. They propose to divide an image into three parts, including black, gray and white. The Z function and the s function are used for fuzzy division of the 2D histogram. Then, QGA is used to find the combination of 12 parameters belonging to, which has the largest value. This technique is used to improve the image segmentation and the implication of their work is that the
According to [1], cancer of the brain and other nervous system is the 10th leading cause of death, and the 5year survival rate for brain cancer patients is 34% for men and 36% for women. Moreover, the World Health Organization (WHO) claims that around 400,000 people worldwide have a brain tumor, and 120,000 people have died in the past few years. In addition, it is expected that approximately 86,970 new cases of primary malignant © 2022, IRJET
|
Impact Factor value: 7.529
|
ISO 9001:2008 Certified Journal
|
Page 486