ISSN 2394-3777 (Print) ISSN 2394-3785 (Online) Available online at www.ijartet.com International Journal of Advanced Research Trends in Engineering and Technology (IJARTET) Vol. 9, Issue 2, February 2022 All Rights Reserved © 2022 IJARTET 20 Analysis of Brain Tumor Classification using Pre-Trained CNN models Dr. M. Thamarai 1 , Dhivyaa S P 2 1 Professor, ECE Department, Sri Vasavi Engineering College, Tadepalligudem, Andhra Pradesh, India. 2 Senior System Engineer, Infosys, Bangalore, India. Abstract: Brain tumor detection is one of the crucial tasks in medical image processing. The difference between normal cells and infected cells is very less and almost both appear similar. So, the detection by the radiologist is inaccurate and there is a need for an automated system for brain tumor detection. This paper proposes an automated brain tumor classification system using 3D Magnetic Resonance Brain Images using Convolution neural network transfer learning concept. The transfer learning concept is used to modify or fine-tune the standard CNNs according to the user applications. This concept reduces the huge amount of input data requirements and minimizes the training and thus in computation time of the process. The top layers of benchmark CNN architectures like VGG16, ResNet 50, and InceptionV3 are fine-tuned and utilized for tumor detection. The performance of the CNN structures is analyzed in terms of performance metrics such as Accuracy, specificity, sensitivity, and various losses. Keywords: Brain tumor detection, magnetic resonance imaging, convolution neural network, and Transfer learning I. INTRODUCTION Recently, Deep learning evolves as a major area for researchers, because of its high prediction accuracy and less error rate. Deep Learning algorithms/Structures perform better than humans, even in a huge volume of data. DL algorithms performance is parallel to the input data. So, we can say deep learning networks are data-hungry networks and we need a huge amount of data to make the network learn. The transfer learning concept introduced in Machine learning helps to utilize CNN applications with less available data. Convolution neural networks are the advanced structure in Deep Learning and are designed especially for various image processing applications such as segmentation, feature extraction, and enhancement. The applications of CNN in medical image processing are mainly for the classification of cancer cells such as breast cancer, lung cancer, and brain tumor detection. The conventional automatic segmentation methods need a classifier with features which is a challenging task. CNN algorithm solves complex features such as healthy brain tissues and tumor tissues through multimodal MRI brain images. This paper discusses brain tumor classification from MRI Brain images. The human brain does a complex job in controlling the other organs that work with billions of cells. A brain tumor occurs when an uncontrolled division of cells forms an abnormal group of cells around or inside the brain. These groups can affect the normal functionality of brain activity and destroy healthy cells (Kavitha et. al, 2016). Brain tumor classifies as grade I, grade II, grade III, and grade IV. Grade I and Grade II are termed as lower-grade (Benign). Grade III and Grade IV are termed as high grade (malignant). Benign tumors are nonprogressive (non-cancerous) so considered to be less aggressive, they originated in the brain and grow slowly; also, they cannot spread anywhere else in the body. However, a malignant tumor is cancerous because the cancerous cell grows rapidly in irregular boundaries. The tumor cells which originated in the brain itself is called a primary malignant tumor and the tumor cells originated in any other part of the body and spread to the brain are called a secondary malignant tumor (KhambhataKruti G et al, 2016, Kaur. G et al, 2016) The medical modalities are X-ray, CT (Computed Tomography) and MRI (Magnetic Resonance Imaging). MRI is one of the best imaging techniques that researchers use for the detection of brain tumors due to its high resolution and interpretation in images. So, it is used in brain tumor detection and treatment phases. MRI images are more suitable for automatic brain tumor analysis because of their ability to provide a lot of information about the brain structure and abnormalities within the brain tissues due to the high resolution of the images [2,5]. Researchers presented different automated approaches for brain tumor detection and type classification using brain MRI images since it became possible to scan and load medical images to the computer.