Evaluation Effectiveness of Brain Tumors Detection using Deep Learning: A Comparative Analysis

Abstract

An important health issue in India is brain tumor, which add to the nation's overall cancer burden. The number of cases of brain tumors rising from last decades so research on this issue is important for improving timely diagnosis, treatment and patient health. Latest technologies such as Deep Learning (DL) shows significant potential for detecting brain tumors at early stage. This surpassing traditional diagnostic methods and helps healthcare professionals for detection of brain tumors at early stages. In this paper we perform comparative analysis of Brain tumors detection using machine learning (ML) and Deep learning. DL model such as Convolutional neural network (CNN), VGG19, ResNetV2, DenseNet121 were used to train the model.

Introduction: Unusual cell growth in the brain or central spinal canal is called a brain tumors. The two main categories of these tumors are benign and malignant. Malignant cells are cancerous, however benign ones are not. Malignant tumors can be aggressive, spreading to other sections of the brain and possibly compromising general health, whereas benign tumors typically develop slowly and do not infect neighbouring tissues. A primary brain tumor is a kind of tumor that doesn't spread from other regions of the body; instead, it starts in the brain or adjacent tissues. Different brain cell types can give rise to secondary brain tumors, which can then be categorized according to the cell of origin. Brain tumors, which impact millions of people annually worldwide, pose a serious problem in the fields of neurology and oncology. Effective treatment and better patient outcomes depend on their early discovery and precise diagnosis. The diagnosis of brain tumors has historically mostly depended on invasive techniques and expert interpretations of medical imaging. But new developments in machine learning (ML) have brought forward creative strategies that could change this environment.

Objectives: To study and analyze the various techniques for the Brain Tumor disease prediction using neural network. Data preprocessing of selected dataset. (Normalization and gray scaling). Comparative Analysis of the existing system to access effectiveness and efficiency of proposed model. Use classification techniques using deep learning algorithm and develop a model. Performance analysis using evaluation metric.

Methods: In the realm of medical imaging, deep learning has emerged as a powerful instrument, particularly for detecting and classifying brain cancers. DL models have demonstrated greater accuracy in diagnosing brain tumors through MRI scans. Various types of brain tumors, including gliomas, meningiomas, and pituitary tumors, can also be diagnosed using deep learning models. In the first phase of model building we collect the dataset from different sources. This dataset include MRI scan images from different radiologist. After that we pre-process the dataset such as resize the images, rotate the images and convert the images into grayscale image. We should remove the noise from the images to extract the features. We use high pass filtering, low pass filtering, Gaussian filters to remove the noise from the images. Different types of features we can extract from the dataset such as shape based features, intensity based features and model based features. In the next phase we perform classification of the images using CNN, ANN, autoendoder models. In this phase we can classify the images into healthy image and tumour image.

Results: A comparative study of the literature review conducted by several writers is presented

Conclusions: DL has emerged as a powerful model, particularly for the detection and classification of brain tumors. In this paper we discuss about detail comparative analysis of DL model used for brain tumors detection.