Brain Tumor Classification Using ML and DL Approaches
Shubhangi Solanki, Uday Pratap Singh, Siddharth Singh Chouhan
Abstract
In medical image processing, categorizing brain tumors is one of the most critical and challenging challenges that must be solved. Because manual classification carried out with the assistance of humans often leads to inaccurate diagnoses and projections, we have turned to machine learning to help us. The proposed work includes a performance evaluation of autonomous brain tumor identification from MR imaging as well as a CT scan utilizing basic image processing approaches depending on multiple hard and soft computing techniques. Furthermore, six classical classifiers were used to identify brain tumors in the photos. Then, to incorporate deep learning methods into our study, we used Convolutional Neural Networks and Deep Convolutional Neural Networks for brain tumor identification. The outcome of the conventional ML with the best accuracy method SVM was analyzed to the outcome of the conventional CNN. The performance of Deep CNN models is assessed, and it is discovered that VGG16 outperforms all other Deep learning models. When contrasted with these traditional classification methods, the SVM's findings were the most accurate. After that, a CNN is used, which, in contrast to the conventional classifiers, shows a discernible rise in overall performance that is much superior. It was decided to test several different layers of CNN with a variety of split ratios for the dataset. The outcomes of the experiments show that a five-layered CNN with an 80:20 split ratio is capable of achieving the greatest performance accuracy possible, which is 97.86%.