A COMPARATIVE STUDY FOR BRAIN TUMOUR DETECTION USING MAGNETIC RESONANCE IMAGES

Abstract

Brain tumours are abnormal neural growths that can be fatal and develop an immense variety of symptoms. Accurate categorization of brain tumours is crucial for optimizing therapeutic approaches and improving overall patient outcomes. In this study, we propose a comparative brain tumour detection system using 6 machine learning algorithms which are Naïve Bayes algorithm, KNN, Random Forest, Ada Boost, SVM and CNN. In the study Experimental results demonstrate that deep learning models, particularly CNN outperforms the other algorithm with the accuracy of 98.16. The suggested system is divided into three primary phases: feature extraction, classification, and picture pre-processing. In the pre-processing stage, noise in Magnetic Resonance Imaging (MRI) scans is reduced, and contrast is improved. This comparison analysis clarifies the advantages and disadvantages of each approach and offers guidance on which models to use depending on particular computational and clinical needs. Subsequent research endeavours will investigate the amalgamation of these models into a cohesive structure to optimise their mutual advantages for heightened precision in diagnosis. The practical ramifications of these findings for clinical settings are covered in the thesis conclusion. Future research is advised to investigate hybrid models that combine the advantages of several algorithms and incorporate cutting-edge methods like generative adversarial networks and transfer learning to improve detection accuracy. In conclusion, this thesis offers a thorough comparison of various machine learning methods for brain tumour identification, stressing the advantages and disadvantages of each approach.