INTELLIGENT MACHINE VISION FRAMEWORK FOR BRAIN TUMOUR DETECTION AND CLASSIFICATION USING MAGNETIC RESONANCE IMAGES (MRI)
Abstract
Brain tumour detection from Magnetic Resonance Imaging (MRI) scans remains a challenging task in medical diagnosis, as accurate identification of abnormal brain tissues requires careful examination of complex image structures. Early and reliable detection can help medical professionals make timely decisions and support appropriate treatment planning. With the advancement of artificial intelligence and computer vision, image processing techniques have become increasingly useful for analysing medical images and identifying potential tumour regions. This study explores an image-processing-based approach for brain tumour analysis using the Computer Vision and Image Processing (CVIP) tools environment. The framework supports several essential stages, including image pre-processing, feature extraction, and classification, to improve the quality of MRI images and facilitate the identification of tumour-related patterns. During pre-processing, image enhancement techniques can help improve visual clarity and reduce unwanted variations, while feature extraction focuses on identifying relevant characteristics that distinguish tumour and non-tumour regions. These extracted features are then used by different classification algorithms to evaluate their ability to recognise brain tumour images. The study compares three classifiers: Naive Bayes, which achieves an accuracy of 90%; a Support Vector Machine (SVM) based on the Bag of Visual Words (BoVW) approach, which reaches 97.3%; and a Convolutional Neural Network (CNN), which obtains the highest reported accuracy of 98.5%. These results indicate that the CNN model performs better than the other two classifiers under the reported experimental conditions. Its superior performance may be attributed to its ability to learn meaningful image features and capture complex spatial patterns directly from image data. Furthermore, comparing these classification methods provides useful insight into the effectiveness of traditional machine learning and deep learning approaches for medical image analysis. The findings highlight the potential of combining image processing with intelligent classification techniques to support computer-aided brain tumour detection. However, further evaluation using larger and more diverse MRI datasets is necessary to assess the models' generalisation ability and reliability in different clinical settings. Overall, this research demonstrates the promise of CNN-based classification for improving automated brain tumour image analysis, while emphasising that such systems should complement, rather than replace, expert medical judgement.
Keywords : Deep Learning, Brain Tumour, Classification, SVP, Machine Vision.
