RVM-MR image brain tumour classification using novel statistical feature extractor

• Published in: International journal of information technology (Singapore. Online) Vol. 15; no. 5; pp. 2395 - 2407
• Main Authors: Dixit, Asmita, Thakur, Manish Kumar
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.06.2023; Springer Nature B.V
• Subjects: Accuracy; Algorithms; Artificial Intelligence; Brain; Brain cancer; Classification; Computer Imaging; Computer Science; Critical components; Datasets; Entropy; Feature extraction; Image acquisition; Image processing; Image Processing and Computer Vision; Kernel functions; Kurtosis; Localization; Machine Learning; Magnetic resonance imaging; Medical imaging; Memory management; Neural networks; Original Research; Pattern Recognition and Graphics; Regression analysis; Skewness; Software Engineering; Statistical analysis; Support vector machines; Tumors; Vision; Wavelet transforms; Wavelet transform; DCT; Brain MR image; Relevance vector machine; PCA
• ISSN: 2511-2104; 2511-2112
• DOI: 10.1007/s41870-023-01277-9

Diagnosis of Brain Tumor is a prominent area of research in biomedical image processing to renovate the radiological machine with acquired magnetic resonance (MR) images. Accurate localization of tumor region and classification of tumors are two critical components which help for subsequent prognosis and further treatment. This paper suggests three major phases of processing of Brain MR Images to diagnose the tumor effectively. First phase includes preprocessing and localization of brain tumor through fuzzy C-mean based segmentation. Second phase performs a novel statistical feature extraction step by combining relevant contradictory features. Lastly, third phase introduces an improved Support Vector Machine (SVM) classifier as Relevance Vector machine (RVM) which employs minimal selection of kernel function to reduce the computational complexity that implies a significant memory management. The proposed method is experimentally validated on three standard datasets and experimental result yields better accuracy as compare to other existing approaches.