Advancements In Brain Tumor Detection: A Deep Learning Approach

Abstract

A mass forms, grouping abnormal cells in the brain, known as a brain tumor. The cerebrum, safeguarded by the hard skull, faces potential dangers from any growth within its confined space. Brain tumors come in two forms: carcinogenic (dangerous) or benign (harmless). Regardless of type, tumors can escalate pressure within the skull, posing a threat to the brain. Emergencies arise with brain tumors, often leading to fatalities. Each year, approximately 6 lakh individuals in India receive a brain tumor diagnosis. Magnetic Resonance Imaging (MRI) scans are the primary diagnostic tool, revealing the presence of brain tumors. Our focus is on automating brain tumor detection from MRI images using deep learning techniques. Initial steps involve preprocessing, converting RGB images to grayscale and employing noise removal filters to enhance segmentation accuracy. Segmentation proceeds through k-means clustering and active contour methods. Feature extraction follows, utilizing discrete wavelet transform and principal component analysis. Machine learning techniques achieve an impressive 98% accuracy. Further enhancement is achieved through CNN classification, reaching 99.2% accuracy. Comparative analysis favors deep learning over traditional machine learning methods. This automated approach bypasses the need for radiologists, streamlining brain tumor identification. Our proposal introduces a robust system for brain tumor classification using deep learning.