Automated Brain Tumor detection using multi-label images of MRI scans and CNNs

Abstract

Numerous medical diagnostic applications now heavily rely on automatic defect detection in medical imaging. Automatically detecting tumors by MRI is essential for treatment planning because it offers details on aberrant tissues. Due to the volume of data required, this strategy is impracticable. As a result, to lower the rate of human death, trustworthy and automatic classification techniques are required. Automated tumor detection techniques are therefore being created to free up radiologist time and attain proven accuracy. Brain tumor, which develops as a result of the abnormal development and division of brain cells, eventually turn into brain cancer. The study of human health benefits greatly from the use of computer vision since it is used to eliminate the requirement for precise human judgement. The most reliable diagnostic tools include MRI scans, CT scans, and X-rays. Secure imaging techniques within magnetic-resonance imaging (MRI). In this study, the noises present in an MR image were removed using a morphological opening to the pre-processing. Binary thresholding and Neural Network segmentation methods were then used to accurately detect tumors. Our model will assess if the person has a brain tumor or not. To increase the accuracy of different models and scaling methods such as Efficient B2, B3, and B6, we want to test and experiment with them.