DEEP LEARNING-BASED BRAIN TUMOR DETECTION WITH ITS CLASSIFICATION

Abstract

In this study we use convolutional neural networks (CNNs) to present a novel approach to brain tumor segmentation. Using deep learning substantially improves the precision and speed of brain tumor segmentation from medical imaging data. In order to train and evaluate the CNN architecture, This model use a large and diverse dataset of brain scans (Br35H) that has been carefully selected from many sources. This ensures excellent performance and generalizability, allowing our system to handle new, unexplored data successfully. Our CNN architecture is meticulously constructed to detect and segment brain tumors with high accuracy. The CNN model uses convolutional and pooling layers to extract important features from input pictures, followed by fully connected layers to deliver the final segmentation output.Performance enhancements include batch normalization, transfer learning, and data augmentation. Experimental results show that our suggested approach is effective in segmenting brain tumors with a 90% accuracy rate and dependability, exceeding current methods in precision and loss, and reaching state-of-the-art performance on benchmark datasets. There are four forms of brain tumors that we can identify and categorize: "glioma," "meningioma," "no-tumor," and "pituitary." This technique also as potential for other medical imaging use including tumor tracking, treatment response assessment, and personalized medicine, these encouraging results show that our method has the power to be a useful tool in medical image analysis for brain tumor early detection and treatment planning, facilitating earlier diagnosis, improving treatment outcomes, and improving patient care.

Author

Dr. R. JAYANTHI, DIVYASRIDHA .M , JEEVA.V , MOHANA PREETA .C , VIMAL.S