Identification and segmentation of tumor using deep learning and image segmentation algorithms

Abstract

Brain tumor is a typical mass of tissue that develops when cells proliferate and divide excessively. Brain tumor perception requires a great deal of work and experience from the medical professional in order to identify the tumor's precise location. If a brain tumor is not discovered in a timely manner, it affects a person's ability to function normally and raises the death rate. This study focuses on tumor segmentation and tumor detection using magnetic resonance imaging (MRI) images. This work helps the medical professional to precisely identify the tumor location and segmentation process provides cost effective data storage. The YOLOv8s model is utilized for tumor identification, while the image segmentation technique is employed for tumor segmentation. The images come from an open-source dataset used for research, and Roboflow 100 transforms them into .yaml files that are congenial with YOLOv8s. To train the model the dataset is split into training, validation and testing. Proposed model consist of dataset which comprises 639 images, of which 453 are utilized for training, 122 for validation, and 64 for testing, resulting in a ratio of 71:19:10. The dataset is subjected to preprocessing and augmentation. The suggested model performance is assessed depending on the parameters like precision, recall, mAP50 and mAP50-95.