International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 03 | Mar 2025
p-ISSN: 2395-0072
www.irjet.net
AUTOMATED BRAIN TUMOR DETECTION AND SEGMENTATION USING RESNET AND YOLO Verannagari Soumya1, B Nitish Kumar 2 , Mrs. K. Muthulakshmi 3 1B.Tech Student, Dept of Computer Science and Engineering, Bharath Institute of Higher Education And Research,
Tamil Nadu, India
2B.Tech Student, Dept of Computer Science and Engineering, Bharath Institute of Higher Education And Research,
Tamil Nadu, India Associate Professor, Dept of Computer Science and Engineering, Bharath Institute of Higher Education And Research, Tamil Nadu, India ---------------------------------------------------------------------***--------------------------------------------------------------------3
Abstract - Brain tumour represent a major threat to
errors, requires a lot of physical labour, and requires knowledge. Furthermore, tumour characteristics including size, shape, and location vary greatly throughout patients, making detection and segmentation difficult. Medical image analysis was transformed by deep learning, which made it possible to automatically and accurately diagnose diseases like brain tumours. By recognising patterns in images, Convolutional Neural Networks (CNNs) are especially well-suited to detect and categorise tumours. Using YOLO for real-time detection and ResNet-50 for feature extraction, the study aims to create an autonomous system for brain tumour identification and segmentation. Using MRI scans, the deep CNN ResNet-50 offers more complex and sophisticated features to distinguish tumours from healthy tissue. It is a good model for medical picture classification because of its residual learning, which aids in deep training without vanishing gradients. The real-time object detector YOLO is incredibly quick. ResNet-50 may be used to extract features from MRI images, and YOLO can be utilised for detection in order to improve the efficiency and accuracy of brain tumour identification. It will be evaluated on metrics of accuracy, precision, recall, IOU, and dice coefficient for accurate detection and segmentation after being trained on publicly accessible MRI data sets that comprise pictures of brain tumours. Improved diagnostic effectiveness, less dependence on human interpretation, and faster brain tumour detection to facilitate early intervention and treatment planning are the expected results. Automation may revolutionise the diagnosis of brain tumours, enabling radiologists to make diagnoses more quickly and accurately. Deep learning is being used in this study to improve computer-aided diagnosis (CAD) and pave the way for further advancements in medical AI. The findings of this study could result in trustworthy medical diagnostic tools that enhance clinical results and patient care.
human health, and their timely and accurate detection is indispensable for successful therapy and enhanced patient survival. Current diagnosis techniques based on manual analysis of MRI images are labour intensive, error-prone, and heavily dependent on the competence of radiologists. The goal of this project is to create an automated brain tumour detection and segmentation system based on deep learning algorithms, namely ResNet-50 and YOLO (You Only Look Once). ResNet-50, a deep convolutional neural network, is used for feature extraction, identifying complex patterns and structural information of brain tumour from MRI images. YOLO, a highly advanced object detection model, is used for real-time tumour localization and segmentation, allowing for quick and precise identification of infected areas. The combination of these two models improves the system's robustness and efficiency by providing both high accuracy and high processing rates. The model will be trained and evaluated on publicly available MRI datasets containing labelled brain tumour images. Performance evaluation will be carried out by metrics including accuracy, precision, recall, IOU (Intersection over Union), and Dice similarity coefficient to ensure correct detection and segmentation outcomes. Through the automation of the diagnosis process, this study hopes to decrease the burden of medical experts, reduce the delay in diagnostics, and increase the likelihood of early intervention among patients. Key Words: Brain tumour, MRI, Deep learning, ResNet-50, YOLO, Feature extraction, Convolutional Neural Networks (CNNs), Real-time tumor localization, Dice similarity coefficient, IOU, Precision, Recall.
1.INTRODUCTION If not discovered in their early stages, brain tumours are dangerous and have a high fatality rate. Because MRIs are non-invasive and have great resolution, they are the first imaging method utilised to diagnose brain tumours. However, traditional diagnosis depends on radiologists' visual interpretation, which takes a lot of time, is prone to
© 2025, IRJET
|
Impact Factor value: 8.315
2. LITERATURE SURVEY 1) Conventional methods for detecting brain tumors include traditional machine learning, radiologists' manual segmentation, and feature approaches like edge detection,
|
ISO 9001:2008 Certified Journal
|
Page 1172