BRAIN TUMOR DETECTION USING DEEP TRANSFER LEARNING
Mr. PANDIYAN M1, Ms. JENISHA E2, Ms. MADHURANJANI G3, Ms. BRINDHA DHARSHINI S4
1Assistant Professor Level – III, Information Science and Engineering & Bannari Amman Institute of Technology
2UG Scholar, Computer Science and Engineering & Bannari Amman Institute of Technology
3 UG Scholar, Computer Science and Engineering & Bannari Amman Institute of Technology
4UG Scholar, Computer Science and Engineering & Bannari Amman Institute of Technology
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - One of the leading causes of cancer-related death in humans is brain tumors. The process of diagnosing brain tumors and helping patients choose their course of treatment depends heavily on early identification. An innovative medical imaging method called Magnetic Reasoning Imaging (MRI) aids radiologists in locating the tumor spot. Manually testing the MRI pictures is a laborious process that needs experience. The development of computer-assisted diagnosis, machine learning, and deep learning in particular has made it possible for radiologists to more accurately diagnose brain tumors these days. The project employs Convolutional Neural Networks (CNNs) the classification of MRI images targeting four distinct types of brain tumors: gliomas, meningiomas, pituitary adenomas, and no tumor. By harnessing the power of deep learning, the model aims to enhance the precision and speed of tumor identification, contributing to early diagnosis and improved patient outcomes. But the traditional CNN has limitations, such as the need for large labeled datasets, employment of more layer of neurons, need for large computational resources, more time consumption and the risk of overfitting. To overcome these challenges, the project leverages transfer learning, a technique that allows the model to leverage knowledge gained from pre-trained networks on large datasets. This project involves the use of EfficientNetB0 architecture for the detection of brain tumor. This methodology demonstrates superior classification performance compared to conventional Neural Network. Therefore, this project presents a robust framework for brain tumor detection utilizing deep transfer learning, offering a promising avenue for improving diagnostic accuracy in the realm of neuroimaging. The adoption of transfer learning mitigates the drawbacks associated with CNNs, propelling the development of effective and reliable method for early detection and intervention in brain tumor cases.
Key Words: MRI, CNN, Brain tumor, Transfer Learning, EffiecientNetB0.
