International Journal of Science and Technology (IJST)

MRI-based Detection, Classification and Grade Prediction of Brain Tumor using Transfer Learning and Support Vector Machines

Brain Tumors detection, classification and grade prediction according to WHO standards using MRI images remain a highly challenging task. Conventional machine learning and deep learnings models often fails to perform well when the dataset is very small, inaccurate and imbalanced, which is very common in medical imaging. To overcome these limitations, the proposed approach introduces an MRI- based framework for brain tumor detection, classification and grade prediction using transferring learning and support vector machines. MRI is the most widely used imaging modality for brain tumor detection due to its excellent soft-tissue contrast and non-invasive nature. This experiment focuses on the advancement of an intelligent framework for detection, classification and grade prediction of brain tumor using transfer learning techniques. Initially, a basic convolutional-based neural networks is used to detect the MRI images into tumor-positive and tumor-negative groups. Once the presence of tumor is confirmed, a classification technique is applied to find the types of tumors. In the grade prediction stage, machine learning classifiers such as Support Vector Machine are employed to determine tumor aggressiveness. The proposed approach aim is to improve the diagnostic accuracy of 95% or above, support the clinical decision-making and reduce the human intervention. The proposed framework is the combination of detection, classification and grade prediction. Tumor detection using a basic CNNs model, classification using DenseNet121 and tumor grade prediction using ML classifier. The experimental analysis provides the reliability, stable performance, making the system suitable for the real-world application.