Published Paper Details:

In the fight against brain cancer, accurate brain tumor detection is crucial for early diagnosis and effective treatment. This study explores various machine learning and deep learning techniques to achieve this goal using MRI scans. We investigated the effectiveness of several methods: Convolutional Neural Networks (CNNs) achieved an impressive test accuracy of 86.27%. This demonstrates their ability to learn important features directly from MRI images. Multilayer Perceptrons (MLPs) were explored in two ways. A standalone MLP trained on features extracted using Principal Component Analysis (PCA) reached an accuracy of 76.47%. We also experimented with using an MLP in a transfer learning approach with InceptionV3 for feature extraction. This approach yielded results to be discussed alongside the standalone models. We also compared the performance of several other machine learning techniques alongside the MLPs and CNNs. A standalone MLP, trained on its own without any transfer learning, achieved an accuracy of 52.94%. We also evaluated the VGG16 convolutional neural network, which reached an accuracy of 70.58%. Logistic regression, a common statistical method, yielded an accuracy of 62.74%. Random forest, an ensemble learning technique that combines multiple decision trees, achieved an accuracy of 72.54%. Ada boosting, another ensemble learning method, performed quite well, reaching the highest accuracy (74.50%) among all the non-deep learning models. For other machine learning models, Naive Bayes achieved an accuracy of 68.62%, while SVM (Support Vector Machine) reached 60.78%. Similarly, a decision tree model resulted in an accuracy of 68.62%. Bagging, another ensemble technique, yielded an accuracy of 66.66%. Interestingly, a hybrid model that combined the pre-trained VGG-16 and InceptionV3 models achieved an accuracy of 68.92%.The results reveal that Convolutional Neural Networks (CNNs) were the most successful method, achieving the highest accuracy (86.27%) for brain tumor detection in MRI scans. This suggests that CNNs are particularly adept at learning the critical patterns hidden within the MRI image data.