Identifying Various Types of Brain Tumors using Deep Neural Network based Image Features

MRI images are typically employed for the clinical examination of most brain tumors. The identification and categorization of brain tumors is a crucial and demanding task in the realm of oncology. Inaccurate and belated diagnoses can result in fatalities. Hence, an accurate and efficient method for automatic detection of brain tumors is essential. An effective and viable solution involves employing a deep learning approach that holds the ability to adapt and learn data. However, the majority of research conducted on this has been based on a small dataset, possibly limiting its practical use. This study addresses the limitations of small datasets in previous research by utilizing a substantial dataset of 7023 MRI scans. Various deep learning techniques, including Convolutional Neural Network (CNN), VGG16, InceptionV3, and EfficientNetB3, were implemented. The overall accuracy of EfficientNetB3 is 99.84%, along with a precision, recall, and F1- score of 99%. According to this research, EfficientNetB3 outperforms the CNN, InceptionV3, and VGG16 models. EfficientNetB3 also has the lowest number of total parameters compared to others, at 11.7 million. This suggests that the proposed model can detect and accurately classify brain tumors very efficiently. This is crucial in real-time classification and may prove to be very handy for medical experts to confirm the diagnosis.