Accurate Deep Learning Models for Predicting Brain Cancer at begin Stage

The objective of this research is to explore and compare the performance of several Deep Learning (DL) models and identify the most accurate classification model for predicting brain tumors using MRI images. The research utilizes the dataset of 450 MRI images which include healthy cases, cases with grade 1 & 2 benign tumors, and cases with grade 3 & 4 malignant tumor cases. The dataset is further divided into the training, validation, and testing sets. Each model is then trained and validated on the training and validation sets and further tested on the testing set and the overall performance is assessed and compared. The results demonstrated unique trends among the models, where CNN and ResNet50 have consistently performed the best with the highest accuracy and least data loss. VGG16 and VGG19 have also exemplified great results, although they utilised more epochs to achieve similar accuracy. Based on the results of the study, it is concluded that the appropriate DL architecture for tumor classification should be selected especially in medical fields. In general, CNN and residual networks showed the best performance and should be chosen when accurate tumor classification is the most important requirement. The potential application of the outcomes of the research can be applied in the field of medicine mainly for the identification, classification, detection, and prediction of various diseases.