Fig 1.
Samples of different brain tumor images.
Table 1.
MRI image dataset distribution.
Fig 2.
Proposed novel customized CNN architecture for brain tumor classification.
Fig 3.
Proposed optimized ResNet101 architecture for brain tumor classification.
Table 2.
Hyperparameter settings of the novel customized CNN.
Table 3.
Fold-wise training and validation accuracies of the novel customized CNN.
Table 4.
Hyperparameter settings of the optimized ResNet101.
Table 5.
Fold-wise training and validation accuracies of the optimized ResNet101.
Table 6.
Performance comparison between the novel customized CNN and optimized ResNet101.
Fig 4.
Validation confusion matrix of the optimized ResNet101 model.
Fig 5.
Testing confusion matrix of the optimized ResNet101 model.
Fig 6.
Roc curve illustrating the AUC for the optimized ResNet101.
Fig 7.
Validation confusion matrix of the novel customized CNN.
Fig 8.
Testing confusion matrix of the novel customized CNN.
Fig 9.
Roc curve illustrating the AUC for the novel customized CNN.
Fig 10.
Training progress of the novel customized CNN.
Fig 11.
Training progress of the optimized ResNet101.
Table 7.
Testing results comparison between the novel customized CNN and the optimized ResNet101.
Table 8.
Performance comparison of the novel customized CNN and the optimized ResNet101 with existing studies.