Table 11 Comparative analysis of proposed work with previous works
Contribution | Features employed | Type of classifier (s) | Dataset | Technical environment | Accuracy (%) | k-Fold cross-validation method/data division |
|---|---|---|---|---|---|---|
Badža et al. [5] | Elementary features- model based | CNN (with two convolution layers) | T1-weighted contrast-enhanced MRI (Figshare dataset) | MATLAB | 96.56 | 60% data in training, 20% in validation, 20% in test and tenfold cross-validation |
Pashaei et al. [27] | Elementary features- model based | CNN (with four convolution layers) | T1-weighted contrast enhanced MRI (Figshare dataset) | – | 93.68 | 70% data in training, 30% in testing and tenfold cross-validation |
Gumaei et al. [23] | GIST features | FNN (feedforward neural network) | T1-weighted contrast enhanced MRI (Figshare dataset) | MATLAB | 94.23 | 70% data in training, 30% in testing and fivefold cross validation |
Afshar et al. [21] | Elementary features- model based | CapsNet | T1-weighted contrast enhanced MRI (Figshare dataset) | Keras package, with Tensorflow | 86.56 | – |
Rehman et al. [22] | Fine-tune/Freeze-AlexNet, GoogLeNet, and VGG16 | SVM | T1-weighted contrast enhanced MRI (Figshare dataset) | MATLAB | 98.69 | 70% of data in training, 15% for validation, and 15% in testing |
Abiwinanda et al. [28] | Elementary features- model based | CNN (with two convolution layers) | T1-weighted contrast enhanced MRI (Figshare dataset) | Keras package, with Tensorflow | 84.19 | – |
Proposed approaches | Elementary features-model based | Two CNNs | Brain tumor classification (MRI): four classes | Keras package, with Tensorflow | 96.47 | 90% data in training, 10% in testing |
95.63 |