Table 2.

Different models with their strengths and limitations.

Reference	Model	Contribution	Limitation
[14]	Integrated model of CNN and Transfer Learning	Good classification accuracy on test data.	Very large and complex CNN model
[16]	Novel 3D CNN Method	Robust when training on one dataset and testing on another dataset.	Only designed for 3D images, and low classification accuracy
[36]	Autoencoder Deep Neural Network (ADNN)	Accuracy was improved	High Computation Complexity
[37]	Enhanced Approach using Residual Networks	High accuracy was achieved by considering small dataset.	Poor results on large dataset
[49]	Modified Deep Convolutional Neural Network	Computation complexity was reduced	Low classification accuracy
[60]	AlexNet, Vgg-16, ResNet18, ResNet34, and ResNet50	Used to classify five classes (normal, cerebrovascular, neoplastic, degenerative, and inflammatory)	Low classification accuracy
[61]	Color Moments and artificial neural network	Simple and very fast	Low classification accuracy
[43]	DWT, color moments, and artificial neural network	High accuracy	Good only on small dataset