Differential Evolution and Multiclass Support Vector Machine for Alzheimer’s Classification

Table 1

The proposed DE-MSVM method performance on axial slice images.

Method	Type	Advantages	Limitations
CNN models [11–14, 16–18, 20, 21]	Deep learning method	Feature extraction is efficient, and feature learning process helps to provide reliable classification.	Overfitting problem in the training and validation.
SVM [15, 19, 22–25]	Machine learning model	SVM model has the capacity to handle the high-dimensional dataset.	SVM model has lower efficiency in learning the feature differences, which affects the sensitivity and specificity.
Fuzzy logic [26]	Fuzzy model	Fuzzy model updates the rules to improve the classification process.	Fuzzy model is highly sensitive to outliers and has lower efficiency in feature learning.
GWO-DNN [27]	Deep learning	GWO method selects the parameter for the DNN to improve the classification performance.	GWO method is easily trapped into local optima, and the DNN model suffers from an overfitting problem.