Automatic Detection of Microaneurysms in OCT Images Using Bag of Features

Table 1

Evaluation of the proposed method.

Classification method		Evaluation criteria
		Accuracy	Sensitivity	Specificity	Precision
1	BOF+MLP	96.33%	97.33%	95.4%	95.28%
2	BOF+Gaussian SVM	91.63%	88.27%	95.02%	94.65%
3	BOF+linear SVM	91.13%	86.52%	95.78%	95.40%
4	BOF+KNN	84.19%	73.14%	95.28%	93.95%
5	BOF+Naïve Bayes	87.75%	85.65%	89.91%	89.48%
6	PCA+MLP	92.52%	93.93%	92.22%	92.36%
7	PCA+Gaussian SVM	83.38%	85.53%	81.28%	82.16%
8	PCA+linear SVM	80.94%	77.77%	84.15%	83.19%
9	PCA+KNN	57.32%	14.77%	99.39%	96.32%
10	PCA+Naïve Bayes	76.13%	76.14%	76.15%	76.14%