Towards Automated Three-Dimensional Tracking of Nephrons through Stacked Histological Image Sets

Table 4

Results of the ANN and SVM on the test set of 712 examples. The 5 classes have been condensed into valid and invalid classes for final classification.

Classification algorithm	Predicted class	Target class		Performance indicators (%)
		Valid	Invalid	Accuracy	Precision	Sensitivity
ANN (threshold = 0.3)	Valid	492	32	93.82	93.62	84.61
	Invalid	12	176
SVM with RBF kernel (width = 5)	Valid	475	19	93.25	86.70	90.86
	Invalid	29	189