Segmentation of Endothelial Cell Boundaries of Rabbit Aortic Images Using a Machine Learning Approach

Table 3

Comparison between SVM kernels with varying 𝛾 in steps of 0.01 in each model. The table is obtained from four different images representing a total of approximately 4×10⁶ pixels and four different SVM Kernels. The number of training data pixels was 3844.

𝛾	Accuracy%
	RBF	Sigmoid	Polynomial	Linear
0.14	92.97	90.06	63.49	55.67
0.13	92.52	88.26	64.53	55.67
0.12	92.82	85.46	65.92	55.67
0.11	92.92	82.55	67.68	55.67
0.10	92.96	79.34	69.63	55.67
0.09	88.95	76.47	71.63	55.67
0.08	91.00	74.19	73.85	55.67
0.07	90.49	71.64	76.43	55.67
0.06	88.14	63.77	79.49	55.67
0.05	88.85	69.70	83.50	55.67
0.04	88.67	64.03	88.61	55.67
0.03	80.66	62.59	83.35	55.67
0.02	82.75	61.67	73.19	55.67