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

Table 1

Six statistical features generated from the 3×3 intensity neighborhood of a candidate pixel.

Feature	Formula
Median	Med(𝑥,𝑦)=med(I√(𝑖)√𝑛×𝑛(𝑥,𝑦))
Range	Range(𝑥,𝑦)=(Max(𝑥,𝑦)−Min(𝑥,𝑦))
	whereMax(𝑥,𝑦)=max(𝐼√(𝑖)√𝑛×𝑛(𝑥,𝑦))
	Min(𝑥,𝑦)=min(𝐼√(𝑖)√𝑛×𝑛(𝑥,𝑦))
Energy	𝐸(𝑥,𝑦)=∑^𝑛𝑖=1𝐼^2√(𝑖)√𝑛×𝑛(𝑥,𝑦)
Second, third and fourth order moments	𝑀𝑟=1/𝑛∑^𝑛𝑖=1(𝐼√(𝑖)√𝑛×𝑛(𝑥,𝑦)−𝜇(𝑥,𝑦))^𝑟
	𝜇(𝑥,𝑦)=1/𝑛∑^𝑛𝑖=1𝐼√(𝑖)√𝑛×𝑛(𝑥,𝑦),
	𝐼(𝑥,𝑦) is the image, 𝑟=2,3,4: second, third and fourth order moments,𝑛 is the number of nearest neighbors (in our case this is 9), and 𝑖 is the candidate pixel