Table 4. Feature extraction and lung parenchyma segmentation.

Algorithm 4 Feature Extraction Algorithm
1: Assume that the final four types of sample sets are {Si}, i = 1, 2, 3, 4, then
${\displaystyle \underset{i=1}{\overset{4}{\cup }}{S}_i=X,\underset{(1\le i,j\le 4,i\ne j)}{\forall }(S_i{\displaystyle \cap }{S}_j)=\varnothing }$
2: Calculate the average grayscale values φ(Si) for four sample sets using (9)
3: max(φ(Si)) → pleural tissue(S4)
4: Compute the centroid coordinates (xi,yi) and the coordinate variances ξ2(Si) of the three left sample sets {Si}using (10) and (11)
5: max(ξ2(Si)) → extrathoracic area(S3)
6: two left sample sets → the left(S1) and the right(S2) lung parenchyma
7: Traverse superpixel samples in {S1, S2}
8: For Xi,Xj ∈ S1 ∪ S2, If aj6 = ai6, Xi,Xj → N (The same CT image N whose slice number is ai6), Connect Xi and Xj to Nk.
9: Until all superpixel samples have been input.
10: Sequential output all Nk corresponding to each lung parenchyma image.