Empirical Driven Automatic Detection of Lobulation Imaging Signs in Lung CT

Algorithm 1

Template matching based lobulation detection.

Input: Training data set

/ Obtain the templates /

Select typical lobulation ROIs at 8 different orientations (45 degree interval approximately)

for  each selected lobulation ROI

  for  length-width ratio   do

    Obtain templates by transforming lobulation ROI with length-width ratio

  end

  for  scaling factor   do

    Obtain templates by zooming lobulation ROI with scaling factors

  end

end

/ Lobulation detection /

for  each Sliding Step

  for  each template

    Compute the NCC matrix between and all Sliding Patches in target image;

    Find the max value in NCC, and record its location ;

  end

  Sort recorded locations by the NCC value related to it;

  ;

  Label the ROI located at as lobulation region for current sliding step, ;

end

Output: Best-matched regions    for each sliding step