<< Chapter < Page Chapter >> Page >

Segmentation block diagram

segmentation

Green channel isolation

As all images were in RGB format, green channel isolation simply involved taking the green layer of the matrix containing the image, returning a grayscale image based on the green intensity of the original image. Recall that this was done to match the white blood cell stains.

Thresholding

Thresholding was performed on the grayscale images with MATLAB's multithresh function using Otsu's method, separating the image with 5 different thresholds into 6 different levels. The 2nd most exclusive threshold was used to select nuclei, and the 4th most exclusive threshold was used to select cell bodies, returning two rough binary masks.

Cleaning

These rough masks were then processed to remove debris, fill holes, and smooth edges.
  • Debris was removed by deleting all objects with a centroid within a fixed distance from the margins or smaller than a fixed number of pixels. The objects and their centroids were found using MATLAB's regionprops function.
  • Holes were filled using MATLAB's imfill function and imclose function.
  • Since debris overlapping with the cell or nucleus of interest were not removed with the previous debris clearing function, a binary erosion was performed on the mask, the debris clearing function was repeated, and a binary dilation was performed on the mask, using MATLAB's imerode and imdilate functions, respectively. This erosion and dilation was also used to smooth the mask's edges. In future iterations, the hole filling function imclose might be moved after this second debris clearing function, as imclose might interfere with imerode's ability to separate overlapping objects.

Masking

Finally, the binary mask was applied to the grayscale image in an element-wise matrix multiplication, returning a grayscale image of just the cell or nucleus in question.

Get Jobilize Job Search Mobile App in your pocket Now!

Get it on Google Play Download on the App Store Now




Source:  OpenStax, Automatic white blood cell classification using svm and neural networks. OpenStax CNX. Dec 16, 2015 Download for free at http://legacy.cnx.org/content/col11924/1.5
Google Play and the Google Play logo are trademarks of Google Inc.

Notification Switch

Would you like to follow the 'Automatic white blood cell classification using svm and neural networks' conversation and receive update notifications?

Ask