Blood Cell Identification

Segment and classify normal and burr cells

Start with an image of red blood cells:

In[1]:=

img=

;

Separate foreground and background with an automatic threshold:

In[2]:=

binary=ColorNegate@Binarize[img]

Out[2]=

Segment overlapping and touching blood cells by identifying every cell as a peak in the distance transform:

In[3]:=

dt=DistanceTransform[binary];ImageAdjust[dt]

Out[3]=

Find the centers of each cell using a maxima detector:

In[4]:=

marker=MaxDetect[dt,1]

Out[4]=

Segment each individual blood cell and color them all differently by performing a watershed segmentation to capture the scope of each cell:

In[5]:=

segments=ImageData[binary,"Bit"]WatershedComponents[ColorNegate@dt,marker];Colorize[segments]

Out[5]=

Create a mask for all cells:

In[6]:=

mask=Image[segments,"Binary"]

Out[6]=

Highlight the tentacles of burr cells via a morphological top-hat transformation that extracts all shapes smaller than a disk of radius 8 pixels.:

In[7]:=

spikes=TopHatTransform[mask,DiskMatrix[8]]

Out[7]=

Extract burr cells by selecting cells of a certain size that exhibit tentacles of more than 5% of their total area:

In[8]:=

burr=Image[Last@SelectComponents[{spikes,segments},(500<#Area<1500∧0.05<#Mean<1)&],"Bit"];

In[9]:=

HighlightImage[img,{Green,burr}]

Out[9]=

Related Symbols