Segmentation of Blood Smear


(Akshay) #1

Above is the Image of Blood Smear … I Need To Detect and Segment the No of RBC and WBC in the Image…
What Should I Use

  1. U-Net - Considered Best for Medical Images
  2. Faster RCNN
  3. Mask RCNN - Segmentation + Detection

I had Achieved Good Accuracy using the Patch Method in Dectecting the RBC and WBC but I Need To Find The size of WBC or RBC in Pixels …So The Patch Method Does Not Solve the Problem. it is a Kind of Segmentation Problem + detection + Knowing the area of the segment

Pls provide suggestions…


(Yash Katariya) #2

I think U-Net segmentation would work. You can refer the paper https://arxiv.org/abs/1505.04597


(Akshay) #3

hi yash

What About Knowing the area of the different segments…

Pls Suggest


#4

Could a simple image processing skill works?

segment by cnn->extract contour->count contour area

By opencv3,(by c++, it should be piece of cake to convert it to python)

vector< vector<Point> > contours;
findContours(img, contours, RETR_EXTERNAL,CV_CHAIN_APPROX_SIMPLE);
cout << "Area " << ": " << contourArea(contours[1]) << endl;

(Akshay) #5

if somebody can help with