search

gsunit / Pectoral-Muscle-Removal-From-Mammograms

Algorithm to segment pectoral muscles in breast mammograms
35 stars 6 forks source link

Issues in removing pectoral muscle for my dataset #1

Open MudasserAfzal opened 3 years ago

MudasserAfzal commented 3 years ago

Hey there. I have tried your code but it is making some error. it's not giving me correct output as like your input images. I am using the mini-MIMA dataset. can you please help me out.

Error is: IndexError: list index out of range

dgvai commented 3 years ago

I also have found the same problem, the solution till I have figured out is to change the image resolution below 500px of width sorts of solving the issue as well as needed to change the sigma value inside apply_canny() function. By doing this changes, I could have removed pectoral muscles of 70% of my dataset. 30% still gives IndexError: list index out of range.

lolo1994a commented 2 years ago

i face the same problem with mini mias,any one find the solution for it?

MudasserAfzal commented 2 years ago

This problem is still not solved. So I used a different approach to my dataset where I split it into grids and only select those grids of the image which has a tumour or segmented ground truth. So through this way, I bypass this process and it helps me to decrease image dimensions as well.

lolo1994a commented 2 years ago

Can I know how you split to grid and get tumor or ground truth ?I try to use threshold segmented but not work and i use mini mias dataset

MudasserAfzal commented 2 years ago

sorry for being late. so basically you can do it by your choice, lets i choose a grid of 500x500 so now i start taking first grid then adding stride of 500 i keep convolving and divide the image. so after that i only select the ones in which pixel values are 1 otherwise i discard. similarly i do the same with the original feature image where i consider the same grid size and select the one where there exist 1 in corresponding ground truth.