The extclsshpinfo.py "find the match from LIDC-IDRI annotation part"， I think there are some mistake

uci-cbcl / DeepLung

WACV18 paper "DeepLung: Deep 3D Dual Path Nets for Automated Pulmonary Nodule Detection and Classification"

GNU General Public License v3.0

153 stars 54 forks source link

The extclsshpinfo.py "find the match from LIDC-IDRI annotation part"， I think there are some mistake #4

Closed mabihuan closed 6 years ago

mabihuan commented 6 years ago

Dear wentao,

When I read the extclsshpinfo.py, I find the "find the match from LIDC-IDRI annotation part" make some mistake.

First，the 71th line in the extclsshpinfo.py，voxcrd[-1] = sliceim.shape[0] - voxcrd[0]，it aim to do filp for z，so I think it should be voxcrd[0] = sliceim.shape[0] - voxcrd[0].
Second，the 154th~156th lines， voxcrd for z, y, x, but lidcant for x, y, z. so I think it should be dist = math.pow(voxcrd[0] - lidcant[5], 2) # z dist += math.pow(voxcrd[1] - lidcant[4], 2) # y dist += math.pow(voxcrd[2] - lidcant[3], 2) # x

wentaozhu commented 6 years ago

For the second, I think you are right. I have corrected it and now you can get better performance.

For the first, I am very sorry that have forgotten why I used the subtract. Maybe you can check again and if it is wrong again, I will update it.

Thank you very much for the help and interest!

mabihuan commented 6 years ago

I run the code modified by voxcrd[0] = sliceim.shape[0] - voxcrd[0] , get the result below , I print the LUNA16 points z,y,x in 1th to 3th column and LIDC points z,y,x in 4th to 6th column that match. The last column is the mindst. _20180615093301

The next is your code voxcrd[-1] = sliceim.shape[0] - voxcrd[0] result. _20180615102641

As you can see the result voxcrd[-1] = sliceim.shape[0] - voxcrd[0] did not get the real match between LUNA16 and LIDC.

wentaozhu commented 6 years ago

Thank you very much for the debug information! I have corrected it.

mabihuan commented 6 years ago

Dear wentao,

I want to ask you another question base on this code error. According to the error, the label you get for LUNA16 from LIDC will be wrong. Then the result of the nodule classification moudle in your paper DeepLung is wrong. Can I understand that？

wentaozhu commented 6 years ago

Dear Bihuan,

I do not have time to rerun the experiments. But in my first version, I just used the y axis to measure the distance. Maybe it contains some inaccurate cases, but it is not wrong.

Hope you can get better results!

Good luck, Wentao

anruoxi123 commented 5 years ago

Did you have the following error running the extclssphinfo.py file? I hope you can give me some advice! afec094bd61851dc4e06f2e176e4af5

CDeepA commented 5 years ago

Did you have the following error running the extclssphinfo.py file? I hope you can give me some advice!

Hey @anruoxi123 . Did you find a solution to the problem?

wentaozhu commented 5 years ago

LIDC-IDRI has more images than LUNA16. Try to remove extra images in LIDC-IDRI.

CDeepA commented 5 years ago

Dear wentaozhu.

Thank you for your help, your were right. An important note, the folders also had to be renamed after study id and series id.