Belarus Tuberculosis Dataset Downloading Link

[zeeshannisar]

Can you please get me the downloading link for Belarus Tuberculosis Dataset. I am using the following link referred on the official website of Belarus but it's not working.
Dataset Downloading Link

[frapa]

The website appears to be down. That is really unfortunate if you are trying to reproduce the results. I think the only thing to do is trying to contact the administrators and see if they can provide the data.

[frapa]

I managed to get a copy of the data from a backup at my university. Unfortunately it is not the original data, but the pre-processed version of it (scaled to 512*512 and center-cropped).

The repository has been updated to include a belarus folder with a zip archive of this data.

All cases in the belarus dataset are TB-positive.

[zeeshannisar]

Thank you Dear.

[YalinZheng]

Dear Francesco, could you also share the manual segmentation of this dataset please?

[frapa]

There is no manual segmentation, we're talking about a pure classification dataset with image-wide labels.

[YalinZheng]

It seems there are segmentation here of 107 images, https://github.com/frapa/tbcnn/tree/master/belarus. However the size might be different to the images you shared. It would be wonderful if you could find the originals. Yalin

…

On Sun, Mar 15, 2020 at 8:39 AM Francesco Pasa ***@***.***> wrote: There is no manual segmentation, we're talking about a pure classification dataset with image-wide labels. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOI6FJRHNHGR5NN2PIHQGXDRHSH2XANCNFSM4IHBBLAA> .

[frapa]

@YalinZheng I think the Shanghai and maybe also the Montgomery datasets have lung masks, but the belarus dataset if you are interesed, but I'm pretty sure the belarus dataset does not have them.

[YalinZheng]

Sorry for the wrong link. The link to 107 segmentation is https://drive.google.com/file/d/1MinxZjUH06VMy5JXbirQxBAmmwpyc3Rk/view?usp=sharing. Yalin

…

On Thu, Mar 26, 2020 at 4:33 PM Francesco Pasa ***@***.***> wrote: @YalinZheng <https://github.com/YalinZheng> I think the Shanghai and maybe also the Montgomery datasets have lung masks, but the belarus dataset if you are interesed, but I'm pretty sure the belarus dataset does not have them. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOI6FJSJDDHZKM4LXSNLNQ3RJN7VHANCNFSM4IHBBLAA> .

[zeeshannisar]

I had trained a model to segment out the lungs in a chest X-ray using Shanghai and Montgomery and datasets and tried to segment out the lungs from Belarus but this didn't work for me.

[manaschuri11]

@frapa Can u also send the dataset containing cropped images of normal or TB negative X-rays

[eltonbrasil]

Hello Everyone,
When I try to run train.py I get the following error:

  img = scipy.misc.imread() 
  AttributeError: module 'scipy.misc' has no attribute imread

Someone can help me? Many thanks in advance.

[zeeshannisar]

@eltonbrasil the version is deprecated. try with installing the older version of scipy or use other libraries to read the image for example Pillow, cv2, imageio. For more details and practical help visit https://stackoverflow.com/questions/15345790/scipy-misc-module-has-no-attribute-imread.

[frapa]

The canonical way is replacing a scipy.misc.imread with imagio.imread. If you had the time to make a PR after that, it wouof be great!

[LinYJ19961119]

I managed to get a copy of the data from a backup at my university. Unfortunately it is not the original data, but the pre-processed version of it (scaled to 512*512 and center-cropped).

The repository has been updated to include a belarus folder with a zip archive of this data.

All cases in the belarus dataset are TB-positive.

Dear @frapa:
Thanks very much for sharing the precious images.I saw a folder of 422 images in the zip, however, there's an issue I want to consult with u:
Are the total 422 images in that folder all TB confirmed?

[frapa]

@LinYJ19961119 Yes, they are app confirmed TB-positive

[LinYJ19961119]

@LinYJ19961119 Yes, they are app confirmed TB-positive

@frapa Thanks for your answering. App confirmed means deep learning confirmed or physicians confirmed?

[frapa]

The data is conforme by physicians.

[LinYJ19961119]

@frapa Thanks!