Hi

It is slightly unclear what kind of help you need? Do you have problems reading in the ground truth from XML files or do you want to extract your own points from images (ie segment the images...) or...

If the probelm is of the first kind, I suggest to look at matlab function xmlread and write own script to get the relevant data. If the latter maybe some simple thresholding is enough or if not then maybe some graph-cut approach might be what you need or...

Thanks Sami, 

I don't know how can extrat countor index in MATLAB.

Technical Report of database said: "for each ROI, a list of contour points is presented between the tags and ."

But in XML file, when I open it via notepad, I found like that:

-

(0.000000, 0.000000, 0.000000)

Point_px

-

(3187.399902, 1992.239990)

In MATLAB when I use xmlread, answer is: "[#document: null]"

How can I extract Mass region and process it or analyse it's feature with contours that database determine?

I enclosed an example xml file.

Hi,

I haven't used myself plist xml-files, but it John Iversen has released some working code to read data in from such files (and I tested that it works). You can find the original contribution from the Mathworks page: [http://www.mathworks.com/matlabcentral/fileexchange/7868-use-mac-os-x-xml-plists-with-matlab/content/xmlplist/XMLPlistToStruct.m]  and the code is attached also on this post as a zip-file.

Just call the loadXMLPlist function with appropriate filename and you should get an struct containing cells of info in the file. eg:

> dat = loadXMLPlist('20586908.xml');

> dat.Images{1}

ans =

   ImageIndex__INTEGER: 0

   NumberOfROIs__INTEGER: 42

   ROIs: {1x42 cell}

> dat.Images{1}.ROIs{18}

ans =

   Area: 0.4281

   Center: '(0.000000, 0.000000, 0.000000)'

   Dev: 98.5360

   IndexInImage__INTEGER: 17

   Max: 1939

   Mean: 1.6520e+03

   Min: 1274

   Name: 'Mass'

   NumberOfPoints__INTEGER: 27

   Point_mm: {1x27 cell}

   Point_px: {1x27 cell}

   Total: 14633524

   Type__INTEGER: 15

You may have to find the suitable type or name in the ROI structs that you need. I hope this gets you going...

very thanks for your help,Sami

Your answer has also been of help to me, Sami, thank you very much! :)

Wen I want to use "Point_px" of Images in order to extract the region of lesion, But Values are not integer, for example (150.203003, 2208.959961)

Do you know how can I show or process just the region of mass?

Hello. I am working with INbreast database too. Anyone knows the specification of XML format for contours? I dont know whats means Point_px and Point_mm.

I am waiting for your help.

I'm working with this database too. I used xml read of Jarek Tuszynski it work perfectly. But now I don't know how to interpret the informatión what exactle means center and point. I hope somebody can help us.

Hi, I can't download the INbreast database, anybody can help me for giving me a download link? Many thanks.

Hi Lei,

I think you cannot download it directly, please request Breast Research Group for the permission. They will provide you the download link and password.

You can sent the email to below address for asking the permission.

[email protected]

 Is there any python code available to convert XML files to contour masks? Or can we directly find somewhere contours masks? Thanks you a lot !

Hi, I am working on the INBreast dataset. I don't know how to extract the contour information from the XML files. What dose the point_px mean? Are they the coordinates of the contour of findings?

Hi, iI am also working with Inbreast . In the pdf joined with the database it is said in page 7 that

The annotations were saved in XML format with the following structure:

1. The tag NumberOfROIs followed by an integer that indicates the number of annotations present in the image;

2. For each ROI, there is a tag Area followed by the value of the area of the current ROI,

3. the tag Center followed by a the coordinates of the point

in the centre of the ROI, the tag Name followed by the type of finding (mass, calcification, distortion, spiculated region) and some other general information’s about the ROI;

It is clear that each Roi is identified by its Center then other information like contour points

The problem is that the contour points exists in the xml file but the center is always 0.0000 0.0000 0.000 for all files. ANY HELP PLEASE

Hello, for those who need to extract the ROIs from the images, Wentao Zhu wrote a code in MATLAB that works fine. I'm attaching his code with a few minor changes, but anyway follow his GitHub link as well.

Note: this code uses the XML files.

https://github.com/wentaozhu/inbreast

Hi, I have a python script that generates binary masks (bool numpy arrays) out of the xml file if anyone is interested.

It runs in python3, but conversion to python2 should take about a minute.

Link: https://gist.github.com/jendelel/3a8e768a8eb9345d49f2a82d02946122

Feel free to ask questions.

Please, How can i download this data?

You need to contact the group by email in the last line of this link.

http://medicalresearch.inescporto.pt/breastresearch/index.php/Get_INbreast_Database

Actually you could use the binary masks provided by the Dataset in the Extras folder. Then use for instance region props from skimage in python to get the ROI and crop the image. If need any help I can provide some code.

In fact I double checked this proposed method with getting the roi from the XML file and the difference is just one pixel

I use the https://github.com/wentaozhu/inbreast but I met the problem:

Error in visinbreast (line 40)

mask(boundbox == 255) = 255;

I don't know how to solve it.

Would anyone can solve my problem?

Thanks!

Felipe André Zeiser I tried the code visinbreast.m and it is working fine for me. The complete database contains 410 images. Among them, 67 are with no findings. So rest of the 343 images should come with ground truth and mask. However, I found that in these 343 images, several images don't have any masking. Can anyone please help?

Hi Susama Bagchi, are you still having difficulties with the base? I can check if I still have the same one saved in my notebook. Does the code have any errors? And sorry about my delay, but I don't check Researchgate regularly.

Felipe André Zeiser Hi, thank you for your reply. The images that were without masking are with no mass and only calcification. Is the ground-truth for calcification images available?

Hi Susama Bagchi , I also need the ground-truth (ROIs) for calcification images, did you obtain it?

Thank you.

Hi Mario Andres Rodriguez Osorio , in INbreast dataset I found that ground truth for calcifications are there as .png files under folder extras.

Hi, I want to ask whether the ground-truth results for mass segmentation are also under folder extras as .png files? Also, is there any way to show the mass segment result in the original file (I mean, like combine the original image and result image together)?

Thank you so much!

Susama Bagchi Hey there, I was wondering from which source have you got your dataset? I have downloaded mine from this source https://drive.google.com/file/d/19n-p9p9C0eCQA1ybm6wkMo-bbeccT_62/view , and there is no extras folder there, just an XML file for the data and it doesn't work so well

Ahmed Khalifa Hi, I downloaded the dataset from the given link provided by Dr. I. C. Moreira who is the in-charge of this dataset. She is very prompt in her response and helpful.

Susama Bagchi Well thanks alot for your quick reply aswell, I will be contacting her soon

With function loadXMLPlist(fname)

These utilities enable the conversion of data written by programs for use in MATLAB (and vice versa).

https://www.mathworks.com/matlabcentral/fileexchange/7868-use-mac-os-x-xml-plists-with-matlab

You can use the function loadXMLPlist

please, put the download link containing the binary masks here