Is there an export function for the annotated voxel coordinates with the various labels used for machine learning?

[SiJing Feng]

Hi I am interested to use ITK SNAP for machine learning project. I am just wondering if ITK SNAP can export the annotated images with its labeled coordinates with the corresponding labels in either CVS or excel format? 

If you or someone from your team can help with above questions it would be much appreciated. 

Kind regards, 

Sijing

[Paul Yushkevich]

Could you explain in a little more detail what you would like to do? What kinds of images and how do you intend to label them? What are you labeling?

Best

Paul

--
You received this message because you are subscribed to the Google Groups "itksnap-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to itksnap-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/itksnap-users/c916da00-68c5-47f5-ba31-47b7eab63516%40googlegroups.com.

--

Paul A. Yushkevich, Ph.D.
Associate Professor
Penn Image Computing and Science Laboratory
Department of Radiology
University of Pennsylvania

[SiJing Feng]

Hi Paul, 

Thank you for the reply. I’m currently working on a project and testing a high ranking Kaggle competition algorithm for recognising pneumothorax on chest x ray. I’m trying to get the dataset from annotation platform to be in the same format as the training dataset from the Kaggle competition. This will allow easier feeding of the dataset into the algorithm, which was built on Kaggle dataset. The Kaggle dataset is in csv format, with image ID in one column, and negative pneumothorax a value of -1, and positive pneumothorax a series of encoded pixel values. This binary mask I wonder if is something that can be outputted from ITK SNAP? 

Kind regards,

Sijing 

To view this discussion on the web visit https://groups.google.com/d/msgid/itksnap-users/CAPR4OGN29%3DnspcV5xCUh5Ka7C5vDrYU%3D5ZZpAi4b3zHqETmykg%40mail.gmail.com.

[Paul Yushkevich]

Yes, you can save the segmentation as a binary mask (Segmentation->Save) and you can do many different manipulations using c3d (itksnap.org/c3d). I think what you are trying to do should be possible although I still don't fully understand what the format of the CSV is (a list of voxel coordinates?)

One possible command for interfacing with machine learning algorithms is -export-patches

Paul

To view this discussion on the web visit https://groups.google.com/d/msgid/itksnap-users/CAAA52fBmivKSJb7OR0jgMUbEzXP%2BB8jJB2uKvHgMQzpR%2BR8qzg%40mail.gmail.com.

[SiJing Feng]

Hi Paul, 

I have attached a copy of the data format I was hoping to have the annotation export to be like. you can see the data are in 2 columns, one with the image study ID, and second column with the "encoded pixels", which I am not sure if represents pixel coordinates, or intensity or area. This dataset format was an output from MD.ai online annotation platform. I am just wondering ITK SNAP could output something of this format from my dataset of DICOM chest X ray images.

Kind regards, 

Sijing

To view this discussion on the web visit https://groups.google.com/d/msgid/itksnap-users/CAPR4OGOiBkgGMuYaJGMX7ioUGUEct0a_RUnCVmp01T3tmn7Fcw%40mail.gmail.com.