Data Augmentation of 3d medical images using SimpleITK

(Jayant Prakash) #1

Hello everyone,

I am new to this amazing library. I am working on classification of lymph nodes using deep learning, therefore, I need to do data augmentation of 3D ROIs (of shape 32X32X32) like 3d rotation, 3d translation along a random vector, etc. I tried to use Transforms like Euler3D but it’s showing error when I use numpy 3d array. In examples Euler3d and other transforms are applied on data points.

Please help me figure out how to apply those useful transform to a 3d numpy array. Transforms. I think this topic will also help others in my domain.

Thanking You

(Ziv Yaniv) #2

Hello @jayant_prakash,

You should not be working directly with augmentation of numpy arrays, rather use SimpleITK images. I believe this Jupyter notebook does what you want, data augmentation in 3D.

If not, please be more specific with respect to what you want to do and we’ll try our best to help.

3 Likes