search

DeepRegNet / DeepReg

Medical image registration using deep learning
Apache License 2.0
569 stars 79 forks source link

Train with Big CT file #814

Closed JiYuanFeng closed 3 years ago

JiYuanFeng commented 3 years ago

Hi! Thank you for your nice work. I would like to train the model with my own dataset. The dataset consists of big CT data (around 370512512). It's ok to train the model by sliding window mechanisms? for example, crop a subcube of 128128128 to train the model. Or it will be better to resize the original data into a smaller one to train the model, and resize the generated transform to warp the original data finally? Could you give me some advice about how I can do it or/and what I need to modify? Thank you for your advice

YipengHu commented 3 years ago

Hi @JiYuanFeng Thanks for the message. I'm afraid DeepReg is not supporting patch-based registration yet - i guess this is what you meant by sliding window mechanisms? So you are only left with resizing for now, which obviously is a highly application-dependent pre-processing step. Happy to discuss further if you share more details with the data, application etc.

JiYuanFeng commented 3 years ago

Hi,Thank you for your useful reply!!! It help me a lot.