Head and neck CT deformable image registration.

[nightandweather]

Hello thank you for sharing your work!, I'm interested in your paper and tried to compare Radiation Oncology TPS deformable image registration algo and state of the art DIR deep learning method.

I applied your oasis folder to my head and neck, rt structure to create data and train data set, and adjusted the number of classes of OASISBrindataset classes and device val functions.

these are the modifications to your code

data_code.zip

example of pkl data set

160192224 was cropped from the center of the ct data set, and rt structures include brainstem, parotid, cochlear, optic_chiasm, spinal cord, and ptv Each rt structure has a set number, and after being transformed by the model, it compares the values to each other in a dice score.

ct_image_label.zip

However, it seems that learning is not going well. Attached is a log file of the tensorboard

https://catholicackr-my.sharepoint.com/:u:/g/personal/dlrkdgns1324_catholic_ac_kr/Ef7-2YSro7dCtHTIBw0FeMcBikA56rnLFgC1fFelr1WO1g?e=q0C9k4

I would appreciate it if you could tell me which part I should fix. I'll be waiting for your answer

Thank you