Training and inferencing at a particular resolution

[Rajesh-ParaxialTech]

Hello
Suppose I want to train a lung nodule detection model using

https://github.com/Project-MONAI/model-zoo/releases/download/hosting_storage_v1/lung_nodule_ct_detection_v0.5.9.zip. Say, i want to train at a resolution of 0.7 mm x 0.7 mm x 1.0 mm. using

https://github.com/Project-MONAI/tutorials/blob/main/detection/luna16_prepare_images.py

and

https://github.com/Project-MONAI/tutorials/blob/main/detection/luna16_prepare_env_files.py.

Right ?

Does this imply :

1.   I must resample my training dataset to this resolution and then involve it in training.
   

2. while inferencing also i must resample any test CT scan image to this resolution (0.7 mm x 0.7 mm x 1.0 mm)  and only then subject the CT scan to inference, for best performance ?
   

Thank you
Rajesh

[KumoLiu]

Hi @Rajesh-ParaxialTech, you can use a custom resolution for training without any issues. However, please note that if you plan to use the pretrained weights from the bundle, the results may not meet your expectations due to the resolution mismatch.
Thanks.