I am currently running the oxygen treatment model from nvidia clara with custom images and EHR data.
I do not completely understand why do I have to include the features for the outcomes, this being FEAT_ED_OD, OUT_OD_24H, OUT_OD_72H ,OUT_DEATH and OUT_DEATH_TIME. What do they mean and how can I obtain them? Because if they in fact are the time of death of the patient, for example, I wouldn´t need his oxygen device estimation, as I already know when he died or will die.
Also, even though I change the image, for the same EHR features I get the same prediction result. Is that right?
Thank you in advance

Those values are only needed to prepare the datasets for training the model. If you just want to use the pre-trained model from NGC for inference, you can put RA for all outcomes (values with OUT_*) in order to prepare data lists json files for inference only. However, FEAT_ED_OD is the oxygen device used when the patient was admitted to the emergency department and is a feature required by the model as input. Please see tutorial/EXAM_EHR_FEATURES_AND_OUTCOMES.xlsx for more details.