pyradiomicsbatch multiprocessing

[sohrabtowfighi]

Can an argument be added (eg 'n_jobs=-1' for all cores) to pyradiomicsbatch which uses multiprocessing to distribute the computations across cores?

[pieper]

We've discussed parallelism but generally consider it should be handled at a higher level of abstraction. For example if you run a big set of jobs on a cluster you would use the job queuing system to manage job allocation.

[sohrabtowfighi]

Isn't that the purpose of pyradiomicsbatch

[fedorov]

The specific meaning of batch that we had in mind is to support processing of multiple datasets in one run, not to distribute that processing across computational resources.

[JoostJM]

@pieper, the parallelisation we discussed can indeed be applied in a batch (i.e. one patient, one core). We've been working with a specialised script that does this, but it was optimized for our dataset. I'll see if I can adapt one to be more general.