Lightweight method to read StudyInstanceUID from DICOM header

[aisosalo]

What would be the most lightweight method to read StudyInstanceUID from DICOM header? I tried to find examples online, but wasn't able to spot one.

[darcymason]

Converting to discussion...

[darcymason]

The simplest answer is to call dcmread with stop_before_pixels=True, then access normally, i.e. ds.StudyInstanceUID.

You could also set specific_tags in the dmcread call to that one tag, but I doubt that would actually be any faster, certainly not by much if it is.

This kind of question often comes up - the problem is that DICOM data elements are random length and have to be read sequentially, there is no good way to jump to a specific one.

[mrbean-bremen]

In principle, there is also the possibility to stop after you have read the tag you want. dcmread does not have an argument that allows that, but if you want to try this, you could use read_partial instead using the stop_when argument (and additionally specificTags as mentioned):

    def stop_condition(tag: BaseTag, vr: Optional[str], length: int) -> bool:
        return tag > 0x0020000D  # StudyInstanceUID

    with open(file_name) as fp:
        ds = read_partial(fp, stop_when=stop_condition, specific_tags=[BaseTag(0x0020000D)])

That would ensure that the tags before StudyInstanceUID are just jumped over, and reading will stop after getting it.
I doubt that this would improve the performance significantly compared to the solution above, but that is the fastest I can think of.

[aisosalo]

In principle, there is also the possibility to stop after you have read the tag you want. dcmread does not have an argument that allows that, but if you want to try this, you could use read_partial instead using the stop_when argument (and additionally specificTags as mentioned):

    def stop_condition(tag: BaseTag, vr: Optional[str], length: int) -> bool:
        return tag > 0x0020000D  # StudyInstanceUID

    with open(file_name) as fp:
        ds = read_partial(fp, stop_when=stop_condition, specific_tags=[BaseTag(0x0020000D)])

That would ensure that the tags before StudyInstanceUID are just jumped over, and reading will stop after getting it. I doubt that this would improve the performance significantly compared to the solution above, but that is the fastest I can think of.

To get the latter option running, importing BaseTag and Optional are required:

from pydicom.tag import BaseTag
from typing import Optional

read_partial is located within filereader.

That should be all, right?

[mrbean-bremen]

Yes, as far as I can see - I didn't test it. You can omit the typing, of course.

[aisosalo]

Thank you for your time. I chose to use the method suggested by @darcymason. Here is the context: 🎗️ [16-bit post-processing] RSNA BC Detection.