iterall() not recursing into all sequences

[marcelzwiers]

I'm a scientist writing an application for other scientists, i.e. I'm not a DICOM nor a python expert. In my application I want to get the value belonging to a certain (user-inputted) DICOM tag. I use ds.iterall() for that and that works well for some sequences but not for others, and I can't wrap my head around to why that is or how to solve this. I looked into the code of iterall() to see if I could fix this, but that looked fine to me (it relies on VR='SQ' to recurse into sequences). I also tried google to find an answer, but found nothing there. Hence I'm reaching out to you.

Here's an example in which Radiopharmaceutical works fine, but not e.g. RadiopharmaceuticalStartTime. Yet I know the latter is there, right next to Radiopharmaceutical, because I see it when I print the sequence value. Looking into the datastructure didn't help me any further (see screenshot at the bottom), I hope you have a suggestion for me on how to proceed?

def printvalue(ds, tagname):
     for elem in ds.iterall():
         if elem.name==tagname:
             print(elem.value)
             return
     print(f'Tagname "{tagname}" not found')

printvalue(ds, 'Radiopharmaceutical')
Fluoro-L-dopa

print(ds[0x0054, 0x0016].value[0]['Radiopharmaceutical'].value)
Fluoro-L-dopa

printvalue(ds, 'RadiopharmaceuticalStartTime')
Tagname "RadiopharmaceuticalStartTime" not found

print(ds[0x0054, 0x0016].value[0]['RadiopharmaceuticalStartTime'].value)
190742.000000

print(ds[0x0054, 0x0016].value)
[(0018, 0031) Radiopharmaceutical                 LO: 'Fluoro-L-dopa'
(0018, 1072) Radiopharmaceutical Start Time      TM: '190742.000000'
(0018, 1074) Radionuclide Total Dose             DS: "194000000.0"
(0018, 1075) Radionuclide Half Life              DS: "6586.2"
(0018, 1076) Radionuclide Positron Fraction      DS: "0.97"
(0018, 1078) Radiopharmaceutical Start DateTime  DT: '20170420190742.000000'
(0054, 0300)  Radionuclide Code Sequence  1 item(s) ---- 
   (0008, 0100) Code Value                          SH: 'C-111A1'
   (0008, 0102) Coding Scheme Designator            SH: 'SRT'
   (0008, 0104) Code Meaning                        LO: '^18^Fluorine'
   (0008, 0105) Mapping Resource                    CS: 'DCMR'
   (0008, 0106) Context Group Version               DT: '20070625000000.000000'
   (0008, 010f) Context Identifier                  CS: '4020'
   ---------
(0054, 0304)  Radiopharmaceutical Code Sequence  1 item(s) ---- 
   (0008, 0100) Code Value                          SH: 'C-B1034'
   (0008, 0102) Coding Scheme Designator            SH: 'SRT'

[darcymason]

Instead of elem.name, it should be:

if elem.keyword == tagname:
   ...

Note keyword rather than name; the latter is the descriptive name including spaces.

Alternatively you could call printvalue with the name including spaces.

It works for 'Radiopharmaceutical' because the name and the keyword happen to be the same.

[marcelzwiers]

Hahaha, wow, so simple, I wasn't looking in that direction at all! But thanks a lot, I have spent (way too) many hours on this :-)
👍

[darcymason]

Happy to help. The keyword/name thing is always a bit of a gotcha; we went with keyword because that is what the DICOM Standard calls it.

As an aside, since you said you are new to this, just want to make sure you know the keywords can be used throughout, e.g. instead of

ds[0x0054, 0x0016].value[0]['Radiopharmaceutical'].value

you can use

ds.RadiopharmaceuticalInformationSequence[0].Radiopharmaceutical

or assign the sequence or sequence items (Dataset) to another variable:

seq = ds.RadiopharmaceuticalInformationSequence
item0 = seq[0]
print(item0.Radiopharmaceutical)
print(item0.RadiopharmaceuticalStartTime)

which I think makes for much more readable code.

[marcelzwiers]

Yes, I saw something like that, I'll keep it in mind, thanks. In this case, that line you mentioned was executed on the command-line, as a test, but it's not in my actual code. As for my code, I now match both on keyword and name (for the convenience of my users), which works very well. I never saw differences between the keyword and the name, except for spaces. I hope won't cause unexpected issues if I do it like this:

def printvalue(ds, tagname):
    for elem in dicomdata.iterall():
        if tagname in (elem.name, elem.keyword):
            print(elem.value)
            break

[darcymason]

I don't see any issues with that - it makes the printvalue function more flexible.

The bigger issue could be speed, if that is important - iterating through the dataset for every item accessed is obviously much slower than going directly to it with lines like in my previous post. With keywords at least, one could simply do if keyword in ds and return it if found, if not ds.iter through it finding sequences and recursively calling the function with each sequence item.

As another point, the function as written will only find the first tag with the name or keyword - if the target name/keyword is in a sequence, then only the value in the first sequence item will be returned.

I'm getting a little off-topic though, so I'll stop there...