DICOM: treat ultrasound ECG data as timelapse

[melissalinkert]

This is the very beginning of basic timelapse support, and does not extend to other modalities.

I don't have data that can be shared at the moment, but the use case that prompts this change is ultrasound data containing an ECG. Without this change, showinf that all planes are Z sections; with this change, all planes are timepoints. I am opening as a draft for feedback, but expect this will be substantially changed (if not closed).

Implementation details aside, this raises a few questions:

• is checking for the presence of specific modality/data type metadata the right way to determine whether a set of planes are a Z stack or a timelapse?
• if checking the modality/data type is the way to determine Z and T dimensions, which specific modalities should be treated as timelapse?

I don't recall if this has come up in discussion with @dclunie / @fedorov in the past, but any thoughts would be welcome.

[fedorov]

@melissalinkert I always thought of bioformats as the library for microscopy imaging. Are you suggesting you are expanding the scope to cover radiology modalities?

I do not have experience with US ECG imaging, so I am mostly responding based on my experience in other radiology modalities.

is checking for the presence of specific modality/data type metadata the right way to determine whether a set of planes are a Z stack or a timelapse?

At least not always. In MR/CT/PET you would typically use ImagePositionPatient to stack the slices into a volume.

I don't have experience working with ultrasound images, but from what I know, they are notorious for the difficulties of figuring out the geometry and reconstructing volume. This page might be helpful to get a perspective: https://github.com/SlicerHeart/SlicerHeart/blob/master/Docs/ImageImportExport.md.

if checking the modality/data type is the way to determine Z and T dimensions, which specific modalities should be treated as timelapse?

I don't think you can make a simple rule based on modality.

[melissalinkert]

Thanks for the feedback, @fedorov. Given that this a more complex issue than it first appears, I'll close this pull request and open something new at a later date.

Bio-Formats has had basic support for other modalities for a long time, in that you should be able to at least read images. Most metadata for radiology modalities does not have a place in the OME schema, though, so is not as well-supported as WSI data. https://downloads.openmicroscopy.org/images/DICOM/samples/ has public examples of files in other modalities that are included in our nightly tests; those have been readable since the early days of Bio-Formats (~2007-2008).

[dclunie]

FYI, the short answer for traditional US and similar XA (non-enhanced-family) multi-frame (NumberOfFrames present and > 1, Multi-frame Module) is FrameTime or FrameTimeVector (Cine Module) that are the target of FrameIncrementPointer.

You can probably find some relevant samples in the images at ftp://medical.nema.org/MEDICAL/Dicom/DataSets/WG12/ and also ftp://dicom.offis.de/pub/dicom/images/