Accurate resizing of ultrasounds using DICOM data

[pCrowley]

Hello and good day! Our group is attempting to standardize a large set of ultrasound images. Our goal is to generate a set of final images so that sizes of structures within the ultrasounds can be directly compared to one another.

The images come from multiple machines, using multiple softwares. To resize, we have multiplied Rows * PhysicalDeltaY and Columns*PhysicalDeltaX, which we thought would lead to a standardized pixel:mm ratio. However, the end result from this effort are images that are disproportionately stretched vertically (or, conversely, not stretched enough horizontally). Round objects become elliptical as a result.

We are hoping to determine a way to accurately resize these images, but have tried many other iterations/variables, all with abnormally skewed images. However, DICOM software like RadiAnt seems to be able to measure accurately in mm images, regardless of software, so we must be missing something critical.

Does anyone have any suggestions on how to standardize this? And how does DICOM reader software determine millimeter distance in an ultrasound? If we can work back from that, identify the DICOM header information involved in that calculation, then I think this could be one step closer to a solution. Thank you for your help!

[Michael Onken]

Hi,

Not really a DCMTK question, but....

Ultrasound pixels must not be quadratic but can be rectangular. I guess you already know that after looking at your images

This information is stored in DICOM attributes, too. Specifically look for Pixel Aspect Ratio and Pixel Spacing. They are part of the Image Pixel Module, see the online version. Please read also the attribute descriptions to understand their relationship to each other and other related attributes.

Best regards,
Michael