Applying Rescale slope and Rescale intercept and Modality LUT Sequence

[vignesh]

Let's say I want to read my DICOM File and extract pixel data to display it on a viewer.

I have gone through the following links,

https://dicom.nema.org/medical/dicom/cu ... .3.10.html.

I have a few Questions regarding applying rescale slope and intercept over an image,

[QUESTION: 1] Do I have to scale every pixel in my image using rescale slope and intercept before displaying it in my dicom viewer? In my understanding rescale slope and intercept is used to convert pixel values to output units which may be HU in the case of CT. If my pixel values are converted to HU values, my image won't be in proper form because images are made of pixels, not HU values. Is this right?

[QUESTION: 2] Let's say, I have a DICOM image whose VOI LUT Function tag has the value LINEAR, Rescale slope is 1 and Rescale Intercept is 0. It also has a Modality LUT Sequence tag. Do I need to apply Modality LUT Sequence over my image before applying LINEAR window leveling? I have also read DCMTK C++ library functions which explicitly mention skipping Modality LUT sequence for following SOP Class UIDs,

1.2.840.10008.​5.​1.​4.​1.​1.​12.​1 [ X​Ray​Angiographic​Image​Storage ]
1.2.840.10008.​5.​1.​4.​1.​1.​12.​2 [ X​Ray​Radiofluoroscopic​Image​Storage ]
1.2.840.10008.5.1.4.1.1.12.3 [ XRay​Angiographic​Bi​Plane​Image​Storage ]

It would be helpful If someone can share sample DICOM file for Modality LUT sequence for any other SOP class UIDs except those 3 mentioned above.

Thankyou!

[Marco Eichelberg]

• Question 1: Yes, you have to apply Rescale Slope and Intercept to every pixel. However, this transformation, called the "Modality LUT Transformation" in DICOM (even if there is not real look-up table), is not intended to convert pixels to output values. See the DICOM pixel transformation pipeline explained in DICOM Part 4.
• Question 2: If you have a Modality LUT Sequence, then you have to apply the look-up table before applying window leveling. It is correct that the Modality LUT should be ignored in the case of the three SOP classes you mention.

That all said, you can find classes in DCMTK that already implement all of this in a quite efficient manner. Check class DicomImage in module dcmimgle. This should provide everything you need (and much more).