ForwardProjection about HU value

Hello, I have a question about the HU value. I input the data collected by CT, which is the real HU value. Is the pixel value of my reconstructed image also the real HU value? I see that the pixel values in the area outside the FOV are all 0, why is it not similar to the -1000HU value of air?

When I ForwardProjection the reconstructed result, I set the area outside the FOV to 0 and -1000 respectively, and found that the sinogram pixels were normal when 0 was set, and abnormal when -1000 was set (about -100000). Is this because of the RTK implementation?

This question seems to be for RTK. But since you already asked here @simon.rit might answer it.

My thoughts: sinogram might need to be strictly non-negative, while reconstructed CT values range from -1000 to … +inf?

1 Like

CT images are never reconstructed in HU. In most cases, they don’t have a unit. If the projections are calibrated, they are reconstructed in an inverse length unit (e.g. \mathrm{cm}^{-1}). After reconstruction, they are converted to HU by measuring the value of air and water in a phantom and linearly rescaling so that air is -1000 and water 0.
In RTK, the software assumes that the projections are line integrals of attenuation coefficients if they are in float or try to convert them (by taking -log of the values divided by an estimate of the I_0 value). Then they are constructed in the inverse length unit for specifying the spacings and origin of the images (usually mm).

1 Like

Thanks! I think so~

Wooo. Thanks for your professional analysis and explanation. Maybe I need to research.