How are the spacing value of the z dimension and thickness (0018, 0050) different in dicom series?

Question

I've been studying some dicom series and find that the thickness attribute and the itkimage.GetSpacing()[2] value are not always consistent.

For example the thickness (0018, 0050) value encoded in the dcm file is 1.5 mm but the corresponding spacing indicated simpleITK on z axis is 1.00. Then what value should I use to indicate the physical distance between adjacent voxel centers on the z axis? If they are different things then What do spacings actually mean?

I retrieve thickness and spacing values in python like this:

//thickness using dicom
thickness = dicom.read_file(dcm_file)[0x0018, 0x0050].value

//spacing using simpleITK
reader = sitk.ImageSeriesReader()
dicom_files = reader.GetGDCMSeriesFileNames(dicom_dir)
reader.SetFileNames(dicom_names)
itkImage= rader.Execute

spacing_x, spacing_y, spacing_z = itkImage.GetSpacing()

I've been searching in the docs but haven't found something like an answer yet. Thanks!

====== Update1 ======

I've checked the Image Position(Patient) (0020,0032) value and they are indeed 1mm away per slice. Then What leaves thickness mean?

Answer 1

There is a minimum slice thickness which a scanner can measure. And there is the distance between measurement positions for adjacent slices. And when creating an image, the scanner does not have to measure the next slice exactly one thickness away, it can be less (making some measurement overlap) or more (creating some gaps in measurements). For adjacent voxel centers, you want spacing. Thickness is extra information, usually of little value.

Answer 2

Slices are computed from continuous signals the scanner acquires. You can imagine the computation as kind of an integral or average. In the mathematical model, the slice thickness is infinitely small, but the pixel values are averaged from signals coming from a cuboid around the slice plane.

So slice thickness and distance between adjacent slices are unrelated to each other and may vary independently.

I find this image quite illustrative: https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSWyeGrs-Ojdhg6GCn7obLfQ3j94c_sD0Hy8Oe1A-HdO6OZ-7H- (unfortunately it is small)