Hello everyone
I have question, I have several CBCTs for same patient taken in different time points, these CBCTs are in different image dimensions and image spacing, and i want to measure and compare the size of a certain segmented structure among these time points using Segment Statistics module .
Dose the difference in image dimensions and image spacing affect the segmentations statistics between different time points?
will a segmentation with larger voxel spacing be computed larger in terms of volume (number of voxels) compared to one with smaller spacing, even if they delineate the same structure?
Hi, since I haven’t received a response to this concern, I decided to remake all CBCTs with consistent dimensions before proceeding with the segmentations. This way, I can avoid any discrepancies, as redoing the segmentations later would be very time- and effort-intensive if I discovered differences in statistical measures. I haven’t verified whether varying image dimensions can actually affect the statistics, but my recommendation is to standardize the dimensions from the start.
Segment statistics computes volumes in physical dimensions, cubic mm or cubic cm. As long as your voxel sizes are correctly specified for each image, the statistics reported will be physically meaningful and comparable between images. Your segmentations may vary slightly because of the differing resolutions, and raw voxel counts will certainly be different, but all other measurements should be in physical units and comparable across images.
Hi there, silly question, what do you mean by voxel sizes are correctly specified?
Am I right to think that given the volume = voxel size (pixel x slice thickness) x voxel number, we shouldn’t actually need to resample imgaes, right？When different CTs come in different voxel sizes or slice thickness, it’s just different scanners or resolution differences. Voxel counts will be different due to different voxel size (aka resolution) but voxel size and number should be in a linear relationship (e.g. for the same volume, if you have higher resolution which is smaller voxel size, you will have higher voxel number and vice versa) as long as the formula is the same, different images should be comparable.
Correct term is voxel spacing, not thickness. Thickness refers to something else.
But yes, for a cube of 4 voxels in each dimension, and the voxel spacing of 1x1x1mm, it represents the same physical volume of 2 voxels in each dimension with voxel spacing of 2x2x2mm (downsampled by factor of 2), or it will be same volume of 8 voxels in each dimension and voxel spacing of 0.5x0.5x0.5mm.
First volume will contain 64 voxels, second one will have 8 voxels, and the third one will have 512 voxels, but they all represent the same physical volume.
Thank you very much Muratmaga, am I also right to say if I wish to re-sample a DICOM at its z-axis (which represents slice thickness in CT), by making a finer-slice CT to a thicker-slice CT (just so that it reduces the time and effort I need to manually correct the slices - 1mm 170 slices to correct v.s. 3mm 51 slices only to correct), the differences between the volume outcome generated by segment statistic is due to the changes in resolution(reduce in resolution) at the sagittal plane of the CT (corresponding to the Z-axis), so over the edges between two different tissues, higher resolution, finner slices, smaller voxels might be able to pick up the tissue slightly better?
If you are mostly doing slice by slice segmentation and manually drawing contours or painting them, and need the detail, yes, then keeping the inslice resolution high at the expense larger spacing at Z dimension is an feasible option.
In ImageStacks you simply set the skip slice option to 2, which means it will only load the every 3rd slice (and adjust the spacing values appropriately).