Dose calculation fails on volumes with orientation other than LPS

[cpinter]

Using pyRadPlan via 3D Slicer 5.9+ / SlicerRT.

If I have a CT that has an orientation anything other than LPS, in which case the IJK to RAS Direction Matrix is

-1  0  0
 0 -1  0
 0  0  1

dose calculation fails (at least photon does). Sometimes it crashes, sometimes it gives an error

like this

Details

Traceback (most recent call last): File "D:\W\sR\Slicer-build\lib\Slicer-5.9\qt-scripted-modules\DoseEngines\pyRadPlanEngine.py", line 87, in calculateDoseUsingEngine dij = calc_dose_influence(ct, cst, stf, pln) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\pyRadPlan\dose\_calc_dose.py", line 45, in calc_dose_influence dij = engine.calc_dose_influence(ct, cst, stf) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\pyRadPlan\dose\engines\_base.py", line 274, in calc_dose_influence dij = self._calc_dose(ct, cst, stf) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\pyRadPlan\dose\engines\_base_pencilbeam.py", line 173, in _calc_dose curr_beam = self._init_beam(dij, ct, cst, scen_stf, i) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\pyRadPlan\dose\engines\_svdpb.py", line 140, in _init_beam beam_info = super()._init_beam(beam_info, ct, cst, stf, i) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\pyRadPlan\dose\engines\_base_pencilbeam.py", line 434, in _init_beam self._compute_ssd(beam_info, ct, density_threshold=self.ssd_density_threshold) File "D:\W\sR\python-install\Lib\site-packages\pyRadPlan\dose\engines\_base_pencilbeam.py", line 524, in _compute_ssd ix_ssd[~ix_nan] = np.argmax(rho[0][~ix_nan] > density_threshold, axis=1) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\numpy\_core\fromnumeric.py", line 1298, in argmax return _wrapfunc(a, 'argmax', axis=axis, out=out, **kwds) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\W\sR\python-install\Lib\site-packages\numpy\_core\fromnumeric.py", line 57, in _wrapfunc return bound(*args, **kwds) ^^^^^^^^^^^^^^^^^^^^ ValueError: attempt to get argmax of an empty sequence

Happy to provide more information if someone starts working on this but needs data, script etc.

[wahln]

This sounds reasonable to me given we mainly worked with LPS data for initial development.

If you need this for the pyRadPlan SlicerRT interface, it might first make sense to do a conversion to LPS in pre-processing @LiBu981 as a workaround.
Apart from this, if you could setup a short script to reproduce this error, potentially with some generated phantom data, we will allocate some resources to support this in our pyRadPlan core as well.