ENH: Add option to compute the Cramer-Rao Lower Bound. #620
Conversation
| if (m_IsSpectralCT) | ||
| { | ||
| rtk::ProjectionsDecompositionNegativeLogLikelihood::MeasuredDataType photon_counts; | ||
| photon_counts.SetSize(forward.size()); | ||
| photon_counts.SetData(forward.data_block()); | ||
| cost->SetMeasuredData(photon_counts); | ||
| cost->ComputeFischerMatrix(in); | ||
| output1It.Set(cost->GetCramerRaoLowerBound()); | ||
| } | ||
| else | ||
| { | ||
| output1It.Set( | ||
| itk::VariableLengthVector<double>(cost->GetVariances(in).data_block(), this->m_NumberOfSpectralBins)); | ||
| } |
There was a problem hiding this comment.
I'm not sure I understand this, it seems to disable the previous mode which was already for m_IsSpectralCT true. Was the behavior different before?
There was a problem hiding this comment.
The function GetVariances was not implemented for spectral CT. It was only returning meaningless results (see here). Furthermore, it was no possible to set the boolean m_ComputeVariances to true with the rtkspectralforwardmodel application.
There was a problem hiding this comment.
Ok... Do you understand what it does for dual energy?
https://github.com/RTKConsortium/RTK/blob/master/include/rtkDualEnergyNegativeLogLikelihood.h#L126
I have a hard time understanding this code and the commit message does not help
There was a problem hiding this comment.
If I understood well, it is mentioned here that the variance of the integrated energy is assumed to be equal to the mean. Then, the function GetVariances is computing this:
which correspond to the mean signal measured by an energy integrating detector.
(This is my understanding but I'm not sure I'm right.)
Add option to obtain the Cramer-Rao Lower Bound (CRLB) when computing the photon counts using rtkSpectralForwardModelImageFilter. The CRLB corresponds to the lower bound of the variance of any unbiased estimator. Therefore, it could be used to estimate the variance that would be obtained after material decomposition.
arobert01
force-pushed
the
cramer-rao
branch
from
fc2ca4b
to
6f0bfc2
Compare
The application was wrongfully removed by this commit (RTKConsortium@b1f0449#diff-22b5c767b274cb860bc3604b376958ddc1bc588a298153c14dcaf6d95db676b1L90)
arobert01
force-pushed
the
cramer-rao
branch
from
6f0bfc2
to
45b0fa2
Compare
|
The last force-pushed improve the way the computation of the Cramer-Rao lower bound was implemented and add the computation of the variance also for spectral CT. |
Add option to obtain the Cramer-Rao Lower Bound (CRLB) when computing the photon counts using rtkSpectralForwardModelImageFilter. The CRLB corresponds to the lower bound of the variance of any unbiased estimator. Therefore, it could be used to estimate the variance that would be obtained after material decomposition.