Click here to


Are you sure ?

Yes, do it No, cancel

Voxel-Specific Proton Stopping Power Ratios and Uncertainties in a CT/MRI Tissue Classification Model

A Witztum*, T Solberg , A Sudhyadhom , University of California San Francisco, San Francisco, CA


(Sunday, 7/29/2018) 1:00 PM - 1:55 PM

Room: Karl Dean Ballroom A2

Purpose: To create voxel-specific proton stopping power ratios (SPRs) and associated uncertainties based on a four-component tissue classification model using CT and MRI imaging.

Methods: The greatest source of uncertainty in the Bethe-Bloch equation (for SPR calculation) is the mean ionization potential (Im). We propose a four-component classification (4CC) system whereby we classify molecules (or voxels) as proportions of water, fat, and protein (by MRI), and hydroxyapatite (HA) by CT. This model is written: ln(I(m)) = (w(water)ln(I(water))) + (w(fat)ln(I(fat))) + (w(protein)ln(I(protein))) + (w(HA)ln(I(HA))) where w is the mass content percentage and I is the mean ionization potential for each molecule. We have shown this model to calculate SPR within 0.2% of a summation over all elemental constituents (Bragg additivity rule).The utility of this model is demonstrated by classifying every voxel on a pelvis CT slice as adipose, prostate, muscle, skin, stool, seminal vesicles, urine, cartilage, cortical bone, and trabecular bone. The ‘true’ 4CC composition (ICRP 23) of each class was used to create an SPR map. Errors were simulated with a normal distribution between 0-2% (water, HA) and 0-10% (fat) and randomized in +/- directions to mimic the errors in MRI/CT based classification. Protein was quantified as the remainder. Voxel SPR uncertainty (2 x standard deviation as percentage of SPR) and values for known CT imaging and range degradation errors are added in quadrature.

Results: A proton (E=250 MeV) SPR map was made along with an uncertainty map. The range of this uncertainty was 0.11-0.25% (quantification errors) and 0.63-0.67% (total).

Conclusion: Using the 4CC method is accurate and gives access to tissue composition specific SPR uncertainty. Once CT/MRI based content quantification techniques are introduced in radiotherapy, individual voxel SPRs and uncertainties should be considered in treatment planning systems. This is an area currently being investigated.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email