ESTRO 37 Abstract book

S937

ESTRO 37

EP-1747 Characterizing proton stopping power ratio uncertainties in a CT/MRI tissue classification model A. Witztum 1 , T. Solberg 1 , A. Sudhyadhom 1 1 University of California- San Francisco, Department of Radiation Oncology, San Francisco, USA Purpose or Objective To analyze the uncertainty in stopping power ratio (SPR) calculations based on potential sources of error in a four- component tissue classification model using CT and MRI imaging. Material and Methods The greatest source of uncertainty in the Bethe-Bloch equation (used to calculate proton SPR) is the mean ionization potential (I m ). One known method to calculate I m is the Bragg additivity rule (BAR) with a summation over all elemental constituents. We propose a reduction to a four-component classification (4CC) system. Using CT and MRI imaging we propose to classify molecules in the human body as proportions of water, fat, and protein (all by MRI) and hydroxyapatite (HA) by CT. In doing so we can calculate I m by a reduction of the BAR to: 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 type of molecule. For each molecule, the I-value was obtained from ICRU report 44. For fat and protein, a weighted mean I was calculated using prevalence and I-values for the constituent lipids and amino acids respectively. The ‘true’ mass content percentages were obtained for various soft tissues (tissue) and cortical bone (bone) from ICRU report 44 and others available in the literature. Protein content was considered as the remaining mass. Errors in this value were simulated using a normal distribution of errors between 0-2% (mean=1%) for water and HA, and between 0-10% (mean=5%) for fat, and were randomized in the positive and negative directions. For tissue, HA content was assumed to be 0%. I m calculated with the 4CC system was compared to BAR-calculated values. Proton (E=250MeV) SPR was calculated using I m (BAR) as the reference standard and compared to SPR using I m from: the 4CC system without errors I m,base , mean, and max I m from the distribution with errors. Results Table 1 shows use of the 4CC model results in a <1.7% (tissue) and <1.0% (bone) error in I m compared to BAR. For tissue, I m mean, min, and max errors of up to -1.8%, - 3.5%, and between -0.6-1.2% are seen. Table 2 shows that for tissue and bone, use of the 4CC model results in proton SPR calculations within 0.2% of BAR. The mean SPR from the error distribution still gave results accurate to 0.3%. SPR calculations from the max I m values had an error within 0.4%. Adding the errors for SPR in quadrature to include systematic error from the 4CC system compared to BAR (max 0.2%) and the random error (2σ as a percentage of SPR (4CC) value) from content measurement (max 0.3%) yields a total error of 0.4% for the 4CC model.

Conclusion Using the 4CC method is accurate and robust to errors in mean ionization potential for proton SPR calculations. Errors of up to 2% in content quantification of water and HA and up to 10% for fat still result in <0.5% error in SPR. Content quantification by CT/MRI of each component in the 4CC model has been demonstrated to be an accurate method for SPR determination.