Session Item

The composition of metal implants (MI) is often unknown, leading to wrongful material assignments in radiotherapy (RT) dose planning. CT numbers of MI can be ambiguous due to the combined effects of photoelectric and Compton interactions at low scanning energies (keV) and the high charge (Z) of metals. Further, CT numbers can be incorrect due to beam hardening and photon starvation modelling effects. To address these concerns, we investigate whether 2D dual energy radiographs, simulating 2D CT topogram projections, can provide a bulk effective charge (Z_eff) classification of MI.

We included bone and metals inserts (Cirs and Gammex Inc.) with known charges for modelling covering Z/Z_eff=10 (Inner bone) to 29 (Copper). Z_eff=√(∑_iw_iZ_i^2.98)^1/2.98 where w_i is the fraction of total electrons of material i. 4 unknown materials were evaluated; the head of a hip implant (HipHead) and proximal femoral rod (rod) from Corail Hip® systems, teeth with dental implants (teeth) and a stainless steel alloy (SS_unknw) insert (Gammex Inc.).

Radiographs with (I) and without (I₀) materials were obtained using the on-board imaging system on a Varian TrueBeam v2.7 (Varian Medical Systems) at energies (E)= 70, 80, 100, 120 and 140kV and constant mAs=0.5. Radiographs E>80kV were also acquired with a Ti filter (E_Ti). Projections p(E)=ln[I₀(E)/I(E)] were generated and ratios pR=p(E_High)/p(E_Low) were created where E_High are all E>E_Low. Average pR values were extracted for each material.

A mono exponential empirical model Z_eff=exp[a₀+a₁·pR] was fitted for each dual energy pair to search for an optimal combination. Further, materials with a larger Z and a quadratic pR term were investigated but did not improve the model predictions. Thus a simpler model in a more relevant Z interval was chosen. p(E) and a₀ values < 0 were disregarded as too noisy (I>I₀) or non-physical (Z<1). Only E combinations proving a reasonable fit (R²>0.8) were used for Z_eff prediction.

pR with combinations E_Low-(E_High) =70-(100/120/120_Ti/140_Ti), 80-(100/120/120_Ti/140/140_Ti), 100-(140/140_Ti), 100_Ti-(120/140/140_Ti), 120_Ti-(140/140_Ti) kV all had similar fits (see figure).

The corresponding estimated Z_eff intervals were 22-24 for rod (Ti+SS alloys, Z:20-30), 21-27 for teeth (bone+Au+Sn+Cu, Z:14-45), 28-36 for SS_unknw (Cr+F+Ni+Cu+Mo, Z:24-42) and 30-41 for HipHead (Ti-Al+Co/Cr alloys, Z<27) ± 15-48% range uncertainty. Known model materials could be predicted with 4-21% uncertainty (see table). All estimated Z_eff values of the unknown materials were within a theoretical Z interval except for HipHead which was estimated higher but within the uncertainty range.

This method provides a first step towards allocating MI to material specific dose kernel by estimating their Z_eff from dual energy topograms. Increased mAs impact on imaging plate gain factors for uncertainty reduction and adding MI in anthropomorphic phantoms should be further investigated.