Hybrid in-vivo treatment verification for particle therapy: Multivariate modelling of fast neutron and gamma-ray detection

Given its sensitivity to anatomical variations, par-
ticle therapy is expected to benefit strongly from reliable on-
line treatment verification. Making use of a large fraction of
the secondary particles available, hybrid prompt gamma-ray and
fast neutron imaging is a promising approach for this purpose.
We present a simulation study introducing a novel method
for range reconstruction by means of these particles. To this
end, the passage of a proton beam through a patient and the
detection of secondary particles in a dedicated detector array
were simulated by Monte Carlo particle transport calculations.
Anatomical variations were mimicked by adding or removing
tissue to the patient. The secondary particle production vertices
were reconstructed from the events registered by the detector
based on scatter kinematics and maximum likelihood expectation
maximisation. The vertices were projected onto the beam axis
and the most predictive features of the resulting distributions
were identified from a standardised feature set by forward
feature selection and the Least Absolute Shrinkage and Selection
Operator (LASSO). A range-shift sensitivity of 1 mm or less
was reached at intensities of 1.2×10 7 protons per spot with
fast neutrons, and at 2.6×10 7 protons per spot with prompt
gamma-rays. These results demonstrate the potential of particle
treatment verification by fast neutrons and prompt gamma-rays
and strengthen the potential of this hybrid system for clinical
application.