Model-based machine learning for the recovery of lateral dose profiles of small photon fields in magnetic field

Abstract

Objective:To investigate the feasibility to train artificial neural networks (NN) to recover lateral dose profiles from detector measurements in a magnetic field. Approach:A novel framework based on a mathematical convolution model has been proposed to generate measurement-less training dataset. 2D dose deposition kernels and detector lateral fluence response functions of two air-filled ionization chambers and two diode-type detectors have been simulated without magnetic field and for magnetic field B = 0.35 and 1.5 T. Using these convolution kernels, training dataset consisting of pairs of dose profiles D(x,y) and signal profiles M(x,y) were computed for a total of 108 2D photon fluence profiles ψ(x,y) (80% training / 20% validation). The NN were tested using three independent datasets, where the second test dataset has been obtained from simulations using realistic phase space files of clinical linear accelerator and the third test dataset was measured at a conventional linac equipped with electromagnets. Main results:The convolution kernels show magnetic field dependence due to the influence of the Lorentz force on the electron transport in the water phantom and detectors. The NN show good performance during training and validation with mean square error (MSE) reaching a value of 1e-6 or better. The corresponding correlation coefficients R reached the value of 1 for all models indicating an excellent agreement between expected D(x,y) and predicted D_pred (x,y). The comparisons between D(x,y) and D_pred (x,y) using the three test datasets resulted in gamma indices (1mm/1% global) < 1 for all evaluated data points. Significance:Two verification approaches have been proposed to warrant the mathematical consistencies of the NN outputs. Besides offering a correction strategy not existed so far for relative dosimetry in a magnetic field, this work could help to raise awareness and to improve understanding on the distortion of detector’s signal profiles by a magnetic field.

Magnetic field influence on the light yield from fiber-coupled BCF-60 plastic scintillators of relevance for output factor dosimetry in MR-linacs

“…Figure 8 demonstrated how the maximum dose point for a small field can change position with the B-field (cf [23]). The signal from an ideal scintillator detector without sensitivity to the magnetic field positioned at the maximum dose point for B = 0 T (and remaining there) should therefore decrease if the B-field is changed to, say, ±0.7 T. We, however, consistently saw the scintillator signal go up for stronger B-fields relative to 0 T (e.g.…”

“…Studies aimed for high-energy particle physics have demonstrated that the scintillator light yield tends to increase with magnetic field strength [18,19]. A typical value seems to be 5−10% increase when changing the magnetic flux density (called B-field in most of the following) from 0 T to 2 T. However, little discussion has been devoted to the question of how the scintillator light yield is affected by the magnetic field during fiber-coupled dosimetry related to radiotherapy, and previous studies of this question [20][21][22] have not lead to conclusive results as recently pointed out by Looe et al [23].…”