MRI-based treatment plan simulation and adaptation for ion radiotherapy using a classification-based approach

Abstract

BackgroundIn order to benefit from the highly conformal irradiation of tumors in ion radiotherapy, sophisticated treatment planning and simulation are required. The purpose of this study was to investigate the potential of MRI for ion radiotherapy treatment plan simulation and adaptation using a classification-based approach.MethodsFirstly, a voxelwise tissue classification was applied to derive pseudo CT numbers from MR images using up to 8 contrasts. Appropriate MR sequences and parameters were evaluated in cross-validation studies of three phantoms. Secondly, ion radiotherapy treatment plans were optimized using both MRI-based pseudo CT and reference CT and recalculated on reference CT. Finally, a target shift was simulated and a treatment plan adapted to the shift was optimized on a pseudo CT and compared to reference CT optimizations without plan adaptation.ResultsThe derivation of pseudo CT values led to mean absolute errors in the range of 81 - 95 HU. Most significant deviations appeared at borders between air and different tissue classes and originated from partial volume effects. Simulations of ion radiotherapy treatment plans using pseudo CT for optimization revealed only small underdosages in distal regions of a target volume with deviations of the mean dose of PTV between 1.4 - 3.1% compared to reference CT optimizations. A plan adapted to the target volume shift and optimized on the pseudo CT exhibited a comparable target dose coverage as a non-adapted plan optimized on a reference CT.ConclusionsWe were able to show that a MRI-based derivation of pseudo CT values using a purely statistical classification approach is feasible although no physical relationship exists. Large errors appeared at compact bone classes and came from an imperfect distinction of bones and other tissue types in MRI. In simulations of treatment plans, it was demonstrated that these deviations are comparable to uncertainties of a target volume shift of 2 mm in two directions indicating that especially applications for adaptive ion radiotherapy are interesting.

Implementation of a Novel Algorithm For Generating Synthetic CT Images From Magnetic Resonance Imaging Data Sets for Prostate Cancer Radiation Therapy

“…Use optimized weights and patient MR images to generate synCT Rigidly register synCT to SIM-CT and import into treatment planning system Compare synCT to SIM-CT by determining MAE of synCT SynCTs were rigidly registered to CT-SIMs, and Hounsfield unit (HU) value differences between CT-SIM and synCT were analyzed via mean absolute error (MAE) (17,18) MAEZ…”

Determination of mean ionization potential using magnetic resonance imaging for the reduction of proton beam range uncertainties: theory and application

“…Proton CT offers the potential to directly determine the proton SPR (Schulte and Penfold, 2012) but currently lacks a clinical implementation as a number of challenges await its development (including numerous small-angle scatterings due to Coulomb interactions and high energy requirements to penetrate the body for imaging) (Li et al, 2006). Recently, magnetic resonance imaging (MRI) images converted to pseudoCT or syntheticCT (sCT) scans have been proposed as a reasonable alternative to SECT (Edmund et al, 2014;Koivula et al, 2016;Rank et al, 2013a;Rank et al, 2013b) for SPR determination. In the case of the prior work of sCT to SPR conversion, the efforts have been to simulate kV SECT images and not necessarily to increase the information content beyond SECT and/or improve the accuracy of SPR determination relative to kV SECT.…”

Patch‐based generation of a pseudo CT from conventional MRI sequences for MRI‐only radiotherapy of the brain

“…This has been attributed to partial volume and susceptibility effects as well as a low signal-to-noise ratio (SNR) of the dUTE scans. 8,16,17 Atlas-based methods, on the other hand, estimate pCTs using conventional (non-dUTE) MRI sequences, and must therefore compensate for the bone/air ambiguity in the MRI images. This is achieved using deformable registration of one or multiple atlases of paired MRI/CT scans to the patient MRI scan and then using the warped atlas CT scan(s) as a pCT estimate.…”