AAPM Task Group 264: The safe clinical implementation of MLC tracking in radiotherapy

Keall, P.; Sawant, Amit; Berbeco, R.; Booth, Jeremy; Cho, Byungchul; Cerviño, Laura; Cirino, Eileen T.; Dieterich, Sonja; Fast, Martin F.; Greer, Peter; Rosenschöld, Per Munck af; Parikh, Parag J.; Poulsen, P.R.; Santanam, Lakshmi; Sherouse, George W.; Shi, Jinjing; Stathakis, Sotirios

doi:10.1002/mp.14625

Cited by 54 publications

(74 citation statements)

References 123 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…18 Recent works proposed using low-rank models to reconstruct highly undersampled MRI with subsecond temporal resolution, 19 but these methods currently have long reconstruction times. Currently, none of these methods can achieve the required acceleration factor combined with low-latency reconstruction to estimate motion within 500 ms. 10 Recently, deep learning (DL) has been proposed to speed up MRI reconstruction and motion estimation, achieving performances on par, if not higher, than its non-DL counterparts. [20][21][22][23][24][25] Specifically, DL models allow for fast inference, leaving the time-consuming step to the training phase, which can take hours or days.…”

Section: Introductionmentioning

confidence: 99%

“…For real‐time treatment adaptation, image acquisition and motion estimation must occur with low latency and a high spatiotemporal resolution, 9 that is, the maximum time between a (respiratory) motion event and dose delivery should be

\leq 500

ms 10 . However, real‐time acquisition of three‐dimensional (3D) MRI and computation of a nonrigid deformation vector field (DVF) is challenging due to the long acquisition times of fully sampled MRI (seconds to minutes) and the ill‐posed and underdetermined nature of motion estimation, hindering real‐time motion estimation 11,12 …”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Real‐time 3D motion estimation from undersampled MRI using multi‐resolution neural networks

et al. 2021

View full text Add to dashboard Cite

To enable real-time adaptive magnetic resonance imaging-guided radiotherapy (MRIgRT) by obtaining time-resolved three-dimensional (3D) deformation vector fields (DVFs) with high spatiotemporal resolution and low latency (< 500 ms). Theory and Methods:Respiratory-resolved T 1 -weighted 4D-MRI of 27 patients with lung cancer were acquired using a golden-angle radial stack-of -stars readout. A multiresolution convolutional neural network (CNN) called TEMPEST was trained on up to 32× retrospectively undersampled MRI of 17 patients, reconstructed with a nonuniform fast Fourier transform, to learn optical flow DVFs. TEMPEST was validated using 4D respiratory-resolved MRI, a digital phantom, and a physical motion phantom. The time-resolved motion estimation was evaluated in-vivo using two volunteer scans, acquired on a hybrid MR-scanner with integrated linear accelerator. Finally, we evaluated the model robustness on a publicly-available four-dimensional computed tomography (4D-CT) dataset. Results: TEMPEST produced accurate DVFs on respiratory-resolved MRI at 20-fold acceleration, with the average end-point-error < 2 mm, both on respiratory-sorted MRI and on a digital phantom. TEMPEST estimated accurate time-resolved DVFs on MRI of a motion phantom, with an error < 2 mm at 28× undersampling. On two volunteer scans, TEMPEST accurately estimated motion compared to the self -navigation signal using 50 spokes per dynamic (366× undersampling).At this undersampling factor,DVFs were estimated within 200 ms, including MRI acquisition. On fully sampled CT data, we achieved a target registration error of 1.87 ± 1.65 mm without retraining the model. Conclusion:A CNN trained on undersampled MRI produced accurate 3D DVFs with high spatiotemporal resolution for MRIgRT.

show abstract

Section: Introductionmentioning

confidence: 99%

\leq 500

Section: Introductionmentioning

confidence: 99%

Real‐time 3D motion estimation from undersampled MRI using multi‐resolution neural networks

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Therefore, it is relatively easy to choose and implement an already established algorithm depending on the specific application. Another advantage is that many of the discussed AI approaches have low latency during inference which meet requirements outlined by the AAPM Task Group 264 115 . As discussed above, the crucial next steps to translate AI solutions for intrafraction motion monitoring are (i) the development of large annotated database and (ii) standardisation of the reporting metrics.…”

Section: Discussionmentioning

confidence: 99%

“…The latency of the motion management system needs to be minimised to enable real-time adaption. Real-time is defined by the AAPM Task Group 264 as a system latency below 500 ms. 115 To achieve real-time motion adaption, there is a prolific body of work that looks at motion prediction using ML and DL methods. AI-based predictors have demonstrated high accuracy in predicting the motion of tumours in the lung and abdomen.…”

Section: Discussionmentioning

confidence: 99%

“…Another advantage is that many of the discussed AI approaches have low latency during inference which meet requirements outlined by the AAPM Task Group 264. 115 As discussed above, the crucial next steps to translate AI solutions for intrafraction motion monitoring are (i) the development of large annotated database and (ii) standardisation of the reporting metrics. This will help to ensure that the AI algorithms are robust to different systems and populations, achieving similar impact of the ImageNet data set on the development of deep learning.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A review of artificial intelligence applications for motion tracking in radiotherapy

2021

View full text Add to dashboard Cite

During radiotherapy, the organs and tumour move as a result of the dynamic nature of the body; this is known as intrafraction motion. Intrafraction motion can result in tumour underdose and healthy tissue overdose, thereby reducing the effectiveness of the treatment while increasing toxicity to the patients. There is a growing appreciation of intrafraction target motion management by the radiation oncology community. Real‐time image‐guided radiation therapy (IGRT) can track the target and account for the motion, improving the radiation dose to the tumour and reducing the dose to healthy tissue. Recently, artificial intelligence (AI)‐based approaches have been applied to motion management and have shown great potential. In this review, four main categories of motion management using AI are summarised: marker‐based tracking, markerless tracking, full anatomy monitoring and motion prediction. Marker‐based and markerless tracking approaches focus on tracking the individual target throughout the treatment. Full anatomy algorithms monitor for intrafraction changes in the full anatomy within the field of view. Motion prediction algorithms can be used to account for the latencies due to the time for the system to localise, process and act.

show abstract

Cardiac motion and its dosimetric impact during radioablation for refractory ventricular tachycardia

Harms

Schreibmann

McCall

et al. 2023

J Applied Clin Med Phys

View full text Add to dashboard Cite

Introduction: Cardiac radioablation (CR) is a noninvasive treatment option for patients with refractory ventricular tachycardia (VT) during which high doses of radiation, typically 25 Gy, are delivered to myocardial scar. In this study, we investigate motion from cardiac cycle and evaluate the dosimetric impact in a cohort of patients treated with CR. Methods: This retrospective study included eight patients treated at our institution who had respiratory-correlated and ECG-gated 4DCT scans acquired within 2 weeks of CR. Deformable image registration was applied between maximum systole (SYS) and diastole (DIAS) CTs to assess cardiac motion. The average respiratory-correlated CT (AVG resp ) was deformably registered to the average cardiac (AVG cardiac ), SYS, and DIAS CTs, and contours were propagated using the deformation vector fields (DVFs). Finally, the original treatment plan was recalculated on the deformed AVG resp CT for dosimetric assessment. Results: Motion magnitudes were measured as the mean (SD) value over the DVFs within each structure. Displacement during the cardiac cycle for all chambers was 1.4 (0.9) mm medially/laterally (ML),1.6 (1.0) mm anteriorly/posteriorly (AP), and 3.0 (2.8) mm superiorly/inferiorly (SI). Displacement for the 12 distinct clinical target volumes (CTVs) was 1.7 (1.5) mm ML, 2.4 (1.1) mm AP, and 2.1 (1.5) SI. Displacements between the AVG resp and AVG cardiac scans were 4.2 (2.0) mm SI and 5.8 (1.4) mm total. Dose recalculations showed that cardiac motion may impact dosimetry, with dose to 95% of the CTV dropping from 27.0 (1.3) Gy on the AVG resp to 20.5 (7.1) Gy as estimated on the AVG cardiac . Conclusions: Cardiac CTV motion in this patient cohort is on average below 3 mm, location-dependent, and when not accounted for in treatment planning may impact target coverage. Further study is needed to assess the impact of cardiac motion on clinical outcomes.

show abstract

AAPM Task Group 264: The safe clinical implementation of MLC tracking in radiotherapy

Cited by 54 publications

References 123 publications

Real‐time 3D motion estimation from undersampled MRI using multi‐resolution neural networks

Real‐time 3D motion estimation from undersampled MRI using multi‐resolution neural networks

A review of artificial intelligence applications for motion tracking in radiotherapy

Cardiac motion and its dosimetric impact during radioablation for refractory ventricular tachycardia

Contact Info

Product

Resources

About