Dataset Information

Real-time radial reconstruction with domain transform manifold learning for MRI-guided radiotherapy.

ABSTRACT:

Background

MRI-guidance techniques that dynamically adapt radiation beams to follow tumor motion in real time will lead to more accurate cancer treatments and reduced collateral healthy tissue damage. The gold-standard for reconstruction of undersampled MR data is compressed sensing (CS) which is computationally slow and limits the rate that images can be available for real-time adaptation.

Purpose

Once trained, neural networks can be used to accurately reconstruct raw MRI data with minimal latency. Here, we test the suitability of deep-learning-based image reconstruction for real-time tracking applications on MRI-Linacs.

Methods

We use automated transform by manifold approximation (AUTOMAP), a generalized framework that maps raw MR signal to the target image domain, to rapidly reconstruct images from undersampled radial k-space data. The AUTOMAP neural network was trained to reconstruct images from a golden-angle radial acquisition, a benchmark for motion-sensitive imaging, on lung cancer patient data and generic images from ImageNet. Model training was subsequently augmented with motion-encoded k-space data derived from videos in the YouTube-8M dataset to encourage motion robust reconstruction.

Results

AUTOMAP models fine-tuned on retrospectively acquired lung cancer patient data reconstructed radial k-space with equivalent accuracy to CS but with much shorter processing times. Validation of motion-trained models with a virtual dynamic lung tumor phantom showed that the generalized motion properties learned from YouTube lead to improved target tracking accuracy.

Conclusion

AUTOMAP can achieve real-time, accurate reconstruction of radial data. These findings imply that neural-network-based reconstruction is potentially superior to alternative approaches for real-time image guidance applications.

SUBMITTER: Waddington DEJ

PROVIDER: S-EPMC10809819 | biostudies-literature | 2023 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Real-time radial reconstruction with domain transform manifold learning for MRI-guided radiotherapy.

Waddington David E J DEJ Hindley Nicholas N Koonjoo Neha N Chiu Christopher C Reynolds Tess T Liu Paul Z Y PZY Zhu Bo B Bhutto Danyal D Paganelli Chiara C Keall Paul J PJ Rosen Matthew S MS

Medical physics 20230127 4

<h4>Background</h4>MRI-guidance techniques that dynamically adapt radiation beams to follow tumor motion in real time will lead to more accurate cancer treatments and reduced collateral healthy tissue damage. The gold-standard for reconstruction of undersampled MR data is compressed sensing (CS) which is computationally slow and limits the rate that images can be available for real-time adaptation.<h4>Purpose</h4>Once trained, neural networks can be used to accurately reconstruct raw MRI data wi ...[more]

PMID: 36646444

Similar Datasets

Project description:BackgroundFunctional assessments of the heart by dynamic cardiovascular magnetic resonance (CMR) commonly rely on (i) electrocardiographic (ECG) gating yielding pseudo real-time cine representations, (ii) balanced gradient-echo sequences referred to as steady-state free precession (SSFP), and (iii) breath holding or respiratory gating. Problems may therefore be due to the need for a robust ECG signal, the occurrence of arrhythmia and beat to beat variations, technical instabilities (e.g., SSFP "banding" artefacts), and limited patient compliance and comfort. Here we describe a new approach providing true real-time CMR with image acquisition times as short as 20 to 30 ms or rates of 30 to 50 frames per second.MethodsThe approach relies on a previously developed real-time MR method, which combines a strongly undersampled radial FLASH CMR sequence with image reconstruction by regularized nonlinear inversion. While iterative reconstructions are currently performed offline due to limited computer speed, online monitoring during scanning is accomplished using gridding reconstructions with a sliding window at the same frame rate but with lower image quality.ResultsScans of healthy young subjects were performed at 3 T without ECG gating and during free breathing. The resulting images yield T1 contrast (depending on flip angle) with an opposed-phase or in-phase condition for water and fat signals (depending on echo time). They completely avoid (i) susceptibility-induced artefacts due to the very short echo times, (ii) radiofrequency power limitations due to excitations with flip angles of 10 degrees or less, and (iii) the risk of peripheral nerve stimulation due to the use of normal gradient switching modes. For a section thickness of 8 mm, real-time images offer a spatial resolution and total acquisition time of 1.5 mm at 30 ms and 2.0 mm at 22 ms, respectively.ConclusionsThough awaiting thorough clinical evaluation, this work describes a robust and flexible acquisition and reconstruction technique for real-time CMR at the ultimate limit of this technology.

Dataset Information

Real-time radial reconstruction with domain transform manifold learning for MRI-guided radiotherapy.

Background

Purpose

Methods

Results

Conclusion

Publications

Real-time radial reconstruction with domain transform manifold learning for MRI-guided radiotherapy.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets