Principal component reconstruction (PCR) for cine CBCT with motion learning from 2D fluoroscopy

A review on 4D cone-beam CT (4D-CBCT) in radiation therapy: Technical advances and clinical applications

Cone-beam CT (CBCT) is the most commonly used onboard imaging technique for target localization in radiation therapy. Conventional 3D CBCT acquires x-ray cone-beam projections at multiple angles around the patient to reconstruct 3D images of the patient in the treatment room. However, despite its wide usage, 3D CBCT is limited in imaging disease sites affected by respiratory motions or other dynamic changes within the body, as it lacks time-resolved information. To overcome this limitation, 4D-CBCT was developed to incorporate a time dimension in the imaging to account for the patient's motion during the acquisitions. For example, respiration-correlated 4D-CBCT divides the breathing cycles into different phase bins and reconstructs 3D images for each phase bin, ultimately generating a complete set of 4D images. 4D-CBCT is valuable for localizing tumors in the thoracic and abdominal regions where the localization accuracy is affected by respiratory motions. This is especially important for hypofractionated stereotactic body radiation therapy (SBRT), which delivers much higher fractional doses in fewer fractions than conventional fractionated treatments. Nonetheless, 4D-CBCT does face certain limitations, including long scanning times, high imaging doses, and compromised image quality due to the necessity of acquiring sufficient x-ray projections for each respiratory phase. In order to address these challenges, numerous methods have been developed to achieve fast, low-dose, and high-quality 4D-CBCT. This paper aims to review the technical developments surrounding 4D-CBCT comprehensively. It will explore conventional algorithms and recent deep learning-based approaches, delving into their capabilities and limitations. Additionally, the paper will discuss the potential clinical applications of 4D-CBCT and outline a future roadmap, highlighting areas for further research and development. Through this exploration, the readers will better understand 4D-CBCT's capabilities and potential to enhance radiation therapy.

Dynamic CBCT imaging using prior model-free spatiotemporal implicit neural representation (PMF-STINR)

Objective. Dynamic cone-beam computed tomography (CBCT) can capture high-spatial-resolution, time-varying images for motion monitoring, patient setup, and adaptive planning of radiotherapy. However, dynamic CBCT reconstruction is an extremely ill-posed spatiotemporal inverse problem, as each CBCT volume in the dynamic sequence is only captured by one or a few x-ray projections, due to the slow gantry rotation speed and the fast anatomical motion (e.g. breathing).Approach. We developed a machine learning-based technique, prior-model-free spatiotemporal implicit neural representation (PMF-STINR), to reconstruct dynamic CBCTs from sequentially acquired x-ray projections. PMF-STINR employs a joint image reconstruction and registration approach to address the under-sampling challenge, enabling dynamic CBCT reconstruction from singular x-ray projections. Specifically, PMF-STINR uses spatial implicit neural representations to reconstruct a reference CBCT volume, and it applies temporal INR to represent the intra-scan dynamic motion of the reference CBCT to yield dynamic CBCTs. PMF-STINR couples the temporal INR with a learning-based B-spline motion model to capture time-varying deformable motion during the reconstruction. Compared with the previous methods, the spatial INR, the temporal INR, and the B-spline model of PMF-STINR are all learned on the fly during reconstruction in a one-shot fashion, without using any patient-specific prior knowledge or motion sorting/binning.Main results. PMF-STINR was evaluated via digital phantom simulations, physical phantom measurements, and a multi-institutional patient dataset featuring various imaging protocols (half-fan/full-fan, full sampling/sparse sampling, different energy and mAs settings, etc). The results showed that the one-shot learning-based PMF-STINR can accurately and robustly reconstruct dynamic CBCTs and capture highly irregular motion with high temporal (∼ 0.1 s) resolution and sub-millimeter accuracy.Significance. PMF-STINR can reconstruct dynamic CBCTs and solve the intra-scan motion from conventional 3D CBCT scans without using any prior anatomical/motion model or motion sorting/binning. It can be a promising tool for motion management by offering richer motion information than traditional 4D-CBCTs.

Dynamic CBCT Imaging using Prior Model-Free Spatiotemporal Implicit Neural Representation (PMF-STINR)

Objective

Dynamic cone-beam computed tomography (CBCT) can capture high-spatial-resolution, time-varying images for motion monitoring, patient setup, and adaptive planning of radiotherapy. However, dynamic CBCT reconstruction is an extremely ill-posed spatiotemporal inverse problem, as each CBCT volume in the dynamic sequence is only captured by one or a few X-ray projections, due to the slow gantry rotation speed and the fast anatomical motion (e.g., breathing).

Approach

We developed a machine learning-based technique, prior-model-free spatiotemporal implicit neural representation (PMF-STINR), to reconstruct dynamic CBCTs from sequentially acquired X-ray projections. PMF-STINR employs a joint image reconstruction and registration approach to address the under-sampling challenge, enabling dynamic CBCT reconstruction from singular X-ray projections. Specifically, PMF-STINR uses spatial implicit neural representation to reconstruct a reference CBCT volume, and it applies temporal INR to represent the intra-scan dynamic motion with respect to the reference CBCT to yield dynamic CBCTs. PMF-STINR couples the temporal INR with a learning-based B-spline motion model to capture time-varying deformable motion during the reconstruction. Compared with the previous methods, the spatial INR, the temporal INR, and the B-spline model of PMF-STINR are all learned on the fly during reconstruction in a one-shot fashion, without using any patient-specific prior knowledge or motion sorting/binning.

Main results

PMF-STINR was evaluated via digital phantom simulations, physical phantom measurements, and a multi-institutional patient dataset featuring various imaging protocols (half-fan/full-fan, full sampling/sparse sampling, different energy and mAs settings, etc.). The results showed that the one-shot learning-based PMF-STINR can accurately and robustly reconstruct dynamic CBCTs and capture highly irregular motion with high temporal (~0.1s) resolution and sub-millimeter accuracy.

Significance

PMF-STINR can reconstruct dynamic CBCTs and solve the intra-scan motion from conventional 3D CBCT scans without using any prior anatomical/motion model or motion sorting/binning. It can be a promising tool for motion management by offering richer motion information than traditional 4D-CBCTs.

Dynamic cone-beam CT reconstruction using spatial and temporal implicit neural representation learning (STINR)

Objective. Dynamic cone-beam CT (CBCT) imaging is highly desired in image-guided radiation therapy to provide volumetric images with high spatial and temporal resolutions to enable applications including tumor motion tracking/prediction and intra-delivery dose calculation/accumulation. However, dynamic CBCT reconstruction is a substantially challenging spatiotemporal inverse problem, due to the extremely limited projection sample available for each CBCT reconstruction (one projection for one CBCT volume).Approach. We developed a simultaneous spatial and temporal implicit neural representation (STINR) method for dynamic CBCT reconstruction. STINR mapped the unknown image and the evolution of its motion into spatial and temporal multi-layer perceptrons (MLPs), and iteratively optimized the neuron weightings of the MLPs via acquired projections to represent the dynamic CBCT series. In addition to the MLPs, we also introduced prior knowledge, in the form of principal component analysis (PCA)-based patient-specific motion models, to reduce the complexity of the temporal mapping to address the ill-conditioned dynamic CBCT reconstruction problem. We used the extended-cardiac-torso (XCAT) phantom and a patient 4D-CBCT dataset to simulate different lung motion scenarios to evaluate STINR. The scenarios contain motion variations including motion baseline shifts, motion amplitude/frequency variations, and motion non-periodicity. The XCAT scenarios also contain inter-scan anatomical variations including tumor shrinkage and tumor position change.Main results. STINR shows consistently higher image reconstruction and motion tracking accuracy than a traditional PCA-based method and a polynomial-fitting-based neural representation method. STINR tracks the lung target to an average center-of-mass error of 1-2 mm, with corresponding relative errors of reconstructed dynamic CBCTs around 10%.Significance. STINR offers a general framework allowing accurate dynamic CBCT reconstruction for image-guided radiotherapy. It is a one-shot learning method that does not rely on pre-training and is not susceptible to generalizability issues. It also allows natural super-resolution. It can be readily applied to other imaging modalities as well.

A modern review of the uncertainties in volumetric imaging of respiratory-induced target motion in lung radiotherapy

Radiotherapy has become a critical component for the treatment of all stages and types of lung cancer, often times being the primary gateway to a cure. However, given that radiation can cause harmful side effects depending on how much surrounding healthy tissue is exposed, treatment of the lung can be particularly challenging due to the presence of moving targets. Careful implementation of every step in the radiotherapy process is absolutely integral for attaining optimal clinical outcomes. With the advent and now widespread use of stereotactic body radiation therapy (SBRT), where extremely large doses are delivered, accurate, and precise dose targeting is especially vital to achieve an optimal risk to benefit ratio. This has largely become possible due to the rapid development of image-guided technology. Although imaging is critical to the success of radiotherapy, it can often be plagued with uncertainties due to respiratory-induced target motion. There has and continues to be an immense research effort aimed at acknowledging and addressing these uncertainties to further our abilities to more precisely target radiation treatment. Thus, the goal of this article is to provide a detailed review of the prevailing uncertainties that remain to be investigated across the different imaging modalities, as well as to highlight the more modern solutions to imaging motion and their role in addressing the current challenges.

Low-dose CT with deep learning regularization via proximal forward-backward splitting

Low-dose x-ray computed tomography (LDCT) is desirable for reduced patient dose. This work develops new image reconstruction methods with deep learning (DL) regularization for LDCT. Our methods are based on the unrolling of a proximal forward-backward splitting (PFBS) framework with data-driven image regularization via deep neural networks. In contrast to PFBS-IR, which utilizes standard data fidelity updates via an iterative reconstruction (IR) method, PFBS-AIR involves preconditioned data fidelity updates that fuse the analytical reconstruction (AR) and IR methods in a synergistic way, i.e. fused analytical and iterative reconstruction (AIR). The results suggest that the DL-regularized methods (PFBS-IR and PFBS-AIR) provide better reconstruction quality compared to conventional methods (AR or IR). In addition, owing to the AIR, PFBS-AIR noticeably outperformed PFBS-IR and another DL-based postprocessing method, FBPConvNet.

Dynamic gating window technique for the reduction of dosimetric error in respiratory-gated spot-scanning particle therapy: An initial phantom study using patient tumor trajectory data

Spot-scanning particle therapy possesses advantages, such as high conformity to the target and efficient energy utilization compared with those of the passive scattering irradiation technique. However, this irradiation technique is sensitive to target motion. In the current clinical situation, some motion management techniques, such as respiratory-gated irradiation, which uses an external or internal surrogate, have been clinically applied. In surrogate-based gating, the size of the gating window is fixed during the treatment in the current treatment system. In this study, we propose a dynamic gating window technique, which optimizes the size of gating window for each spot by considering a possible dosimetric error. The effectiveness of the dynamic gating window technique was evaluated by simulating irradiation using a moving target in a water phantom. In dosimetric characteristics comparison, the dynamic gating window technique exhibited better performance in all evaluation volumes with different effective depths compared with that of the fixed gate approach. The variation of dosimetric characteristics according to the target depth was small in dynamic gate compared to fixed gate. These results suggest that the dynamic gating window technique can maintain an acceptable dose distribution regardless of the target depth. The overall gating efficiency of the dynamic gate was approximately equal or greater than that of the fixed gating window. In dynamic gate, as the target depth becomes shallower, the gating efficiency will be reduced, although dosimetric characteristics will be maintained regardless of the target depth. The results of this study suggest that the proposed gating technique may potentially improve the dose distribution. However, additional evaluations should be undertaken in the future to determine clinical applicability by assuming the specifications of the treatment system and clinical situation.

Impact of Esophageal Motion on Dosimetry and Toxicity With Thoracic Radiation Therapy

Purpose:

To investigate the impact of intra- and inter-fractional esophageal motion on dosimetry and observed toxicity in a phase I dose escalation study of accelerated radiotherapy with concurrent chemotherapy for locally advanced lung cancer.

Methods and Materials:

Patients underwent computed tomography imaging for radiotherapy treatment planning (CT1 and 4DCT1) and at 2 weeks (CT2 and 4DCT2) and 5 weeks (CT3 and 4DCT3) after initiating treatment. Each computed tomography scan consisted of 10-phase 4DCTs in addition to a static free-breathing or breath-hold computed tomography. The esophagus was independently contoured on all computed tomographies and 4DCTs. Both CT2 and CT3 were rigidly registered with CT1 and doses were recalculated using the original intensity-modulated radiation therapy plan based on CT1 to assess the impact of interfractional motion on esophageal dosimetry. Similarly, 4DCT1 data sets were rigidly registered with CT1 to assess the impact of intrafractional motion. The motion was characterized based on the statistical analysis of slice-by-slice center shifts (after registration) for the upper, middle, and lower esophageal regions, respectively. For the dosimetric analysis, the following quantities were calculated and assessed for correlation with toxicity grade: the percent volumes of esophagus that received at least 20 Gy (V20) and 60 Gy (V60), maximum esophageal dose, equivalent uniform dose, and normal tissue complication probability.

Results:

The interfractional center shifts were 4.4 ± 1.7 mm, 5.5 ± 2.0 mm and 4.9 ± 2.1 mm for the upper, middle, and lower esophageal regions, respectively, while the intrafractional center shifts were 0.6 ± 0.4 mm, 0.7 ± 0.7 mm, and 0.9 ± 0.7 mm, respectively. The mean V60 (and corresponding normal tissue complication probability) values estimated from the interfractional motion analysis were 7.8% (10%), 4.6% (7.5%), 7.5% (8.6%), and 31% (26%) for grade 0, grade 1, grade 2, and grade 3 toxicities, respectively.

Conclusions:

Interfractional esophageal motion is significantly larger than intrafractional motion. The mean values of V60 and corresponding normal tissue complication probability, incorporating interfractional esophageal motion, correlated positively with esophageal toxicity grade.

Predicting real-time 3D deformation field maps (DFM) based on volumetric cine MRI (VC-MRI) and artificial neural networks for on-board 4D target tracking: a feasibility study

To predict real-time 3D deformation field maps (DFMs) using Volumetric Cine MRI (VC-MRI) and adaptive boosting and multi-layer perceptron neural network (ADMLP-NN) for 4D target tracking. One phase of a prior 4D-MRI is set as the prior phase, MRIprior. Principal component analysis (PCA) is used to extract three major respiratory deformation modes from the DFMs generated between the prior and remaining phases. VC-MRI at each time-step is considered a deformation of MRIprior, where the DFM is represented as a weighted linear combination of the PCA components. The PCA weightings are solved by minimizing the differences between on-board 2D cine MRI and its corresponding VC-MRI slice. The PCA weightings solved during the initial training period are used to train an ADMLP-NN to predict PCA weightings ahead of time during the prediction period. The predicted PCA weightings are used to build predicted 3D DFM and ultimately, predicted VC-MRIs for 4D target tracking. The method was evaluated using a 4D computerized phantom (XCAT) with patient breathing curves and MRI data from a real liver cancer patient. Effects of breathing amplitude change and ADMLP-NN parameter variations were assessed. The accuracy of the PCA curve prediction was evaluated. The predicted real-time 3D tumor was evaluated against the ground-truth using volume dice coefficient (VDC), center-of-mass-shift (COMS), and target tracking errors. For the XCAT study, the average VDC and COMS for the predicted tumor were 0.92  ±  0.02 and 1.06  ±  0.40 mm, respectively, across all predicted time-steps. The correlation coefficients between predicted and actual PCA curves generated through VC-MRI estimation for the 1st/2nd principal components were 0.98/0.89 and 0.99/0.57 in the SI and AP directions, respectively. The optimal number of input neurons, hidden neurons, and MLP-NN for ADMLP-NN PCA weighting coefficient prediction were determined to be 7, 4, and 10, respectively. The optimal cost function threshold was determined to be 0.05. PCA weighting coefficient and VC-MRI accuracy was reduced for increased prediction-step size. Accurate PCA weighting coefficient prediction correlated with accurate VC-MRI prediction. For the patient study, the predicted 4D tumor tracking errors in superior-inferior, anterior-posterior and lateral directions were 0.50  ±  0.47 mm, 0.40  ±  0.55 mm, and 0.28  ±  0.12 mm, respectively. Preliminary studies demonstrated the feasibility to use VC-MRI and artificial neural networks to predict real-time 3D DFMs of the tumor for 4D target tracking.