An algorithm for automatic determination of the respiratory phases in four-dimensional computed tomography

Respiration-phase-matched digital tomosynthesis imaging for moving target verification: a feasibility study

PURPOSE

To develop a respiration-phase-matched digital tomosynthesis (DTS) technique to monitor moving targets, and to evaluate its accuracy for various imaging parameters and anatomical characteristics.

METHODS

Previously developed 3D-DTS techniques, registering onboard DTS (OB-DTS, reconstructed from onboard projections) to reference DTS (R-DTS, reconstructed from DRRs of 3D reference CT), are inadequate to monitor moving targets. The authors' proposed respiration-phase-matched DTS technique registers OB-DTS to R-DTS reconstructed from DRRs generated by the same phase images of 4D reference CT as the corresponding onboard projections. To evaluate the improved accuracy of the author's technique, the authors performed thoracic phantom studies using (1) simulation with the 4D digital extended-cardiac-torso (XCAT) phantom, and (2) experiments with an anthropomorphic motion phantom. The studies were performed for various: respiratory cycle (RC), scan angle, and fraction of RC contained therein. Also, the authors assessed the accuracy of their technique relative to target size/location, and respiration inconsistencies from the R-DTS to OB-DTS.

RESULTS

In both simulation and experimental studies, the respiration-phase-matched DTS technique is significantly more accurate in determining moving target positions. For 324 different scenarios simulated by XCAT, the respiration-phase-matched DTS technique localizes the 3D target position to errors of 1.07 ± 0.57 mm (mean ± S.D.), as compared to (a) 2.58 ± 1.37 and (b) 7.37 ± 4.18 mm, for 3D-DTS using 3D reference CT of (a) average intensity projection and (b) free-breathing CT. For 60 scenarios evaluated through experimental study, the uncertainties corresponding to those above are 1.24 ± 0.87, 2.42 ± 1.80, and 5.77 ± 6.45 mm, respectively. For a given scan angle, the accuracy of respiration-phase-matched DTS technique is less dependent on RC and the fraction of RC included in the scan. Increasing scan angle improves its accuracy. For different target locations, the targets near the chest wall or in the middle of lung provide higher registration accuracy compared to those near the mediastinum and diaphragm. Larger targets provide higher registration accuracy than small targets. Different respiratory cycle inconsistencies from R-DTS to OB-DTS minimally affect the registration accuracy. Increasing the respiratory amplitude inconsistencies will decrease the accuracy.

CONCLUSIONS

The respiration-phase-matched DTS is more accurate and robust in determining moving target positions than 3D-DTS. It has potential application in pretreatment setup, post-treatment analysis, and intrafractional target verification.

A novel technique for markerless, self-sorted 4D-CBCT: feasibility study

PURPOSE

Four-dimensional CBCT (4D-CBCT) imaging in the treatment room can provide verification of moving targets, facilitating the potential for margin reduction and consequent dose escalation. Reconstruction of 4D-CBCT images requires correlation of respiratory phase with projection acquisition, which is often achieved with external surrogate measures of respiration. However, external measures may not be a direct representation of the motion of the internal anatomy and it is therefore the aim of this work to develop a novel technique for markerless, self-sorted 4D-CBCT reconstruction.

METHODS

A novel 4D-CBCT reconstruction technique based on the principles of Fourier transform (FT) theory was investigated for markerless extraction of respiratory phase directly from projection data. In this FT technique, both phase information (FT-phase) and magnitude information (FT-magnitude) were separately implemented in order to discern projections corresponding to peak inspiration, which then facilitated the proceeding sort and bin processes involved in retrospective 4D image reconstruction. In order to quantitatively evaluate the accuracy of the Fourier methods, peak-inspiration projections identified each by FT-phase and FT-magnitude were compared to those manually identified by visual tracking of structures. The average phase difference as assigned by each method vs the manual technique was calculated per projection dataset. The percentage of projections that were assigned within 10% phase of each other was also computed. Both Fourier methods were tested on two phantom datasets, programmed to exhibit sinusoidal respiratory cycles of 2.0 cm in amplitude with respiratory cycle lengths of 3 and 6 s, respectively. Additionally, three sets of patient projections were studied. All of the data were previously acquired at slow-gantry speeds ranging between 0.6°/s and 0.7°/s over a 200° rotation. Ten phase bins with 10% phase windows were selected for 4D-CBCT reconstruction of one phantom and one patient case for visual and quantitative comparison. Line profiles were plotted for the 0% and 50% phase images as reconstructed by the manual technique and each of the Fourier methods.

RESULTS

As compared with the manual technique, the FT-phase method resulted in average phase differences of 1.8% for the phantom with the 3 s respiratory cycle, 3.9% for the phantom with the 6 s respiratory cycle, 2.9% for patient 1, 5.0% for patient 2, and 3.8% for patient 3. For the FT-magnitude method, these numbers were 2.1%, 4.0%, 2.9%, 5.3%, and 3.5%, respectively. The percentage of projections that were assigned within 10% phase by the FT-phase method as compared to the manual technique for the five datasets were 100.0%, 100.0%, 97.6%, 93.4%, and 94.1%, respectively, whereas for the FT-magnitude method these percentages were 98.1%, 92.3%, 98.7%, 87.3%, and 95.7%. Reconstructed 4D phase images for both the phantom and patient case were visually and quantitatively equivalent between each of the Fourier methods vs the manual technique.

CONCLUSIONS

A novel technique employing the basics of Fourier transform theory was investigated and demonstrated to be feasible in achieving markerless, self-sorted 4D-CBCT reconstruction.

Extracting respiratory signals from thoracic cone beam CT projections

The patient respiratory signal associated with the cone beam CT (CBCT) projections is important for lung cancer radiotherapy. In contrast to monitoring an external surrogate of respiration, such a signal can be extracted directly from the CBCT projections. In this paper, we propose a novel local principal component analysis (LPCA) method to extract the respiratory signal by distinguishing the respiration motion-induced content change from the gantry rotation-induced content change in the CBCT projections. The LPCA method is evaluated by comparing with three state-of-the-art projection-based methods, namely the Amsterdam Shroud method, the intensity analysis method and the Fourier-transform-based phase analysis method. The clinical CBCT projection data of eight patients, acquired under various clinical scenarios, were used to investigate the performance of each method. We found that the proposed LPCA method has demonstrated the best overall performance for cases tested and thus is a promising technique for extracting a respiratory signal. We also identified the applicability of each existing method.

A practical method of identifying data loss in 4DCT

BACKGROUND AND PURPOSE

The design, testing and clinical implementation of a simple quality assurance tool which allows quick and accurate identification of regions of data loss and data interpolation in 4DCT data sets is reported.

MATERIALS AND METHODS

A 4DCT model, dependent on gantry rotation time and pitch, was developed to allow an understanding of the data collection and reconstruction processes. To test this model, 4DCT scans of a phantom were acquired using a Siemens SOMATOM Sensation 40 slice CT scanner. A radio-opaque rod mounted under the couch top was present during the phantom scans.

RESULTS

The model predicts that periodic regions of data loss occur when the respiration rate drops below a critical value. These results are verified by experimental data. Regions of data loss result in breaks in the imaged radio-opaque rod.

CONCLUSIONS

Regions of data loss in 4DCT data sets can be difficult to detect. Mounting a radio-opaque rod under the couch top allows regions of data loss and data interpolation to be quickly assessed on a patient by patient basis. This quality assurance tool has been successfully implemented into clinical use. The results of this work have implications for quality assurance programmes for 4DCT scanning.

A new fast and fully automated software based algorithm for extracting respiratory signal from raw PET data and its comparison to other methods

PURPOSE

Respiratory gating in PET is an approach used to minimize the negative effects of respiratory motion on spatial resolution. It is based on an initial determination of a patient's respiratory movements during a scan, typically using hardware based systems. In recent years, several fully automated databased algorithms have been presented for extracting a respiratory signal directly from PET data, providing a very practical strategy for implementing gating in the clinic. In this work, a new method is presented for extracting a respiratory signal from raw PET sinogram data and compared to previously presented automated techniques.

METHODS

The acquisition of respiratory signal from PET data in the newly proposed method is based on rebinning the sinogram data into smaller data structures and then analyzing the time activity behavior in the elements of these structures. From this analysis, a 1D respiratory trace is produced, analogous to a hardware derived respiratory trace. To assess the accuracy of this fully automated method, respiratory signal was extracted from a collection of 22 clinical FDG-PET scans using this method, and compared to signal derived from several other software based methods as well as a signal derived from a hardware system.

RESULTS

The method presented required approximately 9 min of processing time for each 10 min scan (using a single 2.67 GHz processor), which in theory can be accomplished while the scan is being acquired and therefore allowing a real-time respiratory signal acquisition. Using the mean correlation between the software based and hardware based respiratory traces, the optimal parameters were determined for the presented algorithm. The mean/median/range of correlations for the set of scans when using the optimal parameters was found to be 0.58/0.68/0.07-0.86. The speed of this method was within the range of real-time while the accuracy surpassed the most accurate of the previously presented algorithms.

CONCLUSIONS

PET data inherently contains information about patient motion; information that is not currently being utilized. We have shown that a respiratory signal can be extracted from raw PET data in potentially real-time and in a fully automated manner. This signal correlates well with hardware based signal for a large percentage of scans, and avoids the efforts and complications associated with hardware. The proposed method to extract a respiratory signal can be implemented on existing scanners and, if properly integrated, can be applied without changes to routine clinical procedures.

Detection of respiratory motion in fluoroscopic images for adaptive radiotherapy

Respiratory motion limits the potential of modern high-precision radiotherapy techniques such as IMRT and particle therapy. Due to the uncertainty of tumour localization, the ability of achieving dose conformation often cannot be exploited sufficiently, especially in the case of lung tumours. Various methods have been proposed to track the position of tumours using external signals, e.g. with the help of a respiratory belt or by observing external markers. Retrospectively gated time-resolved x-ray computed tomography (4D CT) studies prior to therapy can be used to register the external signals with the tumour motion. However, during treatment the actual motion of internal structures may be different. Direct control of tissue motion by online imaging during treatment promises more precise information. On the other hand, it is more complex, since a larger amount of data must be processed in order to determine the motion. Three major questions arise from this issue. Firstly, can the motion that has occurred be precisely determined in the images? Secondly, how large must, respectively how small can, the observed region be chosen to get a reliable signal? Finally, is it possible to predict the proximate tumour location within sufficiently short acquisition times to make this information available for gating irradiation? Based on multiple studies on a porcine lung phantom, we have tried to examine these questions carefully. We found a basic characteristic of the breathing cycle in images using the image similarity method normalized mutual information. Moreover, we examined the performance of the calculations and proposed an image-based gating technique. In this paper, we present the results and validation performed with a real patient data set. This allows for the conclusion that it is possible to build up a gating system based on image data, solely, or (at least in avoidance of an exceeding exposure dose) to verify gates proposed by the various external systems.

Iterative sorting for four-dimensional CT images based on internal anatomy motion

Current four-dimensional (4D) computed tomography (CT) imaging techniques using multislice CT scanners require retrospective sorting of the reconstructed two-dimensional (2D) CT images. Most existing sorting methods depend on externally monitored breathing signals recorded by extra instruments. External signals may not always accurately capture the breathing status and may lead to severe discontinuity artifacts in the sorted CT volumes. This article describes a method to find the temporal correspondences for the free-breathing multislice CT images acquired at different table positions based on internal anatomy movement. The algorithm iteratively sorts the CT images using estimated internal motion indices. It starts from two imperfect reference volumes obtained from the unsorted CT images; then, in each iteration, thorax motion is estimated from the reference volumes and the free-breathing CT images. Based on the estimated motion, the breathing indices as well as the reference volumes are refined and fed into the next iteration. The algorithm terminates when two successive iterations attain the same sorted reference volumes. In three out of five patient studies, our method attained comparable image quality with that using external breathing signals. For the other two patient studies, where the external signals poorly reflected the internal motion, the proposed method significantly improved the sorted 4D CT volumes, albeit with greater computation time.

Evaluation of geometrical sensitivity for respiratory motion gating by GATE and NCAT simulation

Respiratory motion artifacts can be a significant factor that limits the PET image quality. To improve image quality, surveillance systems have been developed to track the movements of the subject during scanning. Gating techniques utilizing the tracking information, are able to compensate for subject motion, thereby improving lesion detection. In this paper, we present a gating method that utilizes the geometric sensitivity gating (GSG) of a 3D-PET scanner system operating in list event acquisition mode. GSG method has several advantages over the existing methods, it only uses LOR events and is non-invasive, no additional hardware device is required and there is no additional patient preparation required. Using GATE (GEANT4 Application Tomographic Emission) and NCAT (NURBs(Non Uniform Rational B-Splines) Cardiac Torso) software packages with an Allegro PET configuration, realistic simulations of respiratory motion demonstrate that GSG can be used for respiratory gating.

4D treatment planning for scanned ion beams

At Gesellschaft für Schwerionenforschung (GSI) more than 330 patients have been treated with scanned carbon ion beams in a pilot project. To date, only stationary tumors have been treated. In the presence of motion, scanned ion beam therapy is not yet possible because of interplay effects between scanned beam and target motion which can cause severe mis-dosage. We have started a project to treat tumors that are subject to respiratory motion. A prototype beam application system for target tracking with the scanned pencil beam has been developed and commissioned.

To facilitate treatment planning for tumors that are subject to organ motion, we have extended our standard treatment planning system TRiP to full 4D functionality. The 4D version of TRiP allows to calculate dose distributions in the presence of motion. Furthermore, for motion mitigation techniques tracking, gating, rescanning, and internal margins optimization of treatment parameters has been implemented. 4D calculations are based on 4D computed tomography data, deformable registration maps, organ motion traces, and beam scanning parameters.

We describe the methods of our 4D treatment planning approach and demonstrate functionality of the system for phantom as well as patient data.

Is a Single Respiratory Correlated 4D-CT Study Sufficient for Evaluation of Breathing Motion?

PURPOSE

Respiratory correlated computed tomography has been shown to be effective for evaluation of breathing-induced motion of pulmonary tumors. This study investigated whether a single four-dimensional CT study (4D-CT) is representative and sufficient for treatment planning in stereotactic body radiotherapy (SBRT).

METHODS AND MATERIALS

Four repeated helical 4D-CT studies were acquired every 10 min for 10 patients with 14 pulmonary metastases. Patients remained immobilized in a stereotactic body frame (SBF) for 30 min; abdominal compression was applied to seven patients. Using amplitude based sorting, eight phases equally distributed over the breathing cycle were reconstructed for each 4D-CT study. Tumor position was defined in a total of 406 CT series and variability of breathing motion and mean tumor position were evaluated.

RESULTS

Peak-to-peak tumor motion was 9.9 mm +/- 6.8 mm (mean +/- standard deviation) and 9.0 mm +/- 7.4 mm at time point 0 min (t(0)) and t(30), respectively. In one patient with poor pulmonary function, continuous increase of breathing motion from 17.4 mm at t(0) to 28.3 mm at t(30) was seen. In five and two lesions, respectively, a drift of the mean tumor position greater than 3 mm and 5 mm was observed. A borderline significance was calculated for larger tumor position variability in midventilation phases compared with peak-ventilation phases of the breathing cycle (p = 0.08).

CONCLUSION

Treatment planning based on a single 4D-CT study is reliable for the majority of patients. Increased intrafractional uncertainties were seen for patients with poor pulmonary function and with tumors located in the lower lobe.