Four‐dimensional tissue deformation reconstruction (4D TDR) validation using a real tissue phantom

Purpose: Calculation of four dimensional (4D) dose distributions requires the remapping of dose calculated on each available binned phases of the 4DCT onto a reference phase for summation. Deformable image registration (DIR) is usually used for this task. This project validates a type of new algorithm, using 4D tissue deformation reconstruction (4DTDR) from all 4DCT generated images (or projections) to confirm voxel to voxel accuracy for the complete motion path and future 4D dose calculations. Methods: Three electro‐magnetic‐tracking (EMT) fiducial were implanted into fresh porcine liver and animated in a clinically realistic phantom with highly reproducible hysteresis motion pattern previously reported on. After 4DCT acquisition, all data was copied and fiducial locations were masked in the copied set. All images were sent to the 4DTDR for reconstruction. The original 4DCT data (with fiducials) was sampled in 20 CT phase sets and fiducialsˈ coordinates were recorded, resulting in time‐resolved fiducial motion paths. Measured values of fiducial location were compared to that measured through EMT and those calculated by 4DTDR. Results: The performance of the 4DTDR was measured to have average accuracy of 1.86 mm over all 20 phases (4.54 mm max error) in time and location, calculating a motion path that correctly represents measured hysteresis. EMT measurement confirms 4DCT and 4DTDR values. Conclusion: Accuracy of deformation within tissue necessitates the understanding of motion accuracy over the whole motion path of deformation. Validation of the actual voxel path over the period adds a higher level of accuracy, leading away from typically quoted end‐to‐end accuracy to a definition that includes all motion locations at any moment or any location.

ObjectiveTo study the feasibility and the potential benefits of defining the internal gross tumor volume (IGTV) of hepatocellular carcinoma (HCC) using contrast-enhanced 4D CT images obtained by combining arterial-phase (AP) contrast-enhanced (CE) 3D CT and non-contrast-enhanced (NCE) 4D CT images using deformable registration (DR).MethodsTen HCC patients who had received radiotherapy beforehand were selected for this study. The following CT simulation images were acquired sequentially: NCE 4D CT in free breathing, NCE 3D CT and APCE 3D CT in end-expiration breath holding. All 4D CT images were sorted into ten phases according to breath cycle (CT00 ~ CT90). Gross tumor volumes (GTVs) were contoured on all CT images and the IGTV-1 was obtained by merging the GTVs in each phase of 4D CT images. The GTV on the APCE 3D CT image was deformably registered to each 4D CT phase image according to liver shape using RayStationTM 3.99.0.7 version treatment planning system. The IGTV-DR was obtained by merging the GTVs after DR on the 4D CT images. Volume differences among the GTVs and between the IGTV-1 and the IGTV-DR were compared.ResultsThe edge of most lesions could be definitively identified using APCE 3D CT images compared to NCE 4D and 3D CT images. The GTV volume on APCE 3D CT images increased by an average of 34.79% (P < 0.05). There was no significant difference among the GTV volumes obtained using NCE 4D and 3D CT images (P > 0.05). The GTV volumes after DR on 4D CT different phase images increased by an average of 36.29% (P < 0.05), as was observed using the APCE 3D CT image (P > 0.05). Lastly, the volume of IGTV-DR increased by an average of 19.91% compared to that of IGTV-1 (P < 0.05).ConclusionNCE 4D CT imaging alone has the potential risk of missing a partial volume of the HCC. The combination of APCE 3D CT and NCE 4D CT images using the DR technique improved the accuracy of the definition of the IGTV in HCC.

Purpose: Particularly in the case of thoracic radiation therapy, there are substantial inter‐ and intra‐fractional variations in shape, volume and position of treatment targets and the intervening and surrounding normal tissues. The purpose of this work is to develop proton treatment planning strategies for mobile tumors with and without mobile intervening structures based on 4D CT, and to assess the planning strategies using 4D CT data and daily in‐room CT information. Method and Materials: Five treatment planning strategies were evaluated based on (1) free breathing CT with small smearing, (2) free breathing CT with large smearing, (3) average CT with small smearing, (4) average CT with CT numbers inside the tumor volume replaced by higher CT numbers, and (5) maximum intensity projection CT. For a lung patient with large, 1.6 cm, tumor motion and immobile surrounding tissues, treatment plans were designed using strategies 1, 2 and 4. Each treatment plan was recalculated on five daily CT images using bony structure alignment. For an esophagus patient with 3.5 cm tumor motion and large cardiac, liver and spleen motion, treatment plans were designed using strategies 1, 2, 3 and 5. Each treatment plan was recalculated in all 10 phases of the 4D CT. Results: For the lung patient, the tumor coverage evaluated using the five daily CT's is superior when using strategies 2 and 4 as compared to strategy 1. However, the lung sparing is superior using strategy 4 compared to strategy 2. For the esophagus patient, the treatment plans using strategy 2 and 5 covered the targets for each 4D CT phase while the plan using strategy 1 and 3 caused significant under‐dosing. Conclusion: With the strategies developed using 4D CT data, good tumor coverage was achieved for the thoracic patients with large tumor and surrounding tissue motion using proton therapy.

We have previously developed a retrospective 4D‐MRI technique using body area as the respiratory surrogate, but generally, the reconstructed 4D MR images suffer from severe or mild artifacts mainly caused by irregular motion during image acquisition. Those image artifacts may potentially affect the accuracy of tumor target delineation or the shape representation of surrounding nontarget tissues and organs. So the purpose of this study is to propose an approach employing principal component analysis (PCA), combined with a linear polynomial fitting model, to remodel the displacement vector fields (DVFs) obtained from deformable image registration (DIR), with the main goal of reducing the motion artifacts in 4D MR images. Seven patients with hepatocellular carcinoma (2/7) or liver metastases (5/7) in the liver, as well as a patient with non‐small cell lung cancer (NSCLC), were enrolled in an IRB‐approved prospective study. Both CT and MR simulations were performed for each patient for treatment planning. Multiple‐slice, multiple‐phase, cine‐MRI images were acquired in the axial plane for 4D‐MRI reconstruction. Single‐slice 2D cine‐MR images were acquired across the center of the tumor in axial, coronal, and sagittal planes. For a 4D MR image dataset, the DVFs in three orthogonal direction (inferior–superior (SI), anterior–posterior (AP), and medial–lateral (ML)) relative to a specific reference phase were calculated using an in‐house DIR algorithm. The DVFs were preprocessed in three temporal and spatial dimensions using a polynomial fitting model, with the goal of correcting the potential registration errors introduced by three‐dimensional DIR. Then PCA was used to decompose each fitted DVF into a linear combination of three principal motion bases whose spanned subspaces combined with their projections had been validated to be sufficient to represent the regular respiratory motion. By wrapping the reference MR image using the remodeled DVFs, ‘synthetic’ MR images with reduced motion artifacts were generated at selected phase. Tumor motion trajectories derived from cine‐MRI, 4D CT, original 4D MRI, and ‘synthetic’ 4D MRI were analyzed in the SI, AP, and ML directions, respectively. Their correlation coefficient (CC) and difference (D) in motion amplitude were calculated for comparison. Of all the patients, the means and standard deviations (SDs) of CC comparing ‘synthetic’ 4D MRI and cine‐MRI were 0.98±0.01,0.98±0,01, and 0.99±0.01 in SI, AP, and ML directions, respectively. The mean±SD Ds were 0.59±0.09 mm,0.29±0.10 mm, and 0.15±0.05 mm in SI, AP and ML directions, respectively. The means and SDs of CC comparing ‘synthetic’ 4D MRI and 4D CT were 0.96±0.01,0.95±0.01, and 0.95±0.01 in SI, AP, and ML directions, respectively. The mean±SD Ds were 0.76±0.20 mm,0.33±0.14 mm, and 0.19±0.07 mm in SI, AP, and ML directions, respectively. The means and SDs of CC comparing ‘synthetic’ 4D MRI and original 4D MRI were 0.98±0.01,0.98±0.01, and 0.97±0.01 in SI, AP, and ML directions, respectively. The mean±SD Ds were 0.58±0.10 mm,0.30±0.09 mm, and 0.17±0.04 mm in SI, AP, and ML directions, respectively. In this study we have proposed an approach employing PCA combined with a linear polynomial fitting model to capture the regular respiratory motion from a 4D MR image dataset. And its potential usefulness in reducing motion artifacts and improving image quality has been demonstrated by the preliminary results in oncological patients.PACS numbers: 87.57.cp, 87.57.nj, 87.61.‐c

Due to lack of soft-tissue contrast, target distortion for the upper-abdomen targets such as pancreatic tumors is complicated and requiring sufficient remedy. By applying automatic contour propagation, the authors use the information obtained from 4D CT to test if the deformable image registration compensates the respiration-induced distortion of pancreatic tumor in free breathing (FB) CT images. Ten patients with unresected pancreatic cancer treated with either preoperative or definitive chemoradiation were studied. Pancreas GTVs were delineated on the FB CT. Using deformable image registration, the FB GTV contours were propagated to each phase of the 4D CT images taken right after the FB CT, and were compared with the FB GTV to see difference in tumor volume and tumor size along individual dimensions. A one-dimensional tumor motion in proportion to cos4(ωt) was simulated to calculate the probability distribution function for different magnitude of distortions during FB CT scans, and a binary classification test was conducted to analyze the observed results. The probability distribution function predicted that four out of the ten cases would have substantial target distortion given the variation in target motion amplitudes. Three of these four cases show substantial difference in the superior-inferior size of FB GTV compared to the average 4D GTV, taking into account the uncertainties caused by motions perpendicular to the scanning axis and resolution of the CT scanner. The binary classification test yielded a precision of 75% and an accuracy of 90%. Pancreatic GTV distorted due to respiration-induced tumor motion is effectively compensated by contour propagation from free-breathing CT to 4D CT using DIR. Union of GTVs of all breathing phases or IGTV can be genreated from 4D set of GTVs propagated from that of free breathing. This study is partially supported by NIH grant 1R01CA133539-01A2. I do not have conflict of interest.

Objective To evaluate the dose variation of target coverage and organs at risk (OARs) among four planning strategies using spot-scanning carbon-ion radiotherapy for non-small cell lung cancer (NSCLC). Methods Ten NSCLC patients utilizing gating motion control were selected to receive dose calculation over multiple acquired 4DCT images. Four optimizing strategies consisted of intensity-modulated carbon-ion therapy (IMCT-NoAS), IMCT combined with internal gross tumor volume (IGTV) assigned muscle (IMCT-ASM), single beam optimization (SBO)(SBO-NoAS) and SBO combined with IGTV assigned muscle (SBO-ASM). The initial plan was re-calculated after the 4DCT data were reviewed and then compared with the initial plan in the dosimetry. Results For re-calculation plans with two reviewing CTs, all four strategies yielded similar planning target volume (PTV) coverage. Merely IMCT-NoAS strategy presented with relatively significant variations in dose distribution. Dose variation for OARs between initial and re-calculated plans: for all four strategies, V20 of ipsilateral lung was increased by approximately 2.0 Gy (relative biological effective dose, RBE), V30 of heart was increased by approximately 1.0 Gy (RBE) for both IGTV assigned muscle strategies, whereas decreased by approximately 0.2 Gy (RBE) for both IGTV non-assigned muscle strategies. The maximum dose of spinal cord was changed by 2.5 Gy (RBE). Conclusions Carbon-ion radiotherapy is sensitive to the anatomic motion within the tumors along the beam path. When the tumor motion along the head-foot (H-F) direction exceeds 8 mm, SBO-ASM strategy provides better dose coverage of target. Strategies with IGTV assignment may result in dose overshoot to a position deeper than the initial planning dose distribution. Key words: Lung neoplasms/intensity-modulated carbon-ion therapy; Single beam optimization; Spot scanning carbon-ion; Plan evaluation

It has been shown that non-resonant multiphoton photoacoustic spectroscopy (NMPPAS) has a great potential to be used as a high resolution surgical guidance technique during brain tumor surgery due to its ability of non-invasive or minimally invasive tumor differentiation. However, for experimental purposes associated with method validation, the use of real tissues is not always ideal because of issues such as availability, safety, storage, chemical doping, necessary control of size and shape, etc. To overcome these issues, tissue phantoms made from animal tissues and/or biochemical constituents, are often employed for such analyses. This work demonstrates the ability to develop and characterize gelatin based tissue phantoms with comparable optical and acoustic properties to real tissues by doping the phantoms with a scattering substance, 0.3 mm diameter Al2O3 particles. Using these phantoms, light scattering coefficients (ms) of 39 cm-1 have been generated, which are comparable to real brain tissue, thus making them a great alternative to real tissue for validation studies. In addition, this work also investigates the non-fluorescent species NAD+ found in the tissues, to evaluate its potential for being detected by NMPPAS. NMPPAS spectra of NAD+ shows a very promising beginning to determine other chemical species such as flavins, collagen, tryptophan, etc responsible for NMPPAS spectral signatures, associated with tumorogenesis.© (2012) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.

Quantitative Assessment of Lung Function in Stage I and Stage III Lung Radiation Therapy Patients Using 4DCT-based Ventilation Imaging

Purpose: 4D CT is useful clinically for detailed abdominal and thoracic imaging over the course of the respiratory cycle. However, it usually delivers 10∼15 times more radiation dose to the patient as compared to the standard 3D CT, since multiple scans at each couch position are required to obtain the temporal information. In this work we propose a method to obtain high quality 4D CT with low tube current, hence reducing the radiation exposure of patients. Method and Materials: The improvement of the signal‐to‐noise ratio (SNR) of the CT image at a given phase was achieved by superposing the imaging information from other phases with the use of a deformable image registration model. To further reduce the statistical noise caused by low tube current, we developed a novel 4D penalized weighted least square (4D‐PWLS) method to smooth the data spatially and temporally. The method was validated by motion‐phantom and patient studies using a GE Discovery‐ST PET/CT scanner. A Varian RPM respiratory gating system was used to track the motion and to facilitate the phase binning of the 4D CT data. Results: We calculated the SNRs for both studies. The average SNR of 10 mA phantom images increased by more than three‐fold from 0.051 to 0.165 after the proposed 4D‐PWLS processing, without noticeable resolution loss. The patient images acquired at 90mA showed an increase from 2.204 to 4.558 for the end‐inspiration phase, and from 1.741 to 3.862 for the end‐expiration phase, respectively. By examining the subtraction images before and after processing, good edge preservation was also observed in the patient study. Conclusion: By appropriately utilizing the temporal information in 4D‐CT, the proposed method effectively suppresses the noise while preserving the resolution. The technique provides a useful way to reduce the patient dose during 4D CT and is thus valuable for 4D‐radiotherapy.

IML-UNet: A brain-inspired spatiotemporal collaborative encoding method for medical image sequence registration.

To present a new approach to the problem of estimating errors in deformable image registration (DIR) applied to sequential phases of a 4DCT data set. A set of displacement vector fields (DVFs) are made by registering a sequence of 4DCT phases. The DVFs are assumed to display anatomical movement, with the addition of errors due to the imaging and registration processes. The positions of physical landmarks in each CT phase are measured as ground truth for the physical movement in the DVF. Principal component analysis of the DVFs and the landmarks is used to identify and separate the eigenmodes of physical movement from the error eigenmodes. By subtracting the physical modes from the principal components of the DVFs, the registration errors are exposed and reconstructed as DIR error maps. The method is demonstrated via a simple numerical model of 4DCT DVFs that combines breathing movement with simulated maps of spatially correlated DIR errors. The principal components of the simulated DVFs were observed to share the basic properties of principal components for actual 4DCT data. The simulated error maps were accurately recovered by the estimation method. Deformable image registration errors can have complex spatial distributions. Consequently, point-by-point landmark validation can give unrepresentative results that do not accurately reflect the registration uncertainties away from the landmarks. The authors are developing a method for mapping the complete spatial distribution of DIR errors using only a small number of ground truth validation landmarks.

The purpose of the study is to evaluate the accuracy of two deformable image registration algorithms by examining their influence on the dose summation results obtained using 4DCT (four dimensional computed tomography) dose distributions based on '4D' planned and '4D optimal' IMRT (intensity modulated radiation therapy) plans. For ten lung cancer patients, 4D step and shoot IMRT plans were produced. The breathing cycle was divided into ten parts and for each part a set of CT images was acquired. For each patient the treatment plan was copied to the CTs of each phase and subsequently recalculated. Each phase CT was then registered to the average intensity projection (AIP) CT using a deformable image registration (DIR) algorithm and the composite dose distribution was then calculated by summing up the deformed dose distributions from all the phases ('4D' treatment plan). The '4D optimal' treatment plan was created by producing an optimal plan on the CTs of each phase of the respiratory cycle and summing up the deformed dose distributions from all the phases. The results indicate that it is possible to map the dose distributions of different breathing phases in lung using DIR, and that different DIR methods and target characteristics (motion amplitude, size, location) affect the differences between original plan, '4D' and '4D optimal' dose distributions. Although the '4D optimal' plans were designed to achieve 95% target coverage, both of the used DIR methods failed to translate that coverage in some instances. The same variation between these methods was also observed in the '4D' plan comparison. This study shows that it is feasible to perform an acceptably accurate calculation of the composite deformed dose. However, it is important to account for tumor motion and body deformation especially when the tumor volume is small and/or located in the lower lobe of the lung.

Breathing-induced location uncertainties of internal structures are still a relevant issue in the radiation therapy of thoracic and abdominal tumours. Motion compensation approaches like gating or tumour tracking are usually driven by low-dimensional breathing signals, which are acquired in real-time during the treatment. These signals are only surrogates of the internal motion of target structures and organs at risk, and, consequently, appropriate models are needed to establish correspondence between the acquired signals and the sought internal motion patterns. In this work, we present a diffeomorphic framework for correspondence modelling based on the Log-Euclidean framework and multivariate regression. Within the framework, we systematically compare standard and subspace regression approaches (principal component regression, partial least squares, canonical correlation analysis) for different types of common breathing signals (1D: spirometry, abdominal belt, diaphragm tracking; multi-dimensional: skin surface tracking). Experiments are based on 4D CT and 4D MRI data sets and cover intra- and inter-cycle as well as intra- and inter-session motion variations. Only small differences in internal motion estimation accuracy are observed between the 1D surrogates. Increasing the surrogate dimensionality, however, improved the accuracy significantly; this is shown for both 2D signals, which consist of a common 1D signal and its time derivative, and high-dimensional signals containing the motion of many skin surface points. Eventually, comparing the standard and subspace regression variants when applied to the high-dimensional breathing signals, only small differences in terms of motion estimation accuracy are found.

PNMC: Four-dimensional conebeam CT reconstruction combining prior network and motion compensation

In this work we used 4D dose calculations, which include the effects of shape deformations, to investigate an alternative approach to creating the ITV. We hypothesized that instead of needing images from all the breathing phases in the 4D CT dataset to create the outer envelope used for treatment planning, it is possible to exclude images from the phases closest to the inhale phase. We used 4D CT images from 10 patients with lung cancer. For each patient, we drew a gross tumor volume on the exhale‐phase image and propagated this to the images from other phases in the 4D CT dataset using commercial image registration software. We created four different ITVs using the N phases closest to the exhale phase (where N=10, 8, 7, 6). For each ITV contour, we created a volume‐modulated arc therapy plan on the exhale‐phase CT and normalized it so that the prescribed dose covered at least 95% of the ITV. Each plan was applied to CT images from each CT phase (phases 1–10), and the calculated doses were then mapped to the exhale phase using deformable registration. The effect of the motion was quantified using the dose to 95% of the target on the exhale phase (D95) and tumor control probability. For the three‐dimensional and 4D dose calculations of the plan where N=10, differences in the D95 value varied from 3% to 14%, with an average difference of 7%. For 9 of the 10 patients, the reduction in D95 was less than 5% if eight phases were used to create the ITV. For three of the 10 patients, the reduction in the D95 was less than 5% if seven phases were used to create the ITV. We were unsuccessful in creating a general rule that could be used to create the ITV. Some reduction (8/10 phases) was possible for most, but not all, of the patients, and the ITV reduction was small.PACS number: 87.55.D‐