Yu-Chi Hu

Radiotherapy treatment planning for patients with non-small cell lung cancer using positron emission tomography (PET)

PURPOSE Many patients with non-small cell lung cancer (NSCLC) receive external beam radiation therapy as part of their treatment. Three-dimensional conformal radiation therapy (3DCRT) commonly uses computed tomography (CT) to accurately delineate the target lesion and normal tissues. Clinical studies, however, indicate that positron emission tomography (PET) has higher sensitivity than CT in detecting and staging of mediastinal metastases. Imaging with fluoro-2-deoxyglucose (FDG) PET in conjunction with CT, therefore, can improve the accuracy of lesion definition. In this pilot study, we investigated the potential benefits of incorporating PET data into the conventional treatment planning of NSCLC. Case-by-case, we prospectively analyzed planning target volume (PTV) and lung toxicity changes for a cohort of patients. MATERIALS AND METHODS We have included 11 patients in this study. They were immobilized in the treatment position and CT simulation was performed. Following CT simulation, PET scanning was performed in the treatment position using the same body cast that was produced for CT simulation and treatment. The PTV, along with the gross target volume (GTV) and normal organs, was first delineated using the CT data set. The CT and PET transmission images were then registered in the treatment planning system using either manual or automated methods, leading to consequent registration of the CT and emission images. The PTV was then modified using the registered PET emission images. The modified PTV is seen simultaneously on both CT and PET images, allowing the physician to define the PTV utilizing the information from both data sets. Dose-volume histograms (DVHs) for lesion and normal organs were generated using both CT-based and PET+CT-based treatment plans. RESULTS For all patients, there was a change in PTV outline based on CT images versus CT/PET fused images. In seven out of 11 cases, we found an increase in PTV volume (average increase of 19%) to incorporate distant nodal disease. Among these patients, the highest normal-tissue complication probability (NTCP) for lung was 22% with combined PET/CT plan and 21% with CT-only plan. In other four patients PTV was decreased an average of 18%. The reduction of PTV in two of these patients was due to excluding atelectasis and trimming the target volume to avoid delivering higher radiation doses to nearby spinal cord or heart. CONCLUSIONS The incorporation of PET data improves definition of the primary lesion by including positive lymph nodes into the PTV. Thus, the PET data reduces the likelihood of geographic misses and hopefully improves the chance of achieving local control.

Treatment planning for prostate implants using magnetic-resonance spectroscopy imaging

PURPOSE Recent studies have demonstrated that magnetic-resonance spectroscopic imaging (MRSI) of the prostate may effectively distinguish between regions of cancer and normal prostatic epithelium. This diagnostic imaging tool takes advantage of the increased choline plus creatine versus citrate ratio found in malignant compared to normal prostate tissue. The purpose of this study is to describe a novel brachytherapy treatment-planning optimization module using an integer programming technique that will utilize biologic-based optimization. A method is described that registers MRSI to intraoperative-obtained ultrasound images and incorporates this information into a treatment-planning system to achieve dose escalation to intraprostatic tumor deposits. METHODS MRSI was obtained for a patient with Gleason 7 clinically localized prostate cancer. The ratios of choline plus creatine to citrate for the prostate were analyzed, and regions of high risk for malignant cells were identified. The ratios representing peaks on the MR spectrum were calculated on a spatial grid covering the prostate tissue. A procedure for mapping points of interest from the MRSI to the ultrasound images is described. An integer-programming technique is described as an optimization module to determine optimal seed distribution for permanent interstitial implantation. MRSI data are incorporated into the treatment-planning system to test the feasibility of dose escalation to positive voxels with relative sparing of surrounding normal tissues. The resultant tumor control probability (TCP) is estimated and compared to TCP for standard brachytherapy-planned implantation. RESULTS The proposed brachytherapy treatment-planning system is able to achieve a minimum dose of 120% of the 144 Gy prescription to the MRS positive voxels using (125)I seeds. The preset dose bounds of 100-150% to the prostate and 100-120% to the urethra were maintained. When compared to a standard plan without MRS-guided optimization, the estimated TCP for the MRS-optimized plan is superior. The enhanced TCP was more pronounced for smaller volumes of intraprostatic tumor deposits compared to estimated TCP values for larger lesions. CONCLUSIONS Using this brachytherapy-optimization system, we could demonstrate the feasibility of MRS-optimized dose distributions for (125)I permanent prostate implants. Based on probability estimates of anticipated improved TCP, this approach may have an impact on the ability to safely escalate dose and potentially improve outcome for patients with organ-confined but aggressive prostatic cancers. The magnitude of the TCP enhancement, and therefore the risks of ignoring the MR data, appear to be more substantial when the tumor is well localized; however, the gain achievable in TCP may depend quite considerably on the MRS tumor-detection efficiency.

A patient-specific respiratory model of anatomical motion for radiation treatment planning

The modeling of respiratory motion is important for a more accurate understanding and accounting of its effect on dose to cancers in the thorax and abdomen by radiotherapy. We have developed a model of respiration-induced organ motion in the thorax without the commonly adopted assumption of repeatable breath cycles. The model describes the motion of a volume of interest within the patient based on a reference three-dimensional (3D) image (at end expiration) and the diaphragm positions at different time points. The input data are respiration-correlated CT (RCCT) images of patients treated for non-small- cell lung cancer, consisting of 3D images, including the diaphragm positions, at ten phases of the respiratory cycle. A deformable image registration algorithm calculates the deformation field that maps each 3D image to the reference 3D image. A principal component analysis is performed to parameterize the 3D deformation field in terms of the diaphragm motion. We show that the first two principal components are adequate to accurately and completely describe the organ motion in the data of four patients. Artifacts in the RCCT images that commonly occur at the mid-respiration states are reduced in the model-generated images. Further validation of the model is demonstrated in the successful application of the parameterized 3D deformation field to RCCT data of the same patient but acquired several days later. We have developed a method for predicting respiration-induced organ motion in patients that has potential for improving the accuracy of dose calculation in radiotherapy. Possible limitations of the model are cases where the correlation between lung tumor and diaphragm position is less reliable such as superiorly situated tumors and interfraction changes in tumor-diaphragm correlation. The limited number of clinical cases examined suggests, but does not confirm, the models applicability to a wide range of patients.

Correction of motion artifacts in cone-beam CT using a patient-specific respiratory motion model

PURPOSE Respiratory motion adversely affects CBCT image quality and limits its localization accuracy for image-guided radiation treatment. Motion correction methods in CBCT have focused on the thorax because of its higher soft tissue contrast, whereas low-contrast tissue in abdomen remains a challenge. The authors report on a method to correct respiration-induced motion artifacts in 1 min CBCT scans that is applicable in both thorax and abdomen, using a motion model adapted to the patient from a respiration-correlated image set. METHODS Model adaptation consists of nonrigid image registration that maps each image to a reference image in the respiration-correlated set, followed by a principal component analysis to reduce errors in the nonrigid registration. The model parametrizes the deformation field in terms of observed surrogate (diaphragm or implanted marker) position and motion (inhalation or exhalation) between the images. In the thorax, the model is obtained from the same CBCT images that are to be motion-corrected, whereas in the abdomen, the model uses respiration-correlated CT (RCCT) images acquired prior to the treatment session. The CBCT acquisition is a single 360° rotation lasting 1 min, while simultaneously recording patient breathing. The approximately 600 projection images are sorted into six (in thorax) or ten (in abdomen) subsets and reconstructed to obtain a set of low-quality respiration-correlated RC-CBCT images. Application of the motion model deforms each of the RC-CBCT images to a chosen reference image in the set; combining all images yields a single high-quality CBCT image with reduced blurring and motion artifacts. Repeated application of the model with different reference images produces a series of motion-corrected CBCT images over the respiration cycle, for determining the motion extent of the tumor and nearby organs at risk. The authors also investigate a simpler correction method, which does not use PCA and correlates motion state with respiration phase, thus assuming repeatable breathing patterns. Comparison of contrast-to-noise ratios of pixel intensities within anatomical structures relative to surrounding background tissue provides a quantitative assessment of relative organ visibility. RESULTS Evaluation in lung phantom, two patient cases in thorax and two in upper abdomen, shows that blurring and streaking artifacts are visibly reduced with motion correction. The boundaries of tumors in the thorax, liver, and kidneys are sharper and more discernible. Repeat application of the method in one thorax case, with reference images chosen at end expiration and end inspiration, indicates its feasibility for observing tumor motion extent. Phase-based motion correction without PCA reduces blurring less effectively; in addition, implanted markers appear broken up, indicating inconsistencies in the phase-based correction. In structures showing 1 cm or more motion excursion, PCA-based motion correction shows the highest contrast-to-noise ratios in the cases examined. CONCLUSIONS Motion correction of CBCT is feasible and yields observable improvement in the thorax and abdomen. The PCA-based model is an important component: First, by reducing deformation errors caused by the nonrigid registration and second, by relating deformation to surrogate position rather than phase, thus accommodating breathing pattern changes between imaging sessions. The accuracy of the method requires confirmation in further patient studies.

A stereotactic method for the three-dimensional registration of multi-modality biologic images in animals: NMR, PET, histology, and autoradiography.

The objective of this work was to develop and then validate a stereotactic fiduciary marker system for tumor xenografts in rodents which could be used to co-register magnetic resonance imaging (MRI), PET, tissue histology, autoradiography, and measurements from physiologic probes. A Teflon fiduciary template has been designed which allows the precise insertion of small hollow Teflon rods (0.71 mm diameter) into a tumor. These rods can be visualized by MRI and PET as well as by histology and autoradiography on tissue sections. The methodology has been applied and tested on a rigid phantom, on tissue phantom material, and finally on tumor bearing mice. Image registration has been performed between the MRI and PET images for the rigid Teflon phantom and among MRI, digitized microscopy images of tissue histology, and autoradiograms for both tissue phantom and tumor-bearing mice. A registration accuracy, expressed as the average Euclidean distance between the centers of three fiduciary markers among the registered image sets, of 0.2 +/- 0.06 mm was achieved between MRI and microPET image sets of a rigid Teflon phantom. The fiduciary template allows digitized tissue sections to be co-registered with three-dimensional MRI images with an average accuracy of 0.21 and 0.25 mm for the tissue phantoms and tumor xenografts, respectively. Between histology and autoradiograms, it was 0.19 and 0.21 mm for tissue phantoms and tumor xenografts, respectively. The fiduciary marker system provides a coordinate system with which to correlate information from multiple image types, on a voxel-by-voxel basis, with sub-millimeter accuracy--even among imaging modalities with widely disparate spatial resolution and in the absence of identifiable anatomic landmarks.

Reduction of irregular breathing artifacts in respiration-correlated CT images using a respiratory motion model

PURPOSE Respiration-correlated CT (RCCT) images produced with commonly used phase-based sorting of CT slices often exhibit discontinuity artifacts between CT slices, caused by cycle-to-cycle amplitude variations in respiration. Sorting based on the displacement of the respiratory signal yields slices at more consistent respiratory motion states and hence reduces artifacts, but missing image data (gaps) may occur. The authors report on the application of a respiratory motion model to produce an RCCT image set with reduced artifacts and without missing data. METHODS Input data consist of CT slices from a cine CT scan acquired while recording respiration by monitoring abdominal displacement. The model-based generation of RCCT images consists of four processing steps: (1) displacement-based sorting of CT slices to form volume images at 10 motion states over the cycle; (2) selection of a reference image without gaps and deformable registration between the reference image and each of the remaining images; (3) generation of the motion model by applying a principal component analysis to establish a relationship between displacement field and respiration signal at each motion state; (4) application of the motion model to deform the reference image into images at the 9 other motion states. Deformable image registration uses a modified fast free-form algorithm that excludes zero-intensity voxels, caused by missing data, from the image similarity term in the minimization function. In each iteration of the minimization, the displacement field in the gap regions is linearly interpolated from nearest neighbor nonzero intensity slices. Evaluation of the model-based RCCT examines three types of image sets: cine scans of a physical phantom programmed to move according to a patient respiratory signal, NURBS-based cardiac torso (NCAT) software phantom, and patient thoracic scans. RESULTS Comparison in physical motion phantom shows that object distortion caused by variable motion amplitude in phase-based sorting is visibly reduced with model-based RCCT. Comparison of model-based RCCT to original NCAT images as ground truth shows best agreement at motion states whose displacement-sorted images have no missing slices, with mean and maximum discrepancies in lung of 1 and 3 mm, respectively. Larger discrepancies correlate with motion states having a larger number of missing slices in the displacement-sorted images. Artifacts in patient images at different motion states are also reduced. Comparison with displacement-sorted patient images as a ground truth shows that the model-based images closely reproduce the ground truth geometry at different motion states. CONCLUSIONS Results in phantom and patient images indicate that the proposed method can produce RCCT image sets with reduced artifacts relative to phase-sorted images, without the gaps inherent in displacement-sorted images. The method requires a reference image at one motion state that has no missing data. Highly irregular breathing patterns can affect the methods performance, by introducing artifacts in the reference image (although reduced relative to phase-sorted images), or in decreased accuracy in the image prediction of motion states containing large regions of missing data.

Evaluation of deformable image registration and a motion model in CT images with limited features

Deformable image registration (DIR) is increasingly used in radiotherapy applications and provides the basis for a previously described model of patient-specific respiratory motion. We examine the accuracy of a DIR algorithm and a motion model with respiration-correlated CT (RCCT) images of software phantom with known displacement fields, physical deformable abdominal phantom with implanted fiducials in the liver and small liver structures in patient images. The motion model is derived from a principal component analysis that relates volumetric deformations with the motion of the diaphragm or fiducials in the RCCT. Patient data analysis compares DIR with rigid registration as ground truth: the mean ± standard deviation 3D discrepancy of liver structure centroid positions is 2.0 ± 2.2 mm. DIR discrepancy in the software phantom is 3.8 ± 2.0 mm in lung and 3.7 ± 1.8 mm in abdomen; discrepancies near the chest wall are larger than indicated by image feature matching. Markers 3D discrepancy in the physical phantom is 3.6 ± 2.8 mm. The results indicate that visible features in the images are important for guiding the DIR algorithm. Motion model accuracy is comparable to DIR, indicating that two principal components are sufficient to describe DIR-derived deformation in these datasets.

Predictive Treatment Management: Incorporating a Predictive Tumor Response Model Into Robust Prospective Treatment Planning for Non-Small Cell Lung Cancer

PURPOSE We hypothesized that a treatment planning technique that incorporates predicted lung tumor regression into optimization, predictive treatment planning (PTP), could allow dose escalation to the residual tumor while maintaining coverage of the initial target without increasing dose to surrounding organs at risk (OARs). METHODS AND MATERIALS We created a model to estimate the geometric presence of residual tumors after radiation therapy using planning computed tomography (CT) and weekly cone beam CT scans of 5 lung cancer patients. For planning purposes, we modeled the dynamic process of tumor shrinkage by morphing the original planning target volume (PTVorig) in 3 equispaced steps to the predicted residue (PTVpred). Patients were treated with a uniform prescription dose to PTVorig. By contrast, PTP optimization started with the same prescription dose to PTVorig but linearly increased the dose at each step, until reaching the highest dose achievable to PTVpred consistent with OAR limits. This method is compared with midcourse adaptive replanning. RESULTS Initial parenchymal gross tumor volume (GTV) ranged from 3.6 to 186.5 cm(3). On average, the primary GTV and PTV decreased by 39% and 27%, respectively, at the end of treatment. The PTP approach gave PTVorig at least the prescription dose, and it increased the mean dose of the true residual tumor by an average of 6.0 Gy above the adaptive approach. CONCLUSIONS PTP, incorporating a tumor regression model from the start, represents a new approach to increase tumor dose without increasing toxicities, and reduce clinical workload compared with the adaptive approach, although model verification using per-patient midcourse imaging would be prudent.

Interactive semiautomatic contour delineation using statistical conditional random fields framework

PURPOSE Contouring a normal anatomical structure during radiation treatment planning requires significant time and effort. The authors present a fast and accurate semiautomatic contour delineation method to reduce the time and effort required of expert users. METHODS Following an initial segmentation on one CT slice, the user marks the target organ and nontarget pixels with a few simple brush strokes. The algorithm calculates statistics from this information that, in turn, determines the parameters of an energy function containing both boundary and regional components. The method uses a conditional random field graphical model to define the energy function to be minimized for obtaining an estimated optimal segmentation, and a graph partition algorithm to efficiently solve the energy function minimization. Organ boundary statistics are estimated from the segmentation and propagated to subsequent images; regional statistics are estimated from the simple brush strokes that are either propagated or redrawn as needed on subsequent images. This greatly reduces the user input needed and speeds up segmentations. The proposed method can be further accelerated with graph-based interpolation of alternating slices in place of user-guided segmentation. CT images from phantom and patients were used to evaluate this method. The authors determined the sensitivity and specificity of organ segmentations using physician-drawn contours as ground truth, as well as the predicted-to-ground truth surface distances. Finally, three physicians evaluated the contours for subjective acceptability. Interobserver and intraobserver analysis was also performed and Bland-Altman plots were used to evaluate agreement. RESULTS Liver and kidney segmentations in patient volumetric CT images show that boundary samples provided on a single CT slice can be reused through the entire 3D stack of images to obtain accurate segmentation. In liver, our method has better sensitivity and specificity (0.925 and 0.995) than region growing (0.897 and 0.995) and level set methods (0.912 and 0.985) as well as shorter mean predicted-to-ground truth distance (2.13 mm) compared to regional growing (4.58 mm) and level set methods (8.55 mm and 4.74 mm). Similar results are observed in kidney segmentation. Physician evaluation of ten liver cases showed that 83% of contours did not need any modification, while 6% of contours needed modifications as assessed by two or more evaluators. In interobserver and intraobserver analysis, Bland-Altman plots showed our method to have better repeatability than the manual method while the delineation time was 15% faster on average. CONCLUSIONS Our method achieves high accuracy in liver and kidney segmentation and considerably reduces the time and labor required for contour delineation. Since it extracts purely statistical information from the samples interactively specified by expert users, the method avoids heuristic assumptions commonly used by other methods. In addition, the method can be expanded to 3D directly without modification because the underlying graphical framework and graph partition optimization method fit naturally with the image grid structure.PURPOSE Contouring a normal anatomical structure during radiation treatment planning requires significant time and effort. The authors present a fast and accurate semiautomatic contour delineation method to reduce the time and effort required of expert users. METHODS Following an initial segmentation on one CT slice, the user marks the target organ and nontarget pixels with a few simple brush strokes. The algorithm calculates statistics from this information that, in turn, determines the parameters of an energy function containing both boundary and regional components. The method uses a conditional random field graphical model to define the energy function to be minimized for obtaining an estimated optimal segmentation, and a graph partition algorithm to efficiently solve the energy function minimization. Organ boundary statistics are estimated from the segmentation and propagated to subsequent images; regional statistics are estimated from the simple brush strokes that are either propagated or redrawn as needed on subsequent images. This greatly reduces the user input needed and speeds up segmentations. The proposed method can be further accelerated with graph-based interpolation of alternating slices in place of user-guided segmentation. CT images from phantom and patients were used to evaluate this method. The authors determined the sensitivity and specificity of organ segmentations using physician-drawn contours as ground truth, as well as the predicted-to-ground truth surface distances. Finally, three physicians evaluated the contours for subjective acceptability. Interobserver and intraobserver analysis was also performed and Bland-Altman plots were used to evaluate agreement. RESULTS Liver and kidney segmentations in patient volumetric CT images show that boundary samples provided on a single CT slice can be reused through the entire 3D stack of images to obtain accurate segmentation. In liver, our method has better sensitivity and specificity (0.925 and 0.995) than region growing (0.897 and 0.995) and level set methods (0.912 and 0.985) as well as shorter mean predicted-to-ground truth distance (2.13 mm) compared to regional growing (4.58 mm) and level set methods (8.55 mm and 4.74 mm). Similar results are observed in kidney segmentation. Physician evaluation of ten liver cases showed that 83% of contours did not need any modification, while 6% of contours needed modifications as assessed by two or more evaluators. In interobserver and intraobserver analysis, Bland-Altman plots showed our method to have better repeatability than the manual method while the delineation time was 15% faster on average. CONCLUSIONS Our method achieves high accuracy in liver and kidney segmentation and considerably reduces the time and labor required for contour delineation. Since it extracts purely statistical information from the samples interactively specified by expert users, the method avoids heuristic assumptions commonly used by other methods. In addition, the method can be expanded to 3D directly without modification because the underlying graphical framework and graph partition optimization method fit naturally with the image grid structure.

Evaluation of tumor localization in respiration motion-corrected cone-beam CT: Prospective study in lung

PURPOSE Target localization accuracy of cone-beam CT (CBCT) images used in radiation treatment of respiratory disease sites is affected by motion artifacts (blurring and streaking). The authors have previously reported on a method of respiratory motion correction in thoracic CBCT at end expiration (EE). The previous retrospective study was limited to examination of reducing motion artifacts in a small number of patient cases. They report here on a prospective study in a larger group of lung cancer patients to evaluate respiratory motion-corrected (RMC)-CBCT ability to improve lung tumor localization accuracy and reduce motion artifacts in Linac-mounted CBCT images. A second study goal examines whether the motion correction derived from a respiration-correlated CT (RCCT) at simulation yields similar tumor localization accuracy at treatment. METHODS In an IRB-approved study, 19 lung cancer patients (22 tumors) received a RCCT at simulation, and on one treatment day received a RCCT, a respiratory-gated CBCT at end expiration, and a 1-min CBCT. A respiration monitor of abdominal displacement was used during all scans. In addition to a CBCT reconstruction without motion correction, the motion correction method was applied to the same 1-min scan. Projection images were sorted into ten bins based on abdominal displacement, and each bin was reconstructed to produce ten intermediate CBCT images. Each intermediate CBCT was deformed to the end expiration state using a motion model derived from RCCT. The deformed intermediate CBCT images were then added to produce a final RMC-CBCT. In order to evaluate the second study goal, the CBCT was corrected in two ways, one using a model derived from the RCCT at simulation [RMC-CBCT(sim)], the other from the RCCT at treatment [RMC-CBCT(tx)]. Image evaluation compared uncorrected CBCT, RMC-CBCT(sim), and RMC-CBCT(tx). The gated CBCT at end expiration served as the criterion standard for comparison. Using automatic rigid image registration, each CBCT was registered twice to the gated CBCT, first aligned to spine, second to tumor in lung. Localization discrepancy was defined as the difference between tumor and spine registration. Agreement in tumor localization with the gated CBCT was further evaluated by calculating a normalized cross correlation (NCC) of pixel intensities within a volume-of-interest enclosing the tumor in lung. RESULTS Tumor localization discrepancy was reduced with RMC-CBCT(tx) in 17 out of 22 cases relative to no correction. If one considers cases in which tumor motion is 5 mm or more in the RCCT, tumor localization discrepancy is reduced with RMC-CBCT(tx) in 14 out of 17 cases (p = 0.04), and with RMC-CBCT(sim) in 13 out of 17 cases (p = 0.05). Differences in localization discrepancy between correction models [RMC-CBCT(sim) vs RMC-CBCT(tx)] were less than 2 mm. In 21 out of 22 cases, improvement in NCC was higher with RMC-CBCT(tx) relative to no correction (p < 0.0001). Differences in NCC between RMC-CBCT(sim) and RMC-CBCT(tx) were small. CONCLUSIONS Motion-corrected CBCT improves lung tumor localization accuracy and reduces motion artifacts in nearly all cases. Motion correction at end expiration using RCCT acquired at simulation yields similar results to that using a RCCT on the treatment day (2-3 weeks after simulation).