Brian Thorndyke

Model‐based image reconstruction for four‐dimensional PET

Positron emission tonography (PET) is useful in diagnosis and radiation treatment planning for a variety of cancers. For patients with cancers in thoracic or upper abdominal region, the respiratory motion produces large distortions in the tumor shape and size, affecting the accuracy in both diagnosis and treatment. Four-dimensional (4D) (gated) PET aims to reduce the motion artifacts and to provide accurate measurement of the tumor volume and the tracer concentration. A major issue in 4D PET is the lack of statistics. Since the collected photons are divided into several frames in the 4D PET scan, the quality of each reconstructed frame degrades as the number of frames increases. The increased noise in each frame heavily degrades the quantitative accuracy of the PET imaging. In this work, we propose a method to enhance the performance of 4D PET by developing a new technique of 4D PET reconstruction with incorporation of an organ motion model derived from 4D-CT images. The method is based on the well-known maximum-likelihood expectation-maximization (ML-EM) algorithm. During the processes of forward- and backward-projection in the ML-EM iterations, all projection data acquired at different phases are combined together to update the emission map with the aid of deformable model, the statistics is therefore greatly improved. The proposed algorithm was first evaluated with computer simulations using a mathematical dynamic phantom. Experiment with a moving physical phantom was then carried out to demonstrate the accuracy of the proposed method and the increase of signal-to-noise ratio over three-dimensional PET. Finally, the 4D PET reconstruction was applied to a patient case.

Radiation dose reduction in four-dimensional computed tomography.

Four-dimensional (4D) CT is useful in many clinical situations, where detailed abdominal and thoracic imaging is needed over the course of the respiratory cycle. However, it usually delivers a larger radiation dose than the standard three-dimensional (3D) CT, since multiple scans at each couch position are required in order to provide the temporal information. Our purpose in this work is to develop a method to perform 4D CT scans at relatively low current, hence reducing the radiation exposure of the patients. To deal with the increased statistical noise caused by the low current, we proposed a novel 4D penalized weighted least square (4D-PWLS) smoothing method, which can incorporate both spatial and phase information. The 4D images at different phases were registered to the same phase via a deformable model, thereby, a regularization term combining temporal and spatial neighbors can be designed for the 4D-PWLS objective function. The proposed method was tested with phantom experiments and a patient study, and superior noise suppression and resolution preservation were observed. A quantitative evaluation of the benefit of the proposed method to 4D radiotherapy and 4D PET/CT imaging are under investigation.

Impact of Integrated PET/CT on Variability of Target Volume Delineation in Rectal Cancer

Several studies have demonstrated substantial variability among individual radiation oncologists in defining target volumes using computed tomography (CT). The objective of this study was to determine the impact of combined positron emission tomography and computed tomography (PET/CT) on inter-observer variability of target volume delineation in rectal cancer. We also compared the relative concordance of two PET imaging tracers, 18F-fluorodeoxyglucose (FDG) and 18F-fluorodeoxythymidine (FLT), against conventional computed tomography (CT). Six consecutive patients with locally advanced rectal cancer were enrolled onto an institutional protocol involving preoperative chemoradiotherapy and correlative studies including FDG- and FLT-PET scans acquired in the treatment position. Using these image data sets, four radiation oncologists independently delineated primary and nodal gross tumor volumes (GTVp and GTVn) for a hypothetical boost treatment. Contours were first defined based on CT alone with observers blinded to the PET images, then based on combined PET/CT. An inter-observer similarity index (SI), ranging from a value of 0 for complete disagreement to 1 for complete agreement of contoured voxels, was calculated for each set of volumes. For primary gross tumor volume (GTVp), the difference in estimated SI between CT and FDG was modest (CT SI = 0.77 vs. FDG SI = 0.81), but statistically significant (p = 0.013). The SI difference between CT and FLT for GTVp was also slight (FLT SI = 0.80) and marginally non-significant (p < 0.082). For nodal gross tumor volume, (GTVn), SI was significantly lower for CT based volumes with an estimated SI of 0.22 compared to an estimated SI of 0.70 for FDG-PET/CT (p < 0.0001) and an estimated SI of 0.70 for FLT-PET/CT (p < 0.0001). Boost target volumes in rectal cancer based on combined PET/CT results in lower inter-observer variability compared with CT alone, particularly for nodal disease. The use of FDG and FLT did not appear to be different from this perspective.

Grade-point average, changes of major, and majors selected by students leaving engineering

Graduation success, grade-point average, and destination major of ten cohorts of students matriculating and subsequently leaving undergraduate engineering programs at nine southeastern universities are studied from 1987-2002. Grade point averages are frozen at the time students leave engineering to investigate the role of grades in their decision to leave engineering and their choice of a destination major. This study adds to evidence indicating that poor performance is not the primary reason students leave engineering. Students leaving with low grades most likely select business, students with high grades more likely choose natural science majors and, interestingly, 10 to 20% at all performance levels choose education or a social science. The study also found that 10 to 15% of the students leaving engineering at all performance levels changed majors at least a second time before graduating, suggesting that changing majors is, for some, a journey of exploration rather than a matter of settling for ones second choice.

TU‐D‐J‐6C‐08: Enhancing 4D PET Through Retrospective Stacking

Purpose: Four‐dimensional (4D) PET presents challenges distinct from 4D CT owing to radiotracer dose limitations. A single‐bed field‐of‐view (FOV) PET scan typically requires several minutes to acquire adequate data for reconstruction, necessarily spanning several respiratory cycles and smearing the radiotracer signal within a given lesion over an increased volume. Although prospective or retrospective gating captures the PETimage at a single point in the respiratory cycle, restricting the data to events within the gating interval increases the signal‐to‐noise ratio (SNR). We propose a method, coined “retrospective stacking” (RS), to combine the data from the entire respiratory cycle through deformable registration. In addition, we use the transformation maps to generate a 4D PET with statistics comparable to the single RS image.Method and Materials: A single FOV of a pancreatic cancer patient was acquired via the gated PET mode on a GE Discovery ST PET‐CT scanner. These gated images were registered using a mutual information / B‐splines registration algorithm, and superimposed. A 4D PET series spanning the full respiratory cycle was generated, and fused onto a 4D CT.Results: The SNR of the RS image showed an increase of 15% over a single gated reconstruction. Activity‐volume histograms of radiotracer activity surrounding the pancreatic lesion revealed that the ungated PET showed 33% greater tumor volume (using a 40%‐of‐maximum threshold) than the RS image. The reconstructed 4D PET fused well with the 4D CT, providing a clearer view of radiotracer distribution over the respiratory cycle than was possible using gated reconstructions.Conclusion: Retrospective stacking enabled better integration of temporally varying PET and CT series by reducing radiotracer smearing due to respiratory motion, while at the same time increasing the SNR beyond the poorer statistics inherent in gated PET acquisition. Noise‐reduced 4D PETimages could also be generated for fusion with 4D CT.1

TU-D-J-6C-02: Radiation Dose Reduction in 4D Computed Tomography

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 CTimage at a given phase was achieved by superposing the imaging information from other phases with the use of a deformable image registrationmodel. 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.

SU-DD-A4-04: Registration of Prostate MRI/MRSI and CT Studies Using the Narrow Band Approach

Purpose: High‐field MR techniques makes possible to obtain high quality MRI metabolic images of the prostate to accurately identify the intra‐prostatic lesion(s). However, the use of rigid endorectal probe deforms the shape of the prostate gland and the images so obtained are not directly usable in radiation therapy planning. This work applies a narrow band deformable registration model to faithfully map the MRI information onto treatment planningCTimages.Method and Materials: The narrow band is a hybrid method combining the advantages of pixel‐based and distance‐based registration techniques, since the calculation is restricted to those points contained in a region around user‐delineated structures. The narrow band method is inherently efficient because of the use of a priori information of the meaningful contour data. The deformable mapping is described by the B‐spline model. The limited memory algorithm (L‐BFGS) was implemented to optimize a normalized cross correlation metric function. Its convergence behavior was studied by comparing final metrics obtained in 100 registrations self‐registering an MR image starting from 100 randomly initiated positions. The spatial performance of the algorithm was assessed by intentionally distorting an MRIimage and an attempt was then made to register the distorted image with the original one. The MRI‐CT mapping was carried out for two clinical cases. Results: The convergence analysis showed absence of local minima. The technique can restore an MR image from the intentionally introduced deformations with an accuracy of ∼2 mm. On clinical cases the method was capable of producing clinically sensible mapping. The whole registration procedure for a complete 3D study took less than 15 minutes on a standard PC. Conclusion: Both hypothetical tests and patient studies have indicated that narrow‐band based registration is reliable and provides a valuable tool to integrate the ER‐based MRI/MRSI information to guide prostate radiation therapy treatment.

SU‐FF‐J‐02: A Comparison of Amplitude‐ and Phase‐Based 4D CT

Purpose: Four‐dimensional (4D) CT depends on accurate correlation between temporally acquired CT slices and the patients respiratory cycle. One approach is to record the position of an external marker placed on the abdomen or chest during the scan, and retrospectively match the CT data with the phase of the marker motion. While very effective for regular breathing patterns, the phase‐based approach can lead to significant mismatch between adjacent image segments when the respiratory motion exhibits irregularities. We propose a method of extracting amplitude‐based 4D CT from cine‐acquired CTdata sets, and compare the amplitude‐based 4D CT with the phase‐based 4D CT for both phantom and patient data. Method and Materials:CTdata sets were acquired in cine mode on the GE Discovery ST, and motion of an infrared reflecting block was recorded using Varians Real‐time Position Management (RPM) camera. Rather than use the phase‐based calculations of the RPM system, we replaced the phase field with pseudo‐amplitude values spanning the full respiratory cycle (i.e., differentiating inspiration from expiration). The modified respiratory trace file was then sent, along with the cine CT data, to the GE Advantage Workstation for processing. The method was applied to a thoracic phantom moving irregularly in the longitudinal direction, and to an abdominal 4D scan of a lungcancer patient. Results: For both phantom and patient data, the phase‐based 4D CTimages showed boundary mismatches of up to 1 cm between couch positions. The mismatch on the amplitude‐based sets, however, was less than 2 mm throughout the field of view. Conclusion: Phase‐based 4D CT can lead to mismatched slices when the respiratory cycle involves irregularities. In such situations, by replacing phase with a modified definition of amplitude that distinguishes inspiration from expiration, a substantially improved 4D CTimage can be generated.

SU‐EE‐A3‐02: The Potential of FDG‐PET in Delineating the Lumpectomy Cavity for Partial Breast Irradiation Patients

Purpose: To investigate the potential of FDG‐PET imaging for delineating the surgical cavity in post‐operative partial breast irradiation patients. Method and Materials: A DCIS breast cancer patient was imaged with a GE Discovery ST PET‐CT scanner approximately 2 weeks post lumpectomy. Following the treatment planningCT, a single‐bed (15 cm) FDG‐PET scan was dynamically acquired in 5‐sec intervals over 15 mins. The raw PET data was combined to form bins ranging from 30 sec to 15 min. These data were reconstructed by the GEscanner through an iterative OSEM algorithm, and hardware fused to the treatment planningCT. The value of PET in visualizing the lumpectomy cavity border was investigated through visual comparisons of fused PET‐CT images, the evolution of PET intensity for various breast points, and signal‐to‐noise measurements across the lesion. Results: The PETimage showed clear signal enhancement near the lumpectomy cavity. This enhancement formed a ring in each axial slice, matching the locations of surgical clips. Enhancement was also apparent where the cavity border was difficult to evaluate by CT density or clips alone. The ring presented a significantly higher SUV than other breast tissue (2.2 vs. <1.4), while the region inside the ring had a lower SUV than the (presumably benign) glandular tissue in the same breast (1.2 vs. 1.4). The SUV values were transient below 5 min, but remained stable thereafter. Conclusion: FDG‐PET may provide useful information for delineating lumpectomy cavity borders by elucidating regions of enhanced radiotracer uptake due to inflammation. The hypointense PET volume enclosed by the high activity ring may represent fluid and non‐viable tissue stemming from post‐surgical changes, explaining its lower FDG uptake and further supporting the suggestion that the ring corresponds to the cavity border itself. On current hardware, five‐minute scans were required to achieve stable SUVs.

SU‐FF‐T‐77: Accuracy of Gated IMRT Delivery On the Varian Linac Using the Real‐Time Position Management System

Purpose: To investigate the accuracy of gated IMRT delivery on a Varian linear accelerator equipped with the Realtime Position Management (RPM) camera and software. Method and Materials: A non‐uniform dose distribution within a solid water phantom was contoured and planned with IMRT. A sinusoidally oscillating platform simulated superioinferior respiratory motion, and a reflecting block was placed on the surface of the platform to provide a “respiratory” signal to the RPM camera. First, the phantom was stationary while the platform served only to provide the respiratory signal. The respiratory period was 5 sec, and the treatment was delivered in phase‐gated intervals of 6%, 10%, 25% and 50%. Second, the phantom was placed on the platform, with motion amplitude of 6 cm. Here, dose was delivered to the phantom during a small amplitude‐defined interval at end‐expiration, with periods 1.7 sec, 5.3 sec and 12.6 sec. Dose distributions were captured on film. Results: Dose profiles generally showed variation between configurations less than 2% the maximal dose, with shorter‐interval delivery providing slightly less dose than longer‐interval delivery. The only notable difference occurred for the phantom moving with respiratory period of 1.7 sec, where dose fluctuations of nearly 6% occurred at regions of high dose gradient in the direction of motion. It should be noted that the gating interval spanned 15% the respiratory cycle, implying the beam was delivered in only 1.7 × 0.15 = 0.25 sec intervals. Conclusion: Gated IMRT delivery provided dose distributions equivalent to ungated delivery to within clinically acceptable limits. This result held for significant motion amplitude, under a wide range of respiration frequencies and gating intervals. While discrepancies up to 6% arose at high gradient borders for configurations of extremely rapid motion and short beam‐on time, these parameters are very unlikely to be seen in any clinical situation.