Volume 34, Issue 6, June 2007
Index of content:
- General Poster Discussion: Exhibit Hall C
SU‐FF‐I‐01: 3D Computer‐Aided Detection of Masses and Micro‐Calcifications From Cone Beam CT Scans: A Breast Phantom Study34(2007); http://dx.doi.org/10.1118/1.2760377View Description Hide Description
Purpose: Cone Beam Breast Computed Tomography (CBBCT) has emerged as a promising new modality for detection and diagnosis of breast cancer. Compared to conventional X‐ray mammography, CBBCT provides much greater detail of breast tissue in three‐dimensions (3D) as well as improved patient comfort. However, these benefits come at the cost of hundreds of image slices for each scan. Here, a validation study was performed using CBBCT data of a breast phantom with known nodules and micro‐calcifications. A novel 3D computer‐aided detection algorithm based on 3D template‐matching was applied to this dataset to evaluate the performance of CAD‐CBBCT. Method and Materials: The CBBCT dataset of the breast phantom consisted of 371 slices with isotropic resolution of 0.27 mm. The central 250 contiguous slices were used in our analysis. The selected imaging volume contained 8 masses with diameter ranging from 0.5–10mm (appearing on 9–37 pixels), and 13 micro‐calcifications with diameter ranging from 0.2–0.5 mm (appearing on 3–5 pixels). Our CAD algorithm models masses and micro‐calcifications as spheres (templates) of varying sizes and searches for structures similar to the templates throughout the entire imaging volume. Templates of 9 different sizes were applied. The end result is a single index (the normalized cross correlation coefficient) for each mass and calcification candidate. An optimal correlation threshold for each template size was then selected to best identify the nodules and micro‐calcifications. Results: 8 out of 8 masses down to 0.5 mm and 12 out of 13 micro‐calcifications down to 0.2 mm were successfully detected, with only 1 false positive finding. The processing time for each template was ∼80 seconds (on a Mac Pro with 2.66GHz CPU, 5G memory). Conclusion: CBBCT and 3D CAD were able to detect in a breast imaging phantom very small nodules and micro‐calcifications with high sensitivity and specificity.
SU‐FF‐I‐02: Computer‐Aided Mass Detection On Digitized Mammograms Using Adaptive Thresholding and Fuzzy Entropy34(2007); http://dx.doi.org/10.1118/1.2760378View Description Hide Description
Purpose: A segmentation method for detection of masses in digitized mammograms has been developed using two parallel approaches: adaptive thresholding method and fuzzy entropy feature as a CAD scheme. Method and Materials: The algorithm consists of the following steps: a) Preprocessing of the digitized mammograms including identification of region of interest (ROI) as candidate for massive lesion through breast region extraction, b) Image enhancement using linear transformation and subtracting enhanced from the original image, c) Characterization of the ROI by extracting the fuzzy entropy feature, d) Local adaptive thresholding for segmentation of mass areas, e) Combine expert of the last two parallel approaches for mass detection. Results: The proposed method was tested on 78 mammograms (30 normal & 48 cancerous) from the BI‐RADS and local databases. The detected regions validated by comparing them with the radiologists' hand‐sketched boundaries of real masses. The current algorithm can achieve a sensitivity of 87% with 1.57 FP/image. Conclusions: This approach showed that the behavior of local adaptive thresholding and fuzzy entropy technique could be a useful method for mass detection on digitized mammograms. Our results suggest that the proposed method could help radiologists as a second reader in mammographic screening of masses.
Keywords: Breast Masses, Segmentation, Image Enhancement, local adaptive Thresholding, Fuzzy Entropy.
SU‐FF‐I‐03: Investigation of Similarity Measures for Selection of Similar Images for Breast Lesions On Mammograms34(2007); http://dx.doi.org/10.1118/1.2760379View Description Hide Description
Purpose: Similar known images may be helpful for radiologists in the diagnosis of breast lesions on mammograms. However, selected images may not be really similar if a similarity measure is not properly determined. We determined the radiologists' subjective similarity ratings for pairs of lesions and investigated the objective measures that would agree with the radiologists' ratings. Material and Methods: We selected 300 pairs of masses and 300 pairs of clustered microcalcifications for determination of subjective ratings to establish a “gold standard.” These ratings would be useful for determination and evaluation of objective similarity measures. The 300 pairs were randomly grouped into two groups; one group was used as a training set, whereas the other group was used as a test set. This process was repeated for three times. The objective measures based on the distance in the feature space and by use of an artificial neural network (ANN) were compared by a cross validation method. Results: For distance‐based objective measure, the correlations between the subjective ratings and objective measures were 0.50, 0.53, and 0.51 for the mass pairs. When the radiologists' ratings were used as teacher in training of the ANN, the similarity measures were improved, and the correlations were improved to 0.66, 0.70, and 0.67. Conclusion: The results indicate that similar images selected by the ANN‐based similarity measure may be more useful than the images selected by the distance‐based similarity measure in feature space.
34(2007); http://dx.doi.org/10.1118/1.2760380View Description Hide Description
A snake is a deformable curve used to localize region boundaries. Snake deformation is controlled by two terms: the internal and external energy fields. This study focuses on the generation of a hemithoracic cavity external field (HCEF) for the segmentation of hermithoracic cavities in CT scans. HCEF construction is a multi‐step process. First, the lung parenchyma is segmented using a combination of thresholding and shape descriptors and the trachea is delineated by region growing. Next, bone and contrast‐enhanced tissue are segmented by application of a gray‐level threshold. This threshold is chosen to exclude disease that demonstrates lower and more diffuse contrast uptake than mediastinal structures. A hemithoracic cavity bounding structure (HCBS) image is created by combining the trachea, lung parenchyma, and bone/mediastinum segmentations. Edge detection is then applied to the HCBS image, and a gradient vector field is calculated in the space external to the HCBS. Finally, a weighted distance transform is applied to the internal space of each HCBS to ensure that the internal energy field of the snake does not invade bounding structures. This method was applied to 20 CT sections and qualitatively evaluated.
In the presence of lung‐deforming disease, the interface of the lung parenchyma and soft tissue is not suitable for hemithoracic‐cavity segmentation. The HCEF and snake were applied to images of patients with severely deformed lung parenchyma due to mesothelioma. The snake showed excellent results localizing the lateral portion of the hemithoracic cavity and promising results along the mediastinum.
The HCEF was created for the segmentation of the hemithoracic cavities in the presence of lung deforming disease. In contrast to other external fields, the HCEF incorporates knowledge of specific HCBS identified by several independent segmentation steps. This method is designed to be robust in the presence of lung‐deforming disease and demonstrates promising qualitative segmentation results.
34(2007); http://dx.doi.org/10.1118/1.2760381View Description Hide Description
Purpose: In radiological diagnoses, it is important to understand and interpret anatomical changes, specifically those observed in CT scans, which correlate with treatment plans. There are currently many anticancer agents, which affect tumor vasculature growth as well as normal organ vessel growth. Monitoring these effects with imaging may be useful to guide selection and dosing of different treatments. By observing perfusion of certain organs, it is possible to monitor vessel competency. In cases where perfusion is constant over time, it can be assumed that there is no significant change in vasculature, and thus no damage as an effect of chemotherapeutic agents. In contrast, significant changes in perfusion over time, may suggest decease in normal vessel growth as a result of therapy. Methods & Materials: Patients receiving the VEGF inhibitor sorafenib, underwent CTimaging every six weeks, beginning with a baseline study prior to treatment. A “jog scan” was used to track perfusion through the adrenal glands (chosen due to their significant fenestration). Sixteen pairs of adrenal images were obtained per jog scan, and manually contoured using a contouring program. Each of the sixteen scans represents different perfusion time intervals from 0–150 seconds. The mean pixel values of each gland were obtained, and these values were compared over time for any significant changes in pixel value, and thus change in vasculature perfusion over time. Results: The average change in maximum pixel values from baseline to six weeks after treatment initiation shows a change of 4.58% increase in peak pixel value for both adrenal glands. Conclusion: The manual contouring of adrenal glands in conjunction with calculated maximum pixels values shows changes in adrenal perfusion between baseline and beginning therapy. The continued monitoring of perfusion could prove beneficial to the radiologic diagnosis of significant anatomical changes as a result of continuous chemotherapy.
34(2007); http://dx.doi.org/10.1118/1.2760382View Description Hide Description
Purpose: As the utilization of cardiac CT grows, the development of the testing tools to assess the image quality parameters that are associated with cardiac CT protocols is lagging behind. This paper presents a motion system that was designed to provide the moving imaging targets by rotating the CT phantom in order to simulate the heart beat; thus evaluate the performance of a cardiac CT system. Methods and Materials: To assess the image quality of a ultra‐fast/thin slice VCT (GE LightSpeed), we designed a motion device that was capable of rotating CT phantoms at a controlled rotation speed ranging from 0 to 30 rpm. The apparatus consisted of radiation translucent support plates, gear motor, speed controller, rpm sensor, gears and shafts. This motion phantom system was designed to assess the additional image quality parameters associated with cardiac CT protocols such as the motion artifact occurrence, the loss of spatial resolution due to motion, temporal resolution and isotropic resolution during gated and non‐gated protocols. Results: The motion phantom system was designed and made in‐house. The flexibility of the system allowed its usage with multiple existing CT phantoms including the ACR CT phantom, CATPHAN, CIRS Helical CT phantom and AAPM CT performance phantom. By scanning the phantom in stationary and in various rotation speeds, we were able to examine the performance parameters of the VCT scanner and optimize the clinical protocols with regards to slice thickness selections, detector row combinations, pitch and other technique settings. Conclusions: The motion phantom system is an essential tool to evaluate the image quality and verify the manufacturer specifications on temporal resolution and isotropic resolution which should reveal the true capability of a volume CT scanner. This motion system is also useful in optimizing the clinical protocols.
SU‐FF‐I‐07: Single‐ and Dual‐Energy CT Calibration Lines for Assessing the Calcium Content of Lung Nodules: Effects of Patient Body and Lung Nodule Size34(2007); http://dx.doi.org/10.1118/1.2760383View Description Hide Description
Purpose: To determine the concentration of Ca in lung nodules for a CAD technique, lung nodule calibration lines are being derived at locations throughout lung fields. A study was performed to investigate the effects of patient body and lung nodule size on derived calibration lines. Method and Materials: Simulated spherical lung nodules of two concentrations (50 and 100mg/cc ) were employed. Three different diameter nodules (4.8mm, 9.5mm, 16mm) were scanned in a simulated thorax section “A” representing the middle of the chest with large lung regions. The 4.8mm and 9.5mm nodules were also scanned in section “B” representing the upper chest with smaller lung regions. Fat‐rings were added to the phantoms to simulate larger patients. Images were acquired on a GE‐VCT scanner at 80, 120 and 140kVp. The RMS CT♯ displacements between the calibration lines for phantoms with and without fat‐rings were compared. Results: Body‐size had a significant effect on the calibration lines for each single kVp technique. Mean RMS displacements for the 9.5mm nodules at 80, 120 and 140 kVp were 22+/−2, 19+/−3, and 18+/−2HU, respectively for phantom “A”, and 19+/−2, 14+/−1, and 14+/−1HU for “B”. Corresponding displacements for 80kVp–140kVp dual‐energy were much less: 5+/–1HU (“A”) and 6+/−2HU (“B”). Results similar to the 9.5mm were obtained for the 4.8 and 16mm nodules in phantom “A” However, in phantom “B”, the 4.8mm dual‐energy displacements (12+/−4 HU) were about as large as the single‐energy. The phantom “B” study was repeated, and the dual‐energy displacements for the 4.8mm nodules were slightly better (7+/− 4 HU) on one lung side but about as poor (10+/−7HU) on the other. Conclusion: Dual‐energy CTcalibration of the calcium concentration of lung nodules is less sensitive to patient body size than single‐energy calibration. However, the dual‐energy approach may not compensate for patient body size for smaller nodules.
34(2007); http://dx.doi.org/10.1118/1.2760384View Description Hide Description
Introduction:CT generates high resolution anatomical images. However, scanning moving targets, like lungtumors, may create geometrical deformations in shape, volume and position. The objective of this work was to study the geometrical changes produced on CTimages scanning a known moving target. Material and Methods: A motorized phantom capable of variable oscillatory movement with a platform to attach different radio opaque volumes was constructed (prism, cube, pyramid and sphere). A CTGELight Speed was used to scan the volume during oscillatory movement. Axial and helical images were acquired with different CT parameters, X‐ray tube rotation speed and couch speed. Then, images were acquired with a fixed tube speed and different frequency and amplitude of phantom movement. Analysis of target deformation was done on a Varian Eclipse v.7.3 TPS. Center of mass (CM) displacement, volume deformation and area versus position were compared between static and moving targets. Results: For fixed oscillation parameter with different X‐ray tube speed, the average and maximum CM displacement were 6 and 9mm for axial and 5 and 9mm for helical scans and volume deformation were 6% and 18% for axial and 1% and 31% for helical scans. For fixed X‐ray tube speed and variable oscillatory parameter target movement the average and maximum CM displacement were 6mm, 6mm, 3mm and 2mm respectively. Average and maximum deformation for prism, cube, pyramid and sphere were 8.3% – 44%, 0.4% – 10%, 5.8% – 9.8% and 1.6% – 6.6% respectively. A mathematical predictive model can be established for set conditions. Conclusions: Displacement and deformity varies independently with X‐ray tube rotation speed and oscillatory movement. Movement may produce inaccuracy for PTV definition.
34(2007); http://dx.doi.org/10.1118/1.2760385View Description Hide Description
Purpose: As the clinical utilization of CT grows, it becomes more important to manage patient dose without compromising image quality, especially in children. A special effort should be made to reduce pediatric patient dose through age‐ and size‐specific protocols. Methods & Materials: A pediatric 64 slice VCT scanner(GE Lightspeed) was tested with a group of cylindrical acrylic phantoms with diameters ranging from 6 – 32 cm. In addition, anthropomorphic phantoms (CIRS adult and pediatric dosimetry verification phantoms) were employed to correlate the CTDI values with the skin doses measured by a solid state dosimeter (Unfors PSD). The dose affecting factors included: kVp, mAs, beam filtration, beam collimation, pitch, patient size, detector configuration and dose reduction techniques such as mA modulation and post‐processing. Various techniques and their combinations were included in this study. Finally, clinical protocols for pediatric applications were evaluated and adjusted based upon the measured patient dose and image quality. Results: The automated CTDI values displayed on the system agreed with our measurements when the standard phantom sizes were used, i.e., 16 cm in diameter for head, 32 cm for adult body and 16 cm for pediatric body. However, the measured dose differed from the automated CTDI by a factor of 1.72 for a reduced head phantom size of 6 cm in diameter and a factor of 3.2 for a reduced body phantom size of 10 cm in diameter. Patient age also played an important role in estimating effective dose. The changing beam filtration caused a variation of up to 42% in the in‐air dose output. Concurrently, noise from the phantom images was evaluated. Conclusion: The clinical protocols were established based upon the dose level corresponding to the patient size and age as well as the tolerable noise level corresponding to the specific clinical applications.
SU‐FF‐I‐10: A New Method to Perform CT Gantry Tilt Angle Quality Control Using Commercially Available Phantom34(2007); http://dx.doi.org/10.1118/1.2760386View Description Hide Description
Purpose: To develop a filmless method to perform CT gantry tilt angle quality control test by using commercialy available phantom to adapt the modern digital radiological environment. Method and Materials: A Catphan® 500 CT performance phantom (The Phantom Laboratory, Inc.) was set up on the CT patient couch according to manufacturer's instructions with the CTP401 module of the phantom centered in the scan plane. The CTP401 module contains four ramps (two vertical and two horizontal) which are angled ±23° from the scan plane when gantry tilt is 0°. Axial scans through the ramps were acquired over a range of prescribed slice thicknesses to determine slice sensitivity profile width, or true slice thickness (Tslice), by averaging the slice sensitivity profile width derived from all four ramps. Following this, gantry tilt (θ) was measured over the entire range of tilt (−30° to +30°) by scanning the CTP401 module through the vertical ramp of the same sign (−23° or +23°) and measuring the FWHM length (LFWHM,θ) of the ramp image. Then, gantry tilt (θ) was calculated as for gantry tilts less than 23°, and as for gantry tilts greater than 23°. Calculated gantry angles were validated by comparison to gantry angles determined using the conventional film measurement method described in AAPM Report No. 39. Measurements were obtained at nine gantry tilt angles on 16— slice CT systems. Results: Different slice thicknesses obtained at 9 different angles will be presented. On average, gantry tilt measured using the Catphan agreed with the conventional film measurement method within 1°. Conclusion: This new method removes the requirement to use film for gantry tilt measurements. It is easy to calculate and can be performed without purchasing additional phantoms or equipment. It could be applied to clinical CT gantry angle QC in either on diagnostic or simulation CT.
34(2007); http://dx.doi.org/10.1118/1.2760387View Description Hide Description
Purpose: To determine the HVL of CT systems from CTDIdosimetry measurements without additional test, and to overcome the technical difficulties in CT HVL measurements. Method and Materials: The HVL of one multi‐detector CTscanner was measured when the scanner is static, over the available energy range (80∼140 kVp) and with different bow‐tie filter configurations. Commercial CT dose phantoms in compliance with FDA specifications were used to measure the exposures. Nine doses at the central, peripheral, and intermediate positions were measured for both head and body phantoms. Averages were taken for doses measured at axially symmetric positions of each phantom. These averages were compared to the central position dose measurements. The ratios were analyzed along with the HVL values measured under exactly the same conditions and fitted to second power polynomials. Combining the fitting parameters and dosimetry measurements from different machines, the corresponding HVL were predicted and compared to the measured value of each machine. Results: The HVL of the CTscanner were determined to be between 5 to 10 mm Al over the energy range assessed. When the values were plotted against the dose at the peripheral or intermediate position, normalized by center dose, the curves all fit to second power polynomials with R2>0.99. When applying the correlation obtained from the fittings to dosimetry measurements from different machines, the calculated HVLs are within 5% or 0.5 mm Al error of the measured values. Conclusion: This study showed that the data collected for the CTDI measurements can be directly used to estimate the HVL of CTscanners to certain accuracy. This will reduce the time and effort needed for medical physicists to determine the HVL of CT systems during acceptance testing and annual quality assurance surveys.
34(2007); http://dx.doi.org/10.1118/1.2760388View Description Hide Description
Purpose: 4D imaging is becoming increasingly available for treatment planning. Generally, the information are used in (1) determining the range of tumor motion so that the tumor margin can be determined on a patient specific basis (3.5D‐RT); and (2) helping to choose a gating phase and window for respiration‐gated radiotherapy planning (gated‐RT). The purpose of this work is to assess the clinical impact of these treatment strategies by retrospectively studying 12 lungcancer patients. Methods and Materials: 12 lung patients who had underwent 4D‐CT and received gated‐RT were selected. For each of these patients, two additional treatment plans were generated, including 3.5D‐RT plan and conventional 3D‐RT plan (in which a population based margin of 2.5cm is used). The V20, V40, TCP and NTCP are evaluated for the three plans. Additionally, the potential of dose escalation to the target is evaluated for 3.5‐RT and gated‐RT by keeping the lung toxicity at the level of 3D‐RT. Results: 9 out of 12 patients have tumor motion range less than 2.5cm. For these patients, 4D‐CT derived patient specific margin leads to a reduction of tumor margin in 3.5D‐RT plan and thus significantly reduces the ipsilateral lung toxicity. For the other 4 patients, the motion is comparable or greater than 2.5cm and the use of 4D imaging makes it possible to avoid potential underdosing in the peripheral region of the tumor. In all cases, the gated RT plans lead to significantly reductions of V20, V40, and NTCP. A dose escalation in the range of 3∼15% is feasible for all these patients when gated‐RT is employed while keeping the lung toxicity at the level of 3D‐RT. Conclusions: The use of 4D‐CT individualize the definition of tumor margin in 3.5D‐ and gated‐RTs. With reduced margin size, a clinically significant escalation of dose becomes easily achievable.
34(2007); http://dx.doi.org/10.1118/1.2760389View Description Hide Description
Purpose: We developed a cone‐beam CT phantom for bone, metal, and beam hardening artifact evaluation. The two measures that can be tested are resolution and soft tissue discrimination. We will use this phantom to investigate beam filtrations for artifact reduction, resolution, and soft tissue discrimination. Method and Materials: The cylindrical PMMA phantom designed for cone‐beam CT (radius 12 cm, height 19 cm) is non‐uniform in the z‐direction. The phantom incorporates several different structures to test for bone and metal artifacts, as well as small object discrimination. A human skull fragment placed at the center of the cylinder tests for bone artifacts. Metal wires were placed horizontally and vertically through the cylinder. BBs were inserted on‐top of the skull fragment which were used for automatic object alignment. Both of these metal components were also used to generate metal artifacts. The phantom includes a soft tissue discrimination block with five different densities. Two additional blocks are used to test for small object discrimination in the horizontal and vertical directions. 24 BBs were inserted into the rims of the PMMA cylinder (top and bottom of phantom) and were used for the geometric reconstruction of the image acquisition trajectory. Results: The phantom was scanned on a bench‐top cone‐beam CT system with three different k‐edge filters, Er (0.127 mm), Er (0.254 mm), and Yb (0.254 mm). The resulting artifacts in the reconstructed images were assessed qualitatively. The data were reconstructed, and artifacts from the bone, wires, and BBs were observed to vary depending on the specific beam filtration used. Conclusion: We have developed a cone beam CT phantom for artifact evaluation and imaging technique optimization. The potential imaging techniques that can benefit are breast, brain, and extremity cone beam CT. Future work: develop a task specific figure of merit for artifact evaluation.
SU‐FF‐I‐14: Region‐Of‐Interest (ROI) Cone‐Beam Computed Tomography (CBCT) Using Rotational Digital Angiography (DA) Acquisitions34(2007); http://dx.doi.org/10.1118/1.2760390View Description Hide Description
Purpose: Region‐of‐interest (ROI) cone beam computed tomography(CBCT) promises significant integral dose reduction to patients during acquisition; however, reconstruction using these data result in truncation artifacts both inside and outside the ROI. We propose a new technique to equalize the intensity in the region outside the ROI to that inside to achieve reconstructions comparable to full‐field‐of‐view (FFOV) CBCT.Method and Materials: A ROI filter comprised of a 2.25 cm diameter central aperture in 0.21 g/cm2gadolinium screens was installed on the x‐ray‐tube assembly of a standard C‐arm gantry. Standard (unsubtracted) Rotational Digital Angiography (DA) acquisitions were performed of a head phantom. FFOV images without the filter in place were also obtained. The location of the ROI in the projection images was identified using edge detection and template matching. Intensities outside the ROI were equalized to those inside the ROI by mapping the intensities at the same percentiles in the cumulative histograms of these regions for every image. The equalized images were reconstructed using the 3D reconstructionsoftware of the acquisition apparatus. Results: Inside the ROI, the reconstruction data are highly correlated to the FFOV data (R2 = 0.87). Outside the ROI, the reconstruction while noisier due to fewer photons is comparable to the FFOV, but some artifacts due to the incomplete equalization and detection of the ROI edges, are visible. Further refinement methods are being investigated. Integral Dose Reduction was estimated to be 80% compared to the FFOV 12″ acquisition. Conclusion: DA‐ROI‐CBCT promises to be a feasible clinical technique for non‐subtracted applications such as those in cardiology and orthopedics allowing substantial integral dose reduction to the patient, while providing similar reconstructions to standard FFOV acquisitions.
Support: NIH Grants R01‐NS43924, R01‐EB002873, Sterbutzel Funds, Toshiba Medical Systems Corporation.
34(2007); http://dx.doi.org/10.1118/1.2760392View Description Hide Description
Purpose: In radiation therapy, it is not uncommon that one is interested only in a region of interest (ROI) of the patient and that the center of the ROI does not coincide with the rotation center. Recent development of CTimaging theory allows the design of innovative approaches to imaging ROIs. The benefits of ROI imaging include the reduction of imaging dose to the patient and of scatter and other artifacts. In this work, we propose a new ROI imaging approach and investigate its potential application to image‐guidedradiation therapy.Method: Scanning approaches have been proposed for imaging such ROIs through Normal varying dynamically illuminationcollimation of imaging x‐ray beam. In this study, we investigate and develop new imaging approach that uses a constant illuminationcollimation while using an effective, general trajectory through shifting detector and source in such a way that the x‐ray illumination always covers the ROI. From such data that contain truncations, the backprojection‐filtration (BPF) algorithm can be used for reconstructing the ROI images.Results: We have performed numerical studies to validate and evaluate this imaging approach. In these studies, we have used digital phantoms to generate data with and without noise. The results of our preliminary numerical studies indicate that ROI imaging without varying the source collimation during data acquisition can be achieved. Conclusion: ROI imaging without varying the source collimation during data acquisition can be achieved through varying the detector and source positions. The implication of the work to image‐guidedradiation therapy can be potentially high.
34(2007); http://dx.doi.org/10.1118/1.2760393View Description Hide Description
Purpose: In spite of over twenty years of computerized tomography(CT) research since the well‐known Feldkamp‐Davis‐Kress (FDK) method was first derived for three‐dimensional Cone‐Beam Computerized Tomographic (CBCT)reconstruction, there is a noticeable lack of practical software implementations available. Medical physicsresearchers needing CBCTreconstructions to prototype more advanced imaging techniques generally need to code the FDK method from scratch or adapt third‐party code that may be sophisticated and inflexible. To address this gap in free software tools, the AAPM Imaging Research Subcommittee has supported the development of OSCaR, a simple‐yet‐flexible open‐source Matlab FDK tool for algorithm development. Method and Materials: OSCaR includes open source, executable, and GUI software (Matlab; The MathWorks, Natick MA) for CBCTreconstructions from 2D projections. As a pre‐processing stage, projection data are parsed from a standard data‐file. Upon specification of a Field‐Of‐View (FOV), voxel size, and reconstruction filter, the 3D sinogram is filtered and back‐projected to produce a 3D reconstruction. The final reconstruction can be exported to various data formats as specified by the user. Results: OSCaR accepts data in a variety of formats accessible to Matlab. A circular source‐detector geometry is assumed, but OSCaR allows specification of the piercing point as a function of the projection angle. The aperture can be freely selected, as can the voxel size and the reconstruction filter. Visualization in 3D and in 2D (e.g., slices) is supported. Conclusion: OSCaR demonstrates flexibility, ease of use, and support of a broad range of input data formats. Upon completion of beta testing, the code will be freely available via the AAPM web‐site to AAPM members. The software is intended for algorithm development and research purposes rather than for clinical or commercial use. The software provides a reference‐able base of code to accelerate new imaging research in CBCT and facilitate multi‐institutional collaboration.
SU‐FF‐I‐17: An Evaluation On the Influence of MU Number and Field Length On MV‐CBCT Image Quality On a Siemens Oncor Linac Machine34(2007); http://dx.doi.org/10.1118/1.2760394View Description Hide Description
Purpose: We present an empirical evaluation of several factors influencing image quality in megavoltage cone beam CT (MV‐CBCT) to provide guidance for optimization of these parameters in a clinical setting. Method and Materials: A commercial system capable of MV‐CBCT imaging was recently installed in our clinic (MVision, Siemens Oncology Care Solutions). The system uses a 41 cm × 41 cm electronic portal imaging device specially optimized for 6 MV cone beam acquisition. Images are reconstructed from a 200 degree rotation of the gantry, resulting in a field of view of approximately 27×27×27 cm3. In preparation for routine use in image guided radiotherapy, we investigated the effect of varying the MU number and scan length on soft tissue contrast and system resolution. MV‐CBCT images were acquired of several phantoms using the Y‐jaws to vary the scan length, from 1.0 to 27.4 cm while keeping the X‐aperture fixed at 27.4 cm. Imaging was repeated for three MU setting of 5, 10, and 20. The signal to noise ratio (SNR) and the contrast to noise ratio (CNR) were determined in a low‐contrast CT phantom. A second phantom containing a series of line pair objects was used to assess spatial resolution. The image quality in terms of SNR,CNR, and spatial resolution were scored with respect to the MU and field length. Results: The SNR distribution exhibited a clear dependence on field length and MU number used in the acquisition. SNR improved significantly with increasing MU and with decreasing field length. CNR showed a similar dependence on both parameters, and was optimal at field lengths between 2 and 5 cm. Spatial resolution was independent of both parameters providing MU>5. Conclusion: With 20 MU and 2∼3 cm field length, MV‐CBCT can produce very good soft tissue contrast providing the possibility of MV‐CBCT based target delineation.
SU‐FF‐I‐18: Optimization of FDK Reconstruction Parameters to Minimize Aliasing and Reduce Metal Artifacts34(2007); http://dx.doi.org/10.1118/1.2760395View Description Hide Description
Purpose: To maximize SNR, suppress metal artifacts and minimize backprojection times for FDK‐based cone‐beam CT(CBCT)reconstructions by optimizing binning, filtering and backprojection parameters, and to investigate the performance of new multi‐core CPU's. Methods and Materials:CBCTreconstruction times can be reduced by filtering and downsampling high resolution flat panel projection data to match the reconstruction matrix pitch, and by using nearest neighbor interpolation (NN) for backprojection. However, metal artifacts and noise aliasing may result. To investigate the tradeoffs involved, a frequency‐domain noise power spectrum (NPS) model was developed. Phantom and clinical CBCT data were acquired using the On‐Board Imager and reconstructed with a range of pre‐processing and backprojection parameters while keeping imageMTF constant. Imagenoise, including the amount of noise aliasing, and metal artifacts were evaluated. Reconstruction times were measured on Xeon workstations comprising either two single‐core, dual‐core, or quad‐core CPU's. Results: Two types of metal artifacts emerged. A moiré pattern is produced if insufficient projection data densities in the transaxial direction are maintained, while radial streaks are produced by insufficiently dense axial data. For best image quality aliased noise should be less than 5% of the total imagenoise, and projection data should be 1.5–2.0× denser than the reconstructed image matrix pitch for backprojection with bilinear interpolation. NN interpolation is not preferred. Although backprojection times increased by ∼50% with these higher projection data densities, overall reconstruction times for relatively large image matrices (512×512×188, 675 projections), were <50 seconds using the quad‐core workstation which is sufficiently fast for IGRT applications. Conclusions: Optimal use of high data densities coupled with bilinear interpolation for backprojection can suppress some metal artifacts and minimize noise aliasing. New multi‐core CPU architectures provide sufficient speed to make such reconstructions clinically practical. Conflict of Interest: Employees of Varian Medical Systems.
34(2007); http://dx.doi.org/10.1118/1.2760396View Description Hide Description
Purpose: To develop an efficient analytical scatter correction algorithm for the On‐Board Imager (OBI) for both the center‐detector and offset‐detector geometries used for cone‐beam computed tomography(CBCT). The offset‐detector geometry is used for larger transaxial field‐of‐views and is particularly challenging due to the asymmetric nature of the associated scatter profile and higher overall scatter‐to‐primary ratios. Methods and Materials: A scatter kernel model was implemented. The cone‐beam was modeled as an array of pencil beams. For each of the pencil beams, a scatter point‐spread function was determined based on measured attenuation values and prior simulations of a polychromatic x‐ray beam directed through uniform material. The total scatter estimate was then derived from the cumulative response of each of the scatter point‐spread functions. The model also included the responses of the detector and anti‐scatter grid. To test the model, a pelvis phantom and a cylindrical water phantom were imaged on a table‐top system with hardware and geometric parameters that matched the OBI in offset‐detector configuration. The accuracies of the estimates were determined by comparisons with experimental scatter measurements. Results: Accuracies of the estimates were excellent away from the edges of the phantoms. Hounsfield Unit (HU) errors in the reconstructed images were reduced from over 20% pre‐correction to <2% after correction in the bulk of the image. However, near the edges of the phantoms, scatter was underestimated which resulted in residual HU errors that were on the order of 10%. Conclusions: The results demonstrate the potential for successful implementation of a computationally efficient scatter‐kernel model for the OBI. Work is underway to improve scatter estimates for rays near the edges of objects. Conflict of Interest: Employees of Varian Medical Systems.
SU‐FF‐I‐20: Dose Calculation On Megavoltage Cone‐Beam CT Images Corrected for Cupping and Missing Data Artifacts34(2007); http://dx.doi.org/10.1118/1.2760397View Description Hide Description
Purpose: Megavoltage cone‐beam CT (MVCBCT) using a 6 MV treatment beam provides routine 3D images of patients that can be used for dose calculation. However, artifacts reduce the accuracy of dose calculation based on these images. The objective of this work is to develop correction methods for the artifacts caused by scatter, the polyenergetic beam and the limited field of view. Method and Materials: Monte‐Carlo simulations of the linac beam and flat panel imager were carried out to ascertain the causes of the cupping artifact. Scatter contribution to the image was characterized using a superposition of pencil kernels. The effects of depth hardening and off‐axis softening of the beam were also quantified. MVCBCT images with missing data artifacts were corrected with completion of the image with a kVCT image and a correction algorithm for the MVCBCT part of the composite image. The cupping artifact corrections were tested with dose calculations on an anthropomorphic head phantom, and missing data artifacts corrections were tested on a calibration phantom for kVCT. Dose calculations were also obtained with kVCT images and gamma index analysis was used to compare the dose distributions. Results: For dose calculation on an anthropomorphic head, the corrected images had 97% of the voxels within the criteria of [3%, 2mm], and 95% for the uncorrected images. Dose calculations on the composite corrected images of the calibration phantom matched for 94% of the voxels, with a criteria of [5%,2mm]. This percentage fell to 72% when the MVCBCT part of the composite image was uncorrected. Conclusion: Dose calculation on head‐and‐neck patients using MVCBCT images is accurate within [3%,2mm], and [5%,2mm] for regions that suffer from missing data artifacts. This level of accuracy should be sufficient to study dose variations during the course of a treatment. Conflict of Interest: Research supported by Siemens.