Volume 33, Issue 6, June 2006
Index of content:
- Imaging Moderated Poster Session: Exhibit Hall F
- Moderated Poster ‐ Area 4 (Imaging): Breast Imaging
SU‐DD‐A4‐01: Quantifying Skin Effects After Accelerated Partial Breast Irradiation Using Digital Infrared Imaging (DII): Preliminary Feasibility Data33(2006); http://dx.doi.org/10.1118/1.2240149View Description Hide Description
Purpose: Accelerated partial‐breast irradiation (APBI) is an emerging radiation technique that challenges standard whole breast irradiation. The larger fraction sizes used in these hypofractionated schedules may increase the risk of late normal tissueeffects. Identification of the causes of variability in radiation sensitivity and normal tissue reactions could have important implications for breast cancer therapy. For this, a quantitative method of estimating early and late skineffects is needed. DII is recording instantaneous skin temperatures that are directly correlated with the skinblood flow, a parameter known to be an indicator of skin reaction. Material and Methods: An infrared digital camera IRSnapShot® was used to image breast cancer patients treated with APBI to a total dose of 3850 cGy over 10 fractions. The plans consisted of multiple external non‐coplanar photons +/− electron beams. The patients were imaged in a controlled temperature room before and after each fraction. They were also be imaged at regular intervals during their follow‐up. Two sets of orthogonal DII images were taken. The images were then transferred into Matlab where a GUI is being developed for image registration, thresholding and data analysis. Results: Five patients were imaged as described, three treated with photons only, and two with a combination of photons and electrons. The increase in maximum skin temperature from the baseline (pre treatment) to treatment completion is on average 2 to 4 degrees and depends on the techniques used, higher for plans including electrons, as expected due to their way of depositing dose.Conclusions: DII generated skin temperature information is a promising quantitative tool to estimate early and late effects in irradiated breast cancer patients. Our goal is to generate an “Index of Radiosensitivity” based on the early pattern of change in skin temperature that will allow individualization of radiotherapeutic prescription.
33(2006); http://dx.doi.org/10.1118/1.2240150View Description Hide Description
Purpose: To design and develop a portable optical imager for early‐stage breast cancer diagnostics, providing great depth information, enhanced data acquisition rates, and minimal patient discomfort. Method and Materials: A unique measurement geometry of simultaneous multiple point source illumination was implemented in the design and development of the hand‐held based optical probe. Simultaneous multiple point detection was carried out using an intensified charge‐coupled camera (ICCD) that can be operated in the continuous wave and frequency domain measurement approaches. The hand‐held based imaging probe has been coupled to the ICCD detection system and the performance characteristics (in terms of measurement accuracy and precision) of the imager is characterized through initial phantom studies under homogeneous conditions. Results: Preliminary simulated studies using simultaneous multiple point illuminationmeasurement geometry over the universally used single point illumination geometry demonstrated an increase in the detected signal strength as well as total interrogated tissue volumes. An optimal number of source and detector fibers used to develop the probe head, minimized the dead volume and improved the data acquisition times. Conclusion: A novel fluorescence‐enhanced imagingsystem was developed using a hand‐held probe and an ICCD camera, enabling the flexible and rapid imaging of any given tissue volume. Further work involves phantom based experimental studies towards 3D optical imaging and tomographic analysis. The final goal is to translate the current laboratory‐based techniques into routine clinical use.
33(2006); http://dx.doi.org/10.1118/1.2240151View Description Hide Description
Purpose: Anti‐scatter grids have been commonly used to reduce the amount of scatter in mammography. However, using grids require increasing the radiation dose to the breast in order to have an acceptable exposure to the image receptor. We used Monte Carlo simulation to optimize liner grid design for mammographyimaging in a way to achieve best contrast improvement with lowest dose to the breast. Materials and Methods: We used Monte Carlo Simulation Code MCNP5 to determine the amount of Scatter to Primary Ratio (SPR) for different x‐ray tube peak voltage (kVp), breast thicknesses, and grid geometries. We used a Molybdenum target/Molybdenum filtered x‐ray spectra, materials and geometrical dimensions that closely mimic the clinical situation. We used a semicircular shaped breast phantom made of 50 % adipose and 50 % glandular tissue equivalent materials. The grid septa were made of lead and inter‐space was made of carbon fiber. Results: Our calculated SPR values agree within 5 % with previously published clinical data. We have obtained significant contrast improvement for low bucky factors. For a 5 cm thick breast equivalent phantom, we found an optimal septa height of 0.9mm, septa thickness of 12μm and an inter‐space thickness of 100μm gives an optimal combination of 0.2 SPR, a 2.43 bucky factor, and a 1.31 contrast improvement factor (8 % error). With this geometry, the maximum SPR was lowered from 0.58 without the grid to 0.2 with the grid. Conclusion: We have optimized the geometry of the linear grid and achieved very significant contrast improvement with low SPR while minimizing the bucky factor and hence the mean‐glandular dose to the breast.
33(2006); http://dx.doi.org/10.1118/1.2240152View Description Hide Description
Purpose: This study's aim was to develop an easily reproducible clinical protocol to predict the average glandular dose (AGD) delivered to patients during routine mammography screening. It incorporates an evaluation of patient specific features, including glandularity, to predict the clinically delivered dose for both the cranio‐caudal (CC) and the medio‐lateral oblique (MLO) views. Method and Materials: The development of a modified homogenous dosimetric breast tissue equivalent phantom series (BRTES‐MOD) based on anthropomorphic measurements of the screening mammography population is central in evaluating the patient's fibroglandular content. It has been constructed with reference to the breast tissue elemental composition tabulated in the International Commission on Radiation Units and Measurements ‐ Report 44, and simulates the compression and variable content of patient's tissuecharacteristics. This study calculates the average glandular dose using entrance skin exposure and dose conversion factors based on fibroglandular content, compressed breast thickness, volumetric and anatomical factors, mammographic unit parameters and modifiable parameters of the BRTES‐MOD phantom. Results: Dose conversion factors were successfully calculated from the patient's fibroglandular content, compressed thickness, unit parameters, and spectral half value layer. An anthropometric population study facilitated the derivation of clinically usable equations to determine patient whole breast area, estimate patient skin layer thickness, and assess optimal placement for the automatic exposure control ionization chamber location. Dose distributions for the study population are presented for both CC and MLO views and compare well with those derived from previous population studies. Conclusion: The designed protocol can be performed within the time of a typical mammography screening appointment, and allows the determination of patient‐specific average glandular dose. The BRTES‐MOD method also provides a quantitative measure of patient specific AGD for the multiple projections comprising screening mammography examinations.
SU‐DD‐A4‐05: Characterization of X‐Ray Scatter and Glandular Dose in Digital Tomosynthesis for Breast Imaging Using Monte Carlo Simulations33(2006); http://dx.doi.org/10.1118/1.2240153View Description Hide Description
Purpose: To study the characteristics of x‐ray scatter and glandular dose in digital tomosynthesis for breast imaging.Method and Materials:Monte Carlo simulations of x‐ray transport in breast tomosynthesis were performed using the Geant4 package [Agostinelli et al, Nucl Instrum Meth A 506: 250–303, 2003]. Scatter‐to‐primary ratio (SPR) maps, maximum SPR, scatter point spread functions (PSF) and glandular dose to the breast were computed at several projection angles while varying compressed breast size, thickness, glandularity and x‐ray spectrum. For validation, the SPR and scatter PSF for the planar mammography view (0 degrees) for various setups were compared with published values [Boone et al, Med Phys 27(10): 2408‐16, 2000 and 27(8): 1818–1831, 2000]. Results: SPR maps and PSF show variations with increasing projection angle, with apparent asymmetry appearing at projection angles beyond 10 degrees. When the projection angle is increased from 0 to 21 ‐degrees, while the breast thickness encountered by the central ray increases by 7.1%, the maximum SPR for a semi‐circular 10 cm radius breast increases by 10.1% and 18.8 % for breast thicknesses of 2 cm and 8 cm, respectively. Dose deposition shows a decrease, varying by 3.8–7.6% for the same thicknesses and projection angles. Conclusion: Since the use of an anti‐scatter grid is not easy to implement in tomosynthesisimaging, the development of software‐based post‐acquisition scatter reduction is important, which requires a good understanding of the scatter effects. This work characterizes the scatter signal present in tomosynthesisimages and shows that x‐ray scatter affects each projection angle differently and therefore each projection must be corrected separately, using appropriate prior knowledge. Decreased glandular dose with increasing projection angle must be taken into account when planning a tomosynthesis clinical protocol. Research supported in part by: NIH‐NIBIB Grant RO1‐ EB002123 and the Georgia Cancer Coalition.
33(2006); http://dx.doi.org/10.1118/1.2240154View Description Hide Description
Purpose: To evaluate the ability of a prototype breast CT scanner to detect micro‐calcifications, and to understand the influence that tube potential and radiation dose have on this. Method and Materials: Commercially available micro‐calcifications (μCa) of various sizes (200 to 425 μm) were embedded inside a 12.7 mm polyethylene tube filled with gelatin (to simulate glandular tissue). The gelatin tube was then placed inside a 14 cm diameter adipose equivalent cylindrical phantom and scanned using various tube potentials (60 to 100 kVp) and tube currents. CTimages were reconstructed with both Ramp and Shepp‐Logan filters, with a reconstructed voxel size of about 320×320×200 μm. The μCa were then evaluated quantitatively using signal‐to‐noise ratio (SNR) metric, and subjective appraisals were made as well. A dedicated breast CT visualization workstation was used for subjective evaluation. Results: Results for 250–280 mm μCa imaged at 80 kVp shown that the μCa are clearly visible when the rod is scanned by itself, but extremely difficult to locate when placed inside the 14 cm phantom. The visualization of the μCa improved overall for larger μCa, and overall visualization improves as the radiation levels are increased, as expected. Conclusion: These initial results suggested that the pixel size may not be a critical factor when determining the ability of the prototype system to visual micro‐calcifications, as the current objects scanned are only about 48% of the reconstructed voxel size. Maximum intensity projection (MIP) display for thick‐slice imaging was found to be most useful for subjective viewing of micro‐calcification clusters.
- Moderated Poster ‐ Area 4 (Imaging): Computed Tomography
33(2006); http://dx.doi.org/10.1118/1.2240233View Description Hide Description
Purpose: To investigate the effects of patient body size and lung size on the CT numbers of lung nodules measured with multi‐detector CTscanners and whether improved accuracy can be obtained with a dual‐energy technique. Method and Materials: Simulated lung nodules consisting of 9.5‐mm diameter spheres containing 50mg/cc and 100mg/cc CaCO3 in a water‐equivalent resin were scanned in two simulated thorax section phantoms with a GE VCT scanner. One phantom (A) represented the middle of the chest. It had large simulated lung regions and simulated ribs, heart and spine. The other (B) represented the upper chest. It had a much wider aspect ratio, smaller simulated lung regions, and simulated ribs, scapula, heart, and spine. Fat rings were added to the phantoms to simulate larger patients. Images were acquired on a GE VCT scanner with high‐resolution techniques (0.53:1 pitch, 0.625‐mm slice thickness and interval) at 80, 120 and 140kVp. Scans were repeated 3 times for reproducibility and analyzed using an automated technique. Results: Body size had a significant effect on the measured mean CT‐numbers of the nodules. For phantom‐A, adding fat rings decreased the overall average CT‐numbers of the 50mg/cc nodules at 120kVp by 15HU and those of the 100mg/cc nodules by 21HU. Corresponding reductions in phantom‐B were 9HU and 13HU. The dual‐energy approach (CT#80kVp‐CT#140kVp) reduces the variability, with a maximum difference of 4HU for all conditions. Lung size had a minimal effect with a maximum difference (nodule CT# phantom A ‐ nodule CT# phantom B) of 4.5 HU. Conclusion: Even with modern multi‐detector CTscanners, beam hardening and x‐ray scatter errors due to body size can result in underestimates of the true CT numbers of lung nodules. A dual‐energy approach compensates for these errors and should be considered especially if it can be implemented using a rapid kVp switching technique.
33(2006); http://dx.doi.org/10.1118/1.2240234View Description Hide Description
Purpose: Because diagnostic Computed Tomography(CT)imaging involves a tradeoff between image quality and radiation risk, there is great interest in determining the effects of stochastic noise on the utility of clinical tasks. The reconstruction processes used in CT result in noise properties that are non‐local and anisotropic in the image domain. A commonly used approximation for computing imagenoise from raw measurement data was empirically tested for validity. A noise variance mapping scheme was used to estimate stochastic noise in complex anatomical scenes and was compared to variance measurements of image simulations generated with controlled amounts of synthetic noise.Methods: The commonly assumed transformation between linear and log variance (σ2=1/Q) was tested for Poisson random numbers with means ranging from less than one to larger than 30. Noise variance maps were generated by filtered back projection using the square of the reconstruction kernel operating on sinogram variance estimates. A series of images was reconstructed by adding Poissonnoise to sinogram data, and the variance of regions of interest in the image sequence was calculated. Results: The approximation that log variance is proportional to the inverse number of quanta fails badly for N<10. Estimated variance maps were found to agree with empirical measurements of image variance. The noise variance in a CTimage is a slowly varying spatial function. Image simulations demonstrated that noise has a texture that is highly anisotropic and can mimic anatomic structures. Conclusions:CTnoise is a complex phenomenon. Variance maps are a useful tool for estimating noise in structured image regions where direct variance measurements fail. Fortunately most clinical scans operate at higher flux levels where the commonly used variance approximation is valid. Low‐dose protocols must be carefully evaluated to determine the effects of stochastic noise on diagnostic performance.
33(2006); http://dx.doi.org/10.1118/1.2240235View Description Hide Description
Purpose: Cone Beam CT(CBCT) kilovoltage imaging devices are increasingly available for daily imaging in radiotherapy departments. Flat‐panel based CBCTscanners present a distinctive set of artifacts due mostly to increased scatter, longer data acquisition time and reduced detector quantum efficiency as compared to helical Fan Beam CT (FBCT) systems. Our purpose is to characterize image quality from FBCT and CBCTscanners based on noise,contrast and dose, using FBCT as a benchmark. Method and Materials: we acquired phantom and clinical patient images with a CBCT Varian On‐Board Imager as well as with a FBCT Picker PQ5000 single‐row helical scanner. The CBCTscanner was equipped with antiscatter grid and bowtie filter. By comparing CBCT and FBCT images of a high contrast resolution insert, the CBCT reconstruction voxel size and filter were adjusted until the spatial resolution of the FBCT and CBCTimages was approximately matched. Dose was measured with standard CTDI and Farmer chambers. Noise,contrast and SNR were evaluated and compared. Results:CBCTimages of both phantom and patient were relatively free of streaking and cupping artifacts, indicating that the grid had successfully attenuated most of the scatter. Low contrastdetectability threshold is similar for the two modalities, when CBCTdose is about twice as large as FBCT. Noise and non‐uniformities are more prevalent in patient CBCTimages, but pelvic soft tissue structures are well discernible. For patient and phantom imagesDose×SNR2 is about 4 times lower for FBCT than in CBCT, which is about 1.5–2 times larger than expected, given the measured grid transmission and detector quantum efficiency. Conclusion: In this study, resolution‐matched CBCT and FBCT images could exhibit similar SNRs and contrast‐to‐noise ratios through a combination of increased imagingdose and reduced spatial resolution.
33(2006); http://dx.doi.org/10.1118/1.2240236View Description Hide Description
We compare the consistency and accuracy of two image binning approaches used in 4D‐CT imaging. In 4D‐CT the images and respiratory motion are correlated via RPM Respiratory Gating system (Varian, Palo Alto, CA). In phase binning (PB), RPM assigns each breathing cycle 2π radians, within which the images are grouped. Alternately, the images are assigned bins according to the signal's amplitude (AB). To quantitate both approaches, we used NEMA NU2‐2001 IEC phantom oscillating at random frequencies and amplitudes, simulating patients breathing. 4D‐CT images were obtained using 4‐slice GE Lightspeed CT operating in cine mode. We define consistency error as ability to correctly bin over breathing cycles in the same FOV. Average consistency error in PB ranged from 18%±20% to 30%±35%, while in AB the error ranged from 11%±14% to 20%±24%. For 28mm sphere, PB images were hardly consistent, with error of 43%±47%, while AB images for the same sphere resulted in error of only 18%±22%. In PB, while not all breathing cycles covered all phases, nearly all bins contained sphere slices. AB was more accurate, revealing empty bins where no sphere slices existed. As a proof of principle, we present examples of two NSCLC patients' 4D‐CT lungimages binned by both approaches. While AB can lead to gaps in the images, depending on patients' breathing pattern, PB exhibits no gaps but suffers visible artifacts due to misbinning, yielding images that covered a relatively large amplitude range. AB was more consistent, though often resulting in gaps corresponding to CT slices where no data existed due to patients' breathing. We conclude AB is more accurate than PB, which should be factored into treatment planning and diagnosis.
Work supported by GE Healthcare Technologies and Varian Medical Systems.
33(2006); http://dx.doi.org/10.1118/1.2240237View Description Hide Description
Purpose:Image‐guided adaptive radiotherapy proposes to use sequential CT studies to track anatomical change during treatment via deformable image registration. These CT studies can be acquired with either conventional fan‐beam CT systems or more novel cone‐beam CT techniques. However, cone‐beam CTimages can have higher noise levels and more imaging artifacts than fan‐beam CT, which might impact registration accuracy. We have investigated the effect of these image quality differences on the deformable registration of fan‐beam and simulated cone‐beam CTs. Method and Materials: Our study used two fan‐beam CT studies of a prostate patient, taken ten days apart. A deformable image registration process was used to register the two studies and then transfer treatment planning contours from one CT to the other. The accuracy of the automatically‐transferred contours (and thus of the deformable registration process) was assessed by comparing them to manual contours, with the differences evaluated with respect to inter‐observer variability in the manual contours. Then one of the fan‐beam CTs was modified to include higher noise and cupping artifacts characteristic of cone‐beam CT and the tests were repeated. Changes in registration accuracy were detected by monitoring changes in the automatically‐transferred contours. Results: We found that the additional noise and the cupping artifact caused no appreciable loss of registration accuracy at magnitudes up to and exceeding what would normally be found in an actual cone‐beam CT.Conclusions: We conclude that deficiencies in cone‐beam CT quality that might reduce manual contouring accuracy do not necessarily reduce image registration and automatic contouring accuracy.
33(2006); http://dx.doi.org/10.1118/1.2240238View Description Hide Description
Introduction: The aim of this study is to present a conceptually new method for metal artifact reduction (MAR), especially for patients who have multiple metal objects with small sizes. Metallic implants such as dental fillings cause serious artifacts in reconstructedCTimages. Although the previous methods based on conventional projection‐interpolation successfully reduced artifacts in the case of large metal objects such as hip prostheses, their performance appears to depend highly on the complexity of the structures examined and they are very sensitive in correctly detection of missing projections resulting still many artifacts in the final reconstruction for the case of multiple‐near metal objects. Methods and Materials: The proposed method is based on modifying the raw CT data acquired during patient's examination. First, the projection data affected by metal objects (missing projections) are detected in sinogram using a simple thresholding algorithm. Then, the missing projections are replaced by corresponding 180 degrees projections, which are not affected by metal objects. The idea beyond the replacing scheme is due to the fact that the two projections along the same path but in the opposite sides would be the same in the absence of table motion. So, in the presence of table motion, like an helical CT exam, the opposite side projections still constitute very good approximations for the corresponding missing projections. In order to make the replacing scheme more reliable, we start the process simultaneously from each side of missing projections area. Finally, the modified sinogram is transferred back to the CT scanner device where CT slices are regenerated using the built‐in reconstruction operator.
Experimental Results: The resulting tomography by the proposed approach show significant improvements in image quality, especially for regions near the metallic implants, compared to those by interpolation‐based approaches.
- Moderated Poster ‐ Area 3 (Imaging): Image Segmentation, Visualization, and Registration
TU‐EE‐A3‐01: 2D‐3D Registration of Portal Images with the Planning CT for Detection of Patient Positioning Errors33(2006); http://dx.doi.org/10.1118/1.2241599View Description Hide Description
Purpose: To compare the use of 2D‐3D automatic registration of portal images with the planning CT for detection of patient positioning errors and the use of 3D‐3D registration of MVCBCT with the planning CT.Method and Materials: Two prototype programs were used to carry out 2D‐3D and 3D‐3D image registrations. To assess the accuracy and robustness of these programs, 25 sets of 2D portal images, 25 sets of megavoltage conebeam CT (MVCBCT) images with known positioning shifts were acquired. A planning CT of the RANDO was also acquired. The known shifts between these image sets were ranged from −17mm to 4 mm, −20mm to 5mm and −12mm to 6mm, with uncertainty of 4.278mm, 5.359mm, 3.396mm along the latitude, longitude and vertical directions Results: The average differences between 2D‐3D method and the known shifts were −0.632±0.318 mm, −0.121±0.437 mm, −0.416±0.346 mm, compared to 3D‐3D method of 1.487±0.342 mm, −0.127±0.528 mm, 0.083±0.48 mm along the latitude, longitude and vertical directions. The average differences between 2D‐3D and 3D‐3D image registration methods were 0.86±0.286 mm, −1.39±0.347 mm, −0.33±0.303 mm Conclusion: Both 3D‐3D and 2D‐3D registration methods can detect positioning errors within 1 mm. For a rigid body, 2D‐3D method is sufficient. Conflict of Interest: This project is partly funded by SIEMENS.
33(2006); http://dx.doi.org/10.1118/1.2241600View Description Hide Description
Purpose: To implement and validate a 2D‐3D registration method for determining 3D patient position in external beam radiotherapy using orthogonal EPIDimages and megavoltage digitally reconstructedradiographs (MDRRs). To test the methods dependence on cost function, image pre‐processing and parameter space sample density, and determine the dependence of registered rotations on setup translations and vice versa. Method and Materials: Orthogonal EPIDimage of a humanoid phantom in different poses (3D rotations and translations) were acquired in anterior‐posterior and latero‐lateral view. The EPIDimages were registered with a data base of orthogonal MDRRs, calculated as projection images through the phantom's CT data set at rotation angles within ±5°. Registration results were compared for three different cost functions (least‐squares, cross‐correlation and mutual information), different image pre‐processing techniques (unsharp masking, histogram matching) and for isolated and combined rotations and translations. The influence of setup translations on registration results for rotations, and vice versa, was investigated and compared with a simple model. Results:Image pre‐processing improves registration precision by more than a factor 2. Three dimensional translations were registered with better than 0.5 mm (one standard deviation) when no rotations were present. Three‐dimensional rotations registered with a precision of better than 0.2° (1 SD) when no translations were present. Combined rotations and translations of up to 4° and 15 mm were registered with a precision of better than 0.4° and 0.7 mm respectively. Mutual information resulted in the most precise registration. Setup translations influence registered rotations, mostly following a simple theoretical model, but not vice versa. Conclusion: Precise registration requires image pre‐processing and benefits from interpolation of the parameter space. Influence of object translation on registration of out‐of‐plane rotations can be significant; these “pseudo rotations” can be corrected using the theoretical model when only one projection image is used for registration (e.g. fluoroscopy).
33(2006); http://dx.doi.org/10.1118/1.2241601View Description Hide Description
Purpose: Intraoperative quantitative C‐arm fluoroscopy guidance depends on discerning the relative pose of images (pose recovery). A possible method is to use radiographic fiducials visible in fluoro images [1,2]. We propose a robust and fast method for segmenting fiducials designed for brachytherapy applications. Methods and materials: The fiducial contains points, lines and ellipses made from BBs and wires. The algorithm integrates the a‐priori knowledge of fiducial's mechanical construction in a cleverly devised workflow. The BB segmentation is achieved using morphological top‐hat transform. This information serves as a heuristic input to line segmentation realized by a curve tracing algorithm which operates on edge image, followed by augmenting information from intensity image. Once the lines are segmented, this information feeds to the ellipse extraction step. For ellipse segmentation, intensity image is morphologically processed to eliminate background noise, followed by elimination of BB‐s and lines from the information obtained in prior steps. The resulting image consists of only ellipse segments. A fast variation of Hough transform is used to rectify the full ellipse from the segments. Results: The fiducial algorithm identified all the features (BBs, lines and ellipses) visible to human eye in all ten clinical images. Next the accuracy of fiducial segmentation was assessed numerically by feeding the results to the pose recovery algorithm of . The fiducial was moved on an accurate mechanical platform (as ground truth) while the C‐arm was stationary. We reconstructed the relative poses with an accuracy of 1.2 mm in translation and 0.3 degrees in rotational based on the segmented fiducials. Conclusions: The algorithm makes effective use of a‐priori knowledge and combines the techniques of morphological segmentation, curve tracing, and Hough transform, resulting in a novel curve segmentation strategy.
TU‐EE‐A3‐04: Massive Training Artificial Neural Network (MTANN) to Reduce False Positives Due to Rectal Tubes in Computer‐Aided Polyp Detection33(2006); http://dx.doi.org/10.1118/1.2241602View Description Hide Description
Purpose: One limitation of current computer‐aided detection (CAD) of polyps in CT colonography is a relatively large number of false positives. Rectal tubes are a common source of false positives and may distract the reader from less common polyps in the rectum. Our purpose was to develop a three‐dimensional massive‐training artificial neural network (3D MTANN) for reduction of false positives due to rectal tubes generated by a CAD scheme. Material and Methods: Our database consisted of CT colonography of 73 patients, scanned in both supine and prone positions. Fifteen patients had 28 polyps (15 polyps: 5–9 mm; 13 polyps: 10–25 mm). These cases were subjected to our previously reported CAD scheme that included shape‐based detection of polyps and reduction of false positives with a Bayesian neural network. With this scheme, 96.4% (27/28) by‐polyp sensitivity with 3.1 (224/73) false positives per patient was achieved. To eliminate false‐positive rectal tubes, we developed a 3D MTANN that was trained to enhance polyps and suppress rectal tubes. Results: In the output volumes of the trained 3D MTANN, various polyps were represented by distributions of bright voxels, whereas rectal tubes appeared as darker voxels. The 3D MTANN removed all 20 false‐positive rectal tubes produced by our original CAD scheme without removing any true positives. To evaluate the overall performance, we applied the 3D MTANN to the entire database containing 27 polyps (true positives) and 224 non‐polyps (false positives). The 3D MTANN eliminated 33% (73/224) of non‐polyps without removal of any true positives in an independent test. Conclusion: The 3D MTANN was able to improve the false‐positive rate of our original CAD scheme from 3.1 to 2.1 false positives per patient, while an original by‐polyp sensitivity of 96.4% was maintained. Conflict of Interest: HY, SGA: shareholders, R2 Technology, Inc.
33(2006); http://dx.doi.org/10.1118/1.2241603View Description Hide Description
Purpose: Over the last few years liquid crystal displays have replaced the traditional light box in medical imaging. When display systems are used for primary diagnosis they need to comply with the DICOM GSDF standard. Previous work already demonstrated that limited grayscale depth and viewing angle dependency of LCD results into lower calibration accuracy. As original contribution this paper investigates the impact of spatial noise and non‐uniformities on calibration accuracy. Additionally we will also quantify the relative importance of each of those three shortcomings of medicalLCDs.Method and Materials: To study the effect of non‐uniformity and spatial noise on DICOM GSDF compliance, we compare the observed transfer curve after calibration with the target DICOM GSDF curve and this for multiple locations across the display. The display used is a 5 Mega Pixel monochrome medicalLCD display. We also analyze the typical GSDF compliance metrics dL/L and JNDs/step and compare these plots for the other effects of viewing angle dependency and grayscale depth. Results: Current calibration practice is to characterize the native transfer curve of the display at only one position (often the centre). However there is significant variation in native transfer curve across the display surface and therefore calibration will only be accurate at the position where the characterization took place. On other display positions the average target luminance distortion compared to GSDF ranges from a few JNDs to over 20 JNDs. A technique of spatial noise‐reduction can solve this problem. We also observe that spatial‐noise is much more important than viewing angle and grayscale depth for typical usage. Conclusion: This paper has demonstrated that non‐uniformities and spatial noise can result into poor calibration accuracy. Also a possible solution to the problem has been described. Conflict of Interest: Research sponsored by Barco Medical Imaging Systems.
33(2006); http://dx.doi.org/10.1118/1.2241604View Description Hide Description
Purpose: To evaluate an innovative volumetric display for radiation treatment planning applications. Method and Materials: A volumetric, auto‐stereoscopic display (Perspecta Spatial 3D, Actuality‐Systems Inc., Bedford, MA) has been integrated with the Pinnacle3 TPS for treatment planning. The Perspecta 3D display renders a 25 cm diameter volume that is viewable from any side, floating within a translucent dome. In addition to display all the 3D data exported from Pinnacle, the system provides a 3D cursor and beam placement tools. A 125 point, 5 cm spaced grid centered at isocenter was created in Pinnacle and transferred to Perspecta. A Perspecta 3D ruler verified distances between any two points on the 3D grid. Ten teletherapy beams with various gantry/couch combinations were generated on Pinnacle and verified on Perspecta display. Doses at the same grid points were also compared. CTimages from a QUASAR phantom in 3 orientations were used on Perspecta to confirm beam field size, divergence, etc. Results: In general, the Perspecta system accurately depicted all 3D data exported from Pinnacle. When measured by the 3D ruler, distances between any two points using Perspecta agreed with Pinnacle within the measurement error (typically <0.5 mm). Beam angles were verified through Cartesian coordinate system measurements and also upon rotating the phantom. Field sizes,collimator angles, and beam divergence were similarly confirmed. Isodose surfaces and dose values chosen at arbitrary locations in Perspecta agreed with Pinnacle within ± 2% in an absolute sense, which was governed by human error in coinciding the points. Conclusions: These preliminary results indicate that the Perspecta device is capable of displaying consistent data from the Pinnacle radiotherapytreatment planning system, and may become a valuable tool for visualization and quantitative evaluations in radiation oncology. Conflict of Interest Statement: Actuality Systems Inc. provided the 3D display used in this study.
- Moderated Poster ‐ Area 4 (Imaging): Imaging Dosimetry and Quality Control
TU‐EE‐A4‐01: Impact of Hardware Errors and Patient Movement On Standardized MR QA Tests for Noncartesian Pulse Sequences: Are These Standard QA Tests Sufficient?33(2006); http://dx.doi.org/10.1118/1.2241605View Description Hide Description
Purpose: The American College of Radiology (ACR) phantom was designed during a stage when most pulse sequences were rectilinear. Noncartesian trajectories (including Projection Reconstruction and Spiral imaging) are becoming more widely available on most standard clinical scanners. Our goal was to assess if either 1) ACR quality assurance (QA) tests, 2) Root Mean Square Error (RMSE), or 3) by Visual Inspection, or a combination of the above methods would adequately assess the types of image artifacts associated with these radial methods. Method and Materials: Simulation in Matlab and C was written to reconstruct ACR phantom images using projection and spiral methods. Additive and multiplicative phase distortions and as well as center of k‐space misregistration artifacts were introduced to test their impact on phantom images for both reconstruction methods. Additional artifacts of gradient delay, amplitude modulation and miscalibration of gradient amplitude were tested for spiral reconstruction.Images were then evaluated using the above mentioned techniques. Results: Gradient delay, miscalibration of gradient modulation and k‐space center misregistration affected all tests. However, the ACR QA test did not effectively demonstrate the influence of certain artifacts (phase distortion artifacts and ghosting) in projection and spiral reconstructed images. RSME was effective for most tests in determining the overall severity (i.e. larger RMSE demonstrates more artifacts). However, RMSE was not useful for assessing phase distortion. Conclusions: We noted that for specific artifacts, QA and RMSE tests were not sufficient alone. Visual inspection is time consuming, but is not necessarily a true objective test of performance as ACR QA. A compendium of noncartesian induced hardware artifacts was generated for Physicist Referral and we believe a combination of all above methods 1) ACR QA, 2) RMSE and Visual Inspection would be the appropriate test for validating scanner performance.
33(2006); http://dx.doi.org/10.1118/1.2241606View Description Hide Description
Purpose: To investigate why Image Intensity Uniformity (IIU) tests performed on commercial 3T MRI systems according to the ACR MR Accreditation Program (MRAP) instructions are subject to frequent failures. Method and Materials: The phantom scanning instructions for ACR MRAP tests specify that a dedicated phantom (ACR phantom) must be used to assess the MRIscanners performance. As a part of required test, an assessment of IIU is performed by measuring the Percent Image Uniformity (PIU) according to the specified formula. The data used in calculation of PIU must be acquired for a rigidly defined scanning protocols, and the process of data collection is also precisely specified. For the 3T MRIscanner to pass the test, the measured PIU must be equal or greater than 82%.
PIU tests were performed using the same ACR phantom on five different 3T MR scanners from two different vendors, following the ACR MRAP acquisition and measurement protocol exactly. The scanners tested were installed in routine clinical settings and all head coils, available on site, were tested independently. In addition, if scanner offered multiple gradient performance modes, all modes were tested. Finally, effects of image post‐processing routines, designed to improve the image uniformity and installed on the scanners were investigated as well. Results: A vast majority of PIU tests using non‐postprocessed images failed (more than 95%). All tests that used postprocessed images passed. Upon closer investigation, the likely cause of failures was linked to RF properties of the ACR phantom. Conclusion: The IIU test needs to be carefully reassessed to ensure that it provides a meaningful characterization of MRI scanner's performance, and is not hampered by the limitations imposed by the electromagnetic physics of scanning at 127 MHz, the resonant RF frequency of 3T MRI systems.