KEYWORDS: Radiotherapy, Data processing, Technologies and applications, Medical imaging, Decision support systems, Human-machine interfaces, Head, Neck, Tissues, Tumors, Databases, Prototyping, Data modeling, Information science
The primary goal in radiation therapy is to target the tumor with the maximum possible radiation dose while limiting the radiation exposure of the surrounding healthy tissues. However, in order to achieve an optimized treatment plan, many constraints, such as gender, age, tumor type, location, etc. need to be considered. The location of the malignant tumor with respect to the vital organs is another possible important factor for treatment planning process which can be quantified as a feature making it easier to analyze its effects. Incorporation of such features into the patient’s medical history could provide additional knowledge that could lead to better treatment outcomes. To show the value of features such as relative locations of tumors and surrounding organs, the data is first processed in order to calculate the features and formulate a feature matrix. Then these feature are matched with retrospective cases with similar features to provide the clinician with insight on delivered dose in similar cases from past. This process provides a range of doses that can be delivered to the patient while limiting the radiation exposure of surrounding organs based on similar retrospective cases. As the number of patients increase, there will be an increase in computations needed for feature extraction as well as an increase in the workload for the physician to find the perfect dose amount. In order to show how such algorithms can be integrated we designed and developed a system with a streamlined workflow and interface as prototype for the clinician to test and explore. Integration of the tumor location feature with the clinician’s experience and training could play a vital role in designing new treatment algorithms and better outcomes. Last year, we presented how multi-institutional data into a decision support system is incorporated. This year the presentation is focused on the interface and demonstration of the working prototype of informatics system.
We have developed an imaging informatics based decision support system that learns from retrospective treatment plans
to provide recommendations for healthy tissue sparing to prospective incoming patients. This system incorporates a
model of best practices from previous cases, specific to tumor anatomy. Ultimately, our hope is to improve clinical
workflow efficiency, patient outcomes and to increase clinician confidence in decision-making. The success of such a
system depends greatly on the training dataset, which in this case, is the knowledge base that the data-mining algorithm
employs. The size and heterogeneity of the database is essential for good performance. Since most institutions employ
standard protocols and practices for treatment planning, the diversity of this database can be greatly increased by
including data from different institutions. This work presents the results of incorporating cross-country, multi-institutional
data into our decision support system for evaluation and testing.
We have developed a comprehensive DICOM RT specific database of retrospective treatment planning data for radiation therapy of head and neck cancer. Further, we have designed and built an imaging informatics module that utilizes this database to perform data mining. The end-goal of this data mining system is to provide radiation therapy decision support for incoming head and neck cancer patients, by identifying best practices from previous patients who had the most similar tumor geometries. Since the performance of such systems often depends on the size and quality of the retrospective database, we have also placed an emphasis on developing infrastructure and strategies to encourage data sharing and participation from multiple institutions. The infrastructure and decision support algorithm have both been tested and evaluated with 51 sets of retrospective treatment planning data of head and neck cancer patients. We will present the overall design and architecture of our system, an overview of our decision support mechanism as well as the results of our evaluation.
KEYWORDS: Data modeling, Decision support systems, Databases, Computed tomography, Radiotherapy, Data mining, Tumors, Data processing, Imaging informatics, Machine learning
We have built a decision support system that provides recommendations for customizing radiation therapy treatment plans, based on patient models generated from a database of retrospective planning data. This database consists of relevant metadata and information derived from the following DICOM objects - CT images, RT Structure Set, RT Dose and RT Plan. The usefulness and accuracy of such patient models partly depends on the sample size of the learning data set. Our current goal is to increase this sample size by expanding our decision support system into a collaborative framework to include contributions from multiple collaborators. Potential collaborators are often reluctant to upload even anonymized patient files to repositories outside their local organizational network in order to avoid any conflicts with HIPAA Privacy and Security Rules. We have circumvented this problem by developing a tool that can parse DICOM files on the client’s side and extract de-identified numeric and text data from DICOM RT headers for uploading to a centralized system. As a result, the DICOM files containing PHI remain local to the client side. This is a novel workflow that results in adding only relevant yet valuable data from DICOM files to the centralized decision support knowledge base in such a way that the DICOM files never leave the contributor’s local workstation in a cloud-based environment. Such a workflow serves to encourage clinicians to contribute data for research endeavors by ensuring protection of electronic patient data.
Cancer registries are information systems that enable easy and efficient collection, organization and utilization of data related to cancer patients for the purpose of epidemiological research, evidence based medicine and planning of public health policies. Our research focuses on developing a web-based system which incorporates aspects of both cancer registry information systems and medical imaging informatics, in order to provide decision support and quality control in external beam radiation therapy. Integrated within this system is a knowledge base composed of retrospective treatment plan data sets of 42 patients, organized in a systematic fashion to aid query, retrieval and data mining. A major cornerstone of our system is the use of DICOM RT data sets as the building blocks of the database. This offers enormous practical advantages since it establishes a framework that can assimilate data from different treatment planning systems and across institutions by making use of a widely used standard – DICOM. Our system will help clinicians to assess their dose volume constraints for prospective patients. This is done by comparing the anatomical configuration of an incoming patient’s tumor and surrounding organs, to that of retrospective patients in the knowledge base. Treatment plans of previous patients with similar anatomical features are retrieved automatically for review by the clinician. The system helps the clinician decide whether his dose/volume constraints for the prospective patient are optimal based on the constraints of the matched retrospective plans. Preliminary results indicate that this small-scale cancer registry could be a powerful decision support tool in radiation therapy treatment planning in IMRT.
Di Zhang, Maryam Khatonabadi, Hyun Kim, Matilda Jude, Edward Zaragoza, Margaret Lee, Maitraya Patel, Cheryce Poon, Michael Douek, Denise Andrews-Tang, Laura Doepke, Shawn McNitt-Gray, Chris Cagnon, John DeMarco, Michael McNitt-Gray
KEYWORDS: Diagnostics, Image quality, Computed tomography, Liver, Monte Carlo methods, Scanners, Data modeling, Medical imaging, Radiation effects, Inflammation
Purpose: While several studies have investigated the tradeoffs between radiation dose and image quality (noise) in CT
imaging, the purpose of this study was to take this analysis a step further by investigating the tradeoffs between patient
radiation dose (including organ dose) and diagnostic accuracy in diagnosis of appendicitis using CT. Methods: This
study was IRB approved and utilized data from 20 patients who underwent clinical CT exams for indications of
appendicitis. Medical record review established true diagnosis of appendicitis, with 10 positives and 10 negatives. A
validated software tool used raw projection data from each scan to create simulated images at lower dose levels (70%,
50%, 30%, 20% of original). An observer study was performed with 6 radiologists reviewing each case at each dose
level in random order over several sessions. Readers assessed image quality and provided confidence in their diagnosis
of appendicitis, each on a 5 point scale. Liver doses at each case and each dose level were estimated using Monte Carlo
simulation based methods. Results: Overall diagnostic accuracy varies across dose levels: 92%, 93%, 91%, 90% and
90% across the 100%, 70%, 50%, 30% and 20% dose levels respectively. And it is 93%, 95%, 88%, 90% and 90%
across the 13.5-22mGy, 9.6-13.5mGy, 6.4-9.6mGy, 4-6.4mGy, and 2-4mGy liver dose ranges respectively. Only 4 out of
600 observations were rated "unacceptable" for image quality. Conclusion: The results from this pilot study indicate that
the diagnostic accuracy does not change dramatically even at significantly reduced radiation dose.
Recently published AAPM Task Group 204 developed conversion coefficients that use scanner reported CTDIvol to
estimate dose to the center of patient undergoing fixed tube current body exam. However, most performed CT exams use
TCM to reduce dose to patients. Therefore, the purpose of this study was to investigate the correlation between organ
dose and a variety of patient size metrics in adult chest CT scans that use tube current modulation (TCM).
Monte Carlo simulations were performed for 32 voxelized models with contoured lungs and glandular breasts tissue,
consisting of females and males. These simulations made use of patient's actual TCM data to estimate organ dose. Using
image data, different size metrics were calculated, these measurements were all performed on one slice, at the level of
patient's nipple. Estimated doses were normalized by scanner-reported CTDIvol and plotted versus different metrics.
CTDIvol values were plotted versus different metrics to look at scanner's output versus size.
The metrics performed similarly in terms of correlating with organ dose. Looking at each gender separately, for male
models normalized lung dose showed a better linear correlation (r2=0.91) with effective diameter, while female models
showed higher correlation (r2=0.59) with the anterior-posterior measurement. There was essentially no correlation
observed between size and CTDIvol-normalized breast dose. However, a linear relationship was observed between
absolute breast dose and size. Dose to lungs and breasts were consistently higher in females with similar size as males
which could be due to shape and composition differences between genders in the thoracic region.
Our goal in this paper is to data mine the wealth of information contained in the dose-volume objects used in external
beam radiotherapy treatment planning. In addition, by performing computational pattern recognition on these mined
objects, the results may help identify predictors for unsafe dose delivery. This will ultimately enhance current clinical
registries by the inclusion of detailed dose-volume data employed in treatments. The most efficient way of including
dose-volume information in a registry is through DICOM RT objects. With this in mind, we have built a DICOM RT
specific infrastructure, capable of integrating with larger, more general clinical registries, and we will present the results
of data mining these sets.
The purpose of this study was to investigate the accuracy of Monte Carlo simulated organ doses using cylindrical ROIs
within the organs of patient models as an alternative method to full organ segmentations. Full segmentation and
placement of circular ROIs at the approximate volumetric centroid of liver, kidneys and spleen were performed for 20
patient models. For liver and spleen, ROIs with 2cm diameter were placed on 5 consecutive slices; for the kidneys 1cm
ROIs were used. Voxelized models were generated and both fixed and modulated tube current simulations were
performed and organ doses for each method (full segmentation and ROIs) were recorded. For the fixed tube current
simulations, doses simulated using circular ROIs differed from those simulated using full segmentations: for liver, these
differences ranged from -5.6% to 10.8% with a Root Mean Square (RMS) difference of 5.9%. For spleen these
differences ranged from -9.5% to 5.7% with an RMS of 5.17%; and for kidney the differences ranged from -12.9% to
14.4% for left kidney with an RMS of 6.8%, and from -12.3% to 12.8% for right kidney with an RMS of 6.6%. Full
body segmentations need expertise and are time consuming. Instead using circular ROIs to approximate the full
segmentation would simplify this task and make dose calculations for a larger set of models feasible. It was shown that
dose calculations using ROIs are comparable to those using full segmentations. For the fixed current simulations the
maximum RMS value was 6.8% and for the TCM it was 6.9%.
Pregnant women with shortness of breath are increasingly referred for CT Angiography to rule out Pulmonary Embolism (PE). While this exam is typically focused on the lungs, extending scan boundaries and overscan can add to the irradiated volume and have implications on fetal dose. The purpose of this work was to estimate radiation dose to the fetus when various levels of overscan were encountered.
Two voxelized models of pregnant patients derived from actual patient anatomy were created based on image data. The models represent an early (< 7 weeks) and late term pregnancy (36 weeks). A previously validated Monte Carlo model of an MDCT scanner was used that takes into account physical details of the scanner. Simulated helical scans used 120 kVp, 4x5 mm beam collimation, pitch 1, and varying beam-off locations (edge of the irradiated volume) were used to represent different protocols plus overscan. Normalized dose (mGy/100mAs) was calculated for each fetus.
For the early term and the late term pregnancy models, fetal dose estimates for a standard thoracic PE exam were estimated to be 0.05 and 0.3 mGy/100mAs, respectively, increasing to 9 mGy/100mAs when the beam-off location was extended to encompass the fetus.
When performing PE exams to rule out PE in pregnant patients, the beam-off location may have a large effect on fetal dose, especially for late term pregnancies. Careful consideration of ending location of the x-ray beam - and not the end of image data - could result in significant reduction in radiation dose to the fetus.
Multidetector CT (MDCT) systems offer larger coverage and wider z-axis beams, resulting in larger cone angles. One impact on radiation dose is that while radiation profiles at isocenter are constant when contiguous axial scans are performed, the increased beam divergence from the larger cone angle results in significant surface dose variation. The purpose of this work was to measure the magnitude of this effect. Contiguous axial scans were acquired using an MDCT for two sizes of cylindrical phantoms and an anthropomorphic phantom. Film dosimetry and/or radiation detector measurements were performed on the surface of each phantom. Detailed mathematical models were developed for the MDCT scanner and all phantoms. Monte Carlo simulations of contiguous axial scans were performed for each phantom model. From cylindrical phantoms, film dosimetry at the surface showed differences between peak and valley that reached 50%. From the anthropomorphic phantom, measured values ranged from 7.9 to 16.2 mGy at the phantom surface. Monte Carlo simulations demonstrated these variations in both cylindrical and anthropomorphic phantoms. The magnitude of variation was also related to object size. Even when contiguous axial scans are performed on MDCT, surface radiation profiles show considerable variation. This variation will increase as MDCT cone angles increase and when non-contiguous scans (e.g. pitch > 1) are acquired. The variation is also a function of object size. While average surface doses may remain constant, peak doses may increase, which may be significant for radiation sensitive organs at or near the surface (e.g. breast, thyroid).
The CT image is a representation of the patient's anatomy as measured in terms of such physical characteristics as density, electron density, and atomic number. The process of sampling the patient's molecular composition with the x-ray beam is subject to varied physical effects that degrade the ability of the CT image data set to accurately represent the tissues of the body. To assess the impact of patient and scanner related characteristics on the final CT image a Monte Carlo based modeling has been developed that can simulate such effects as scatter and beam hardening on the reconstructed CT image. By selectively studying the effects of these variables, the model can be used as a design tool for improving the diagnostic capabilities of a CT scanner and /or correcting or eliminating unwanted sources of variation in the CT image. Initial simulations model the unique geometry of Electron Beam CT with a clinical goal of correcting for scanner and patient related physical factors that may cause variations in the assessment of Coronary Artery Calcium.
The field programmable gate array (FPGA) is a promising technology for increasing computation performance by providing for the design of custom chips through programmable logic blocks. This technology was used to implement and test a hardware random number generator (RNG) versus four software algorithms. The custom hardware consists of a sun SBus-based board (EVC) which has been designed around a Xilinx FPGA. A timing analysis indicates the Sun/EVC hardware generator computes 1 multiplied by 106 random numbers approximately 50 times faster than the multiplicative congruential algorithm. The hardware and software RNGs were also compare using a Monte Carlo photon transport algorithm. For this comparison the Sun/EVC generator produces a performance increase of approximately 2.0 versus the software generators. This comparison is based upon 1 multiplied by 105 photon histories.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.