| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
ABSTRACT |
Monday, May 7, 10:00 AM - 12:00 PM
Abstracts 035-045
Moderator(s): Hani Abujudeh, William Weadock, and Walter Carpenter
10:00 AM
Keynote Address: What Do We Mean by Quality in Radiology?
Hani Abujudeh, Massachusetts General Hospital, Boston, MA
10:10 AM
035. Self-Subsidization of Educational Expenses by Senior Radiology Residents
Tilak G. S.*; Baker S. R. UMDNJ - New Jersey Medical School, Newark, NJ
Address correspondence to G. Tilak (gtilak1{at}hotmail.com)
Objective: The purpose of this study is to document the degree of self-subsidization of educational expenses by senior radiology residents.
Materials and Methods: Questionnaires were distributed to all radiology residents (n = 176) attending the week-long New Jersey Medical School board review course held twice in March and May 2006. Respondents (n = 175) documented the number and source of financial support for review courses they would attend, including the AFIP course in radiologic pathology. They also listed the amount of additional financial allowances paid to them by their programs and cited the funding source for the radiology board examinations.
Results: Average AFIP expenditure, including tuition, room, board, and travel, equaled $3,441 ± 52, of which 44% was paid by the residents themselves. Residents attended on average 2 review courses costing $4,116 ± 149, with residents bearing 77% of the costs. The average additional allowance paid to residents by their program was $1,938 ± 156. Total board expenditures of $3,120, including fees and travel, were borne entirely by the residents. Total out-of-pocket expenses for these activities was $7,515 ± 183, which amounted to 15% of senior residents' average annual salary.
Conclusion: The desire by both programs and trainees for success on the radiology board examination has stimulated the growth of preparatory review courses. The enduring popularity of the AFIP course has made this activity, too, a seemingly essential rotation for most radiology residency programs. Each of these off-site opportunities incurs significant financial obligations to residents, and when added to the cost of the board exams, equals 15% of their average annual salary. Thus, radiology residents are subsidizing their education to a considerable degree relative to their salaries.
036. Mashdot: A Web-Based Medical Journal Club
Garg N.4*; Akkanti B.1; Narvy S.3; Sekra A.2; Bader D.4 1. Boston Medical Center, Boston, MA; 2. Kazakh National Medical University, Almaty, Kazakhstan; 3. Keck School of Medicine USC, Los Angeles, CA; 4. St. Vincent Hospital, Worcester, MA
Address correspondence to N. Garg (naveen.garg{at}gmail.com)
Objective: To create a virtual journal club: www.mashdot.com, to supplement and enhance traditional journal clubs. The internet is changing the way that science is read, reported, and reviewed. The wikipedia has become more widely read and is as accurate as Britannica (Nature 2005: 438, 900-901). Blogs.nature.com has a debate on web-based peer review. Journalreview.org has some elements of a virtual journal club, allowing users to perform a Pubmed query, and then, read, rate, and discuss articles from the query results. Mashdot expands on these ideas and the traditional journal club, and adds community features such as moderation, comment tags and ratings, live chat, rss feeds, and customizable pages for users.
Materials and Methods: The site is hosted on Ubuntu distro of Linux operating system using apache server and mysql database. The backend for the site uses an open source community blogging software called slashcode (originally used in slashdot.org, a popular technology news blog). Mashdot is slashcode modified for medical news and discussion. Modifications were made using perl, lisp, mysql, javascript, and html. Article reviews are meta-moderated by topic editors in various specialties.
Results: Features: 1) Comment Tags: Positive comment tags were modified to `Investigable', `TeachingPoint', `Accurate', `Informative', and negative comment tags are, `Spam', `Inaccurate', `Offtopic', and `Troll'. 2) Distributed Moderation: Eligibility to rate comments requires just 1 comment page view, and >50% of users are eligible for moderation. 3) Live Chat. 4) Images and movies in article reviews. 5) Pubmed search. 6) Customizable RSS feeds from medical journals.
Conclusion: Longevity and attendance of traditional journal clubs has been correlated to mandatory attendance, food, and perceived importance to the program director (Alguire. 1998). Mashdot cannot offer virtual food, but it has the potential to increase participation by allowing an interdepartmental, interdisciplinary, and international audience. At St. Vincent's, Mashdot allows residents on away rotations to participate in the journal club. We plan to coordinate the monthly topics of journal club with other institutions to increase efficiency and quality of discussions.
037. Can a Medical Student Learn to Generate Diagnostic Quality 3D CT Angiography Images and Identify the Pathology Demonstrated During an 8-Week Research Experience?
Acosta K.; Manohar C.*; Eade L.; Olson M. Loyola University Medical Center, Maywood, IL
Address correspondence to C. Manohar (cmanohar{at}lumc.edu)
Objective: The purpose was to determine if a medical student with a basic understanding of anatomy and no clinical training could learn to identify normal structures and pathology and then create diagnostic quality 3D CT angiography images.
Materials and Methods: A medical student spent 8 weeks in the CT 3D radiology laboratory. The first 2 weeks were spent observing the acquisition of CT angiography images and postprocessing of these images in the 3D lab. During the next 4 weeks, the student learned how to use a 3D workstation (GE Advantage 4.2) to produce multiplanar reformatted, maximum intensity projection, and volume-rendered images. In the final 2 weeks, the student reviewed and postprocessed twenty random arterial and venous cases. These cases were also independently reviewed and postprocessed by the 3D lab technologist administrator. The student and technologist were asked to list the pertinent findings. Two radiologists, blinded to the author of the postprocessing images, independently reviewed the two sets of images and the lists of findings. Diagnostic quality of the images, including demonstration of anatomy and pathology, was evaluated for the presence and absence of 5 factors: appropriate MIP thickness, appropriate magnification factor, optimal obliquity and correct opacity, and utilization of cut planes in volume rendering. Both sets of images were graded for overall diagnostic quality using a grading system of nondiagnostic, adequate diagnostic, and superior diagnostic. Differences of opinion were resolved by consensus.
Results: Evaluation was based on 19/20 cases. One case was eliminated due to poor contrast enhancement. All postprocessed images were of diagnostic quality. The use of appropriate MIP thickness and magnification factor and correct opacity and utilization of cut planes in volume rendering were comparable for both image sets. The technologist created optimal oblique images in 16 cases compared to 11 cases by the student. The overall diagnostic quality was graded superior in 14 and adequate in 5 cases created by the technologist, and superior in 1 and adequate in 18 by the student. There were a total of 66 pertinent findings. The student listed 37 and the technologist listed 38 of these findings.
Conclusion: After 8 weeks of training, a first-year medical student achieved adequate postprocessing skills and demonstrated understanding of basic anatomy and pathology that was comparable to a dedicated 3D CT technologist.
038. Radiologist Recommendation for Follow-up Exams: What are the Common Reasons?
Chatterji M.; Lee S. I.*; Dreyer K.; Thrall J. H.; Hahn P. F. Massachusetts General Hospital, Boston, MA
Address correspondence to M. Chatterji (mchatterji{at}partners.org)
Objective: Previous analysis identified high-cost outpatient exams recommended by a radiologist in a prior report. This study identifies by organ system the common findings on the prior exam that result in radiologist recommendation for follow-up.
Materials and Methods: 1,334 pairs of prior and follow-up exams were identified systematically from an institutional RIS database of all radiology reports (>125,000) from 7/05–12/05. Each pair represented a high-volume high-cost outpatient radiology exam (brain MR, chest and abdomen CT, and body PET) generated by a recommendation for follow-up by a prior exam within 60 days. The most common recommended pairs—chest X ray or chest CT leading to chest CT (626), brain MR leading to brain MR (120), and abdomen CT leading to abdomen CT (115)—were analyzed. Prior exam reports were reviewed and the finding leading to the recommendation was recorded by two radiologists independently.
Results: Lung nodule is the most common reason for a chest X ray or chest CT to recommend a chest CT, comprising 343/626 (55%) of the pairs and 343/705 (49%) of recommended chest CT exams. Mass follow-up is the most common reason for a brain MR to recommend a brain MR, comprising 27/115 (23%) of the pairs and 25/249 (10%) of recommended MR exams. Renal lesion is the most common reason for an abdomen CT to recommend an abdomen CT, comprising 27/115 (23%) of the pairs and 27/329 (8%) of recommended abdomen CT exams.
Conclusion: The most common findings on the prior exam resulting in a radiologist recommendation for follow-up are lung nodule for chest CT, mass for brain MR, and renal lesion for abdomen CT. Imaging findings generating radiologist recommendation for follow-up should be used to develop evidencebased criteria for standardizing radiologist recommendation for each organ system.
039. 2007 Current Status, Utility, and Growth of Open Source Projects in Radiology
Laks M. P.* Albert Einstein College of Medicine, Montefi ore Medical Center, Bronx, NY
Address correspondence to M. Laks (mlaks2000{at}yahoo.com)
Objective: The purpose of this study is to review the status in 2007 of the major open source radiology software projects active today. We will describe their use, evaluate what they have contributed in the past, and their prospects for future contributions in modern radiology departments.
Materials and Methods: The internet provides many resources to track active open source projects that are announced. There are mailing lists, software users groups, general categorical review sites as well as known internet repositories of open source software such as sourceforge and Savannah.gnu.org. We will describe different options on how to get the software, compile and use it.
Results: We will review the status and prospects of the major projects like Osirix, DCMTK. dcm4che, wustl ctn, amide, aeskulap, jdicom, Image/J. We also discuss the role of 3d slicer the Julius software Frameworks. Other products available for free use such as MeVisLab will be mentioned. We will discuss the advantages of open software development environments such as Linux and the MacOsx as well as available software on Windows platforms.
Conclusion: Open source radiology software is robust and a growing resource. Every radiology department should have individuals who are familiar with these resources, and active deployment will provide great benefit for both clinicians and researchers.
040. Do-it-Yourself PACS/RIS Automation for the Radiologist: Humanizing User-indifferent Computer Interfaces
Yao L.* MRIDx, Bethesda, MD
Address correspondence to L. Yao (yaolawrence{at}yahoo.com)
Objective: Radiologists are at serious risk for repetitive stress injury (RSI) due to the ubiquity of filmless operations and teleradiology, increasing workloads and image data volumes, and dependence on non-ergonomic computer user interfaces. Low- and no-cost automation strategies for minimizing radiologist exposure to RSI are presented.
Materials and Methods: Basic interface modifications can reduce the risk of RSI for radiologists. Effective automation aims to reduce mouse clicks and especially mouse dragging operations. Commercial products such as a customizable trackball are indispensable. Mapping graphical software commands to keyboard shortcuts is also essential, but may not be sufficiently flexible or robust on many viewing software platforms. The open source, free-ware Windows (Microsoft) utility AutoHotkey is introduced as a strategy that carries custom hotkey and mouse/cursor operations beyond what is often possible on commercial user interfaces.
Results: Common and serious shortcomings in typical PACS/RIS interfaces are illustrated. Examples of simple automation solutions to user interface deficiencies are presented. Repetitive mouse clicks required to use a common report verification interface are eliminated by a simple mouse automation script. Another script eliminates the multiple keystrokes required to open a cross-sectional imaging study according to user preferences on a common viewing platform.
Conclusion: The risk of RSI can be reduced, and work efficiency can be increased, by tailoring user interfaces more carefully to radiologists' special needs. This process is often largely equivalent to `getting the mouse out of computing.' The necessary automation methods can be simple to implement, are immensely helpful to the radiologist, and often cost nothing.
041. The Good, the Bad, and the Evil: Problematic User Interfaces in Current Commercial Radiology Software
Laks M. P.* Albert Einstein College of Medicine, Montefi ore Medical Center, Bronx New York, NY
Address correspondence to M. Laks (mlaks2000{at}yahoo.com)
Objective: Over the last few years there has been a rapid deployment of computerized radiology information systems in the US. PACS systems, RIS systems, and hospital information systems have been installed in many departments. Unfortunately, these computer systems are quite complex and difficult to assess fully prior to installation. Beginning users may attribute problems that they encounter to their own lack of familiarity with the software. Fundamental problems can therefore often only be appreciated after a user becomes adept at using these systems.
Materials and Methods: Documentation of problems with user interfaces that have surfaced in rollout of RIS PACS and HIS systems from major vendors at a major medical center.
Results: Software problems with enterprise medical computer systems have less visibility and commentary in the literature than they should, given their importance in everyday work in medical centers. We illustrate this with our work on user interface issues. In our experience, many problems only surfaced well into the installation process of complex systems. Vendors frequently are not responsive to addressing issues.
Conclusion: We would like to do what we can to improve the care of our patients by improving the software tools that we use on a daily basis. We will illustrate general principles of user interface design that are involved in the problems we experienced. We propose ways to disseminate information on the weaknesses experienced by our users of these systems to other potential adopters of these systems. We hope that the free and open discussion of these issues will inspire the vendors to improve these systems.
042. Affect of Gender on Speech Recognition Accuracy
Ali S.3*; Siddiqui K.4; Safdar N.3; Juluru K.2; Kim W.1; Siegel E.3 1. Hospital of the University of Pennsylvania, Philadelphia, PA; 2. Johns Hopkins University, Baltimore, MD; 3. University of Maryland, Baltimore, MD; 4. VA Maryland Health Care System, Baltimore, MD
Address correspondence to S. Ali (sma2b{at}yahoo.com)
Objective: Speech recognition applications are increasingly being utilized within busy radiology departments to enhance report output, while offering other work-flow enhancement features. Key to the success of these applications is recognition accuracy, which can be affected by multiple factors, including computer processor power, microphone quality, background noise, and training. Despite optimization of multiple conditions, anecdotal evidence suggests that there remain variations in recognition accuracy between different users. In this study, we investigate the role of gender differences on speech recognition accuracy.
Materials and Methods: Five male and five female radiology residents were each trained on a commercial speech recognition application. Within a reading environment standardized for background noise and ambient conditions, each resident was asked to dictate a standardized set of 10 radiology reports containing a total of 2,123 words. Utilizing a commercial software solution, the generated reports were compared with the original reports, and error rates were calculated. The error rate was defined as the sum of the number of word insertions and deletions, divided by the total word count for a given report. Statistical analysis was performed using SPSS for Windows (v. 12.0.1), with determination of average error rates between groups and analysis of variance (ANOVA).
Results: Error rates in the male population ranged from 0.025 to 0.139, with a mean of 0.082 (standard deviation = 0.033). Error rates in the female population ranged from 0.015 to 0.206, with a mean of 0.100 (standard deviation = 0.044). The results show a significantly higher rate of recognition error in the female population compared to the male population (p = 0.02).
Conclusion: Our study showed that there is a significantly higher rate of transcription error in women compared to men using a commercial voice recognition application. With increasing use of these applications and increasing volumes of radiology studies being performed, these error rates may have significant negative impact on reporting accuracy and output, disproportionately affecting female radiologists. Causes of this discrepancy may include the differences in the volume and frequency of speech between genders, or more fundamental differences in how these applications were tested at time of development. Further work should investigate the causes of these errors with inclusion of larger sample sizes.
044. MIR-Radsearch: A Secure, HIPAA-Compliant, Google-Based Data-Mining Tool for Radiology Reports
Erinjeri J. P.*; Prior F. W.; Picus D.; Koppel P. Mallinckrodt Institute of Radiology, St. Louis, MO
Address correspondence to J. Erinjeri (joseph.erinjeri{at}mir.wustl.edu)
Objective: To develop a secure, web-based, HIPAA-compliant data-mining tool for radiology reports based on the Google search engine using free and open source technologies.
Materials and Methods: 20 months of radiology reports from the
Mallinckrodt Institute of Radiology (
915,000 studies, 2.8 gigabytes) were
downloaded in text format from our radiology information system (IdxRad) to a
fileserver running the Windows 2003 Server operating system. The reports (or
XML representations of the reports) were indexed using Google Desktop
Enterprise search engine software on the fileserver. In order to provide easy
intranet access to the index by multiple users across the department, we
deployed the Apache webserver on the fileserver, which serves an HTML form to
the users allowing submission of queries to the search engine over our
intranet. Google's XML response was interpreted by a script written in PERL,
and search results were presented to the users via a web browser. The query,
reason for search, search results, and documents visited are logged to
maintain HIPAA compliance. The search engine (Google Desktop), webserver
(Apache), and scripting language (PERL) are open source and/or freely
available.
Results: Indexing of the document took approximated 36 hours,
averaging 25,000 reports per hour. Benchmarks described include the time
to retrieve the 10 most relevant records for a given query. For example,
keyword search of a common term like "patient" yielded the first
10 most relevant results of 915,000 total matches in 0.72 seconds. Keyword
search of a less common term like "moderate cardiomegaly"
identified 7300 matches in 0.43 seconds. Keyword search of a rare term like
"hemangioendothelioma" identified 76 matches in 0.37 seconds.
Retrieval of all 76 matches from this query took 1.86 seconds. Search times
are more sensitive to the number of results requested and number of
simultaneous users, rather than the complexity of the query.
Conclusion: Radiology report data-mining tools like MIR-Radsearch can be used to improve the productivity of academic radiologists performing clinical, educational, research, and administrative tasks. By using the existing Google search algorithm and framework, radiologists can leverage their existing knowledge of the Google's interface, query, and relevancy ranking to quickly perform useful searches.
045. Identification of 3D Surface Reconstructed Facial Images and Implications for Patient Privacy and Security
Chen J. J.1*; Siddiqui K. M.2; Juluru K.2; Kim W.2; Safdar N.1; Siegel E. L.1,2 1. University of Maryland, School of Medicine, Baltimore, MD; 2. VA Maryland Health Care System, Baltimore, MD
Address correspondence to J. Chen (jchen1{at}umm.edu)
Objective: HIPAA requires safeguards to be in place "to ensure the integrity and confidentiality" of a patient's health information, including "full face photographic images and any comparable images." Although 3D surface reconstructed images from a head or facial CT study are not explicitly mentioned, these images have, by some, been considered to fall within the category of "comparable images." This could make it difficult to share these images in a teaching file, for clinical trials or for academic purposes. This study attempts to use patient photographs to determine whether a patient can be readily identified from surface reconstructed images of the face.
Materials and Methods: The study is a prospective, pilot study. The pilot includes three subsets of subjects, with a total of 90 participants. The first subset consists of patients already undergoing CT scans of the head, face, brain, or cerebral vasculature at the University of Maryland Medical Center (UMMC) or the VA Maryland Healthcare System (VAMHCS). Digital photographs and surface reconstructed images of their faces were obtained. The second subset consists of patients undergoing miscellaneous imaging studies at either UMMC or VAMHCS. Forty participants were enrolled with only their facial photographs digitally taken to create a library of facial photographs. The third subset consists of a group of patients who will be asked to match images of the surface reconstructed 3D facial features in the first subset to digital photographs of the individuals in both the first and second subsets of the study. Thirty participants will be enrolled to perform this matching.
Results: Data analysis will be performed to determine the ability of patients and radiologists to successfully match surface reconstructions of a patient's face with their photographs. Preliminary data suggest that subjects have difficulty matching pictures and surface reconstructed data. Detailed analysis will be performed at the completion of the study.
Conclusion: The recently expressed concerns over the inability to truly "anonymize" CT studies of the head/face/brain are yet to be tested in a prospective study. We believe that it is important to test, in a scientific fashion, whether these reconstructed images are indeed a "threat" to patient privacy/security. We plan to investigate techniques to obfuscate facial features both at acquisition time and with post processing techniques.
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
