| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
ABSTRACT |
Tuesday, May 2, 11:20 AM–12:30 PM
Abstracts 114–119
Moderators: William J. Weadock, MD and Walter Carpenter, MD
11:20 AM
114. Evaluation of the Mouse Count on Computers by Radiologists in a Typical Work Environment
Saksena M.A.1*; Braschi M.1,2; Hahn P.F.1; Mueller P.R.1; Harisinghani M.G.1,2; 1. Radiology, Massachusetts General Hospital, Boston, MA; 2. Radiology, Center for Molecular Imaging Research, Boston, MA.
Address correspondence to M.A. Saksena (msaksena{at}partners.org)
Objective: With the move toward PACS reporting systems, radiologists required to carry out most tasks using the common computer mouse. Not only does the radiologist have to use the mouse device to operate the PACS and ancillary systems such as speech recognition and hospital record systems but in addition has to use the mouse during leisure and personal computer use. There is a known association between the mouse counts (the number of times per day a user clicks the mouse) to condition such as wrist pain and forearm in various studies. The purpose of this study was to evaluate the mouse count on computers used during daily reporting in an average radiology department.
Materials and Methods: Freely available software (mouse count, version 1.2a) with ability to detect and record mouse counts was installed on 5 computer systems with varying use in the abdominal imaging division within a large tertiary care hospital. The same software was also installed on the computers used by 3 secretaries aiding in daily workflow within the same department. Daily counts were recorded over a period of 7 days.
Results: The mean daily mouse count on the 5 computers used exclusively by radiologists was 826.8 (range, 460–1.110). The mean daily mouse count on computers used by the secretaries was 1.152 (range, 882–1.462). The mean weekly mouse count on computers used by radiologists was 12.845.6 (11.395–12.337). The highest mean count rate was recorder on the computer used for reading and dictating diagnostic CT exams.
Conclusion: The mouse click rate for radiologists in a busy department equals that of an office desk top in a secretary's office. It is important that radiologists realize the repetitive trauma they undertake and make a concerted efforts minimize the use of the mouse in order to avoid stress injuries related to increased mouse use.
115. Making Online Radiology Teaching Files-A New and Unique Way of Making and Storing Radiology-teaching Files on the Internet
Bagga S.; Radiology, Tufts-New England Medical Centre, Boston, MA.
Objective: Web based learning programs are playing a major role in information and knowledge dissemination. The ready availability of Internet and broadband connections has made it possible for radiologists to access images on the internet very easily. While some academic institutions have dedicated information technology staff helping the radiologists in making and maintaining online teaching files, most radiologists feel handicapped by their lack of knowledge about the computers and the internet. Dependence on information technology staff takes away some of the fun and autonomy from this creative process. This project was conceived to make it very easy for any radiologist with a computer, access to the internet and interesting images and a desire to create online teaching files with complete autonomy.
Materials and Methods: Interesting radiologic and nuclear images were first saved on a portable flash drive. A word document was then created, a brief relevant clinical his-tory of the patient was provided. The pertinent imaging findings along with any correlative images were included. A brief discussion about the findings, clinical and pathologic correlation was done. The radiologic images were imported from the flash drive on the computer using Picasa, a free imaging and editing software available on Google. A blog name of Nukesdoctor was created. All these images were then blogged using this software and following the online instructions. The word document about the findings and the clinico-pathologic correlation was cut and pasted on the blog.
Results: The entire process of downloading the images to the computer using the Picasa software was very simple and intuitive. The images could be easily cropped and edited and blogged using the same software without any problems.
Conclusion: The creation of this blog for the purpose of creating online teaching files was a very simple, intuitive and easy process. It can give a lot of academic satisfaction to the creating radiologist with no need for any help from information technology personnel, thus proving complete autonomy and creative independence to the radiologist.
11:40 AM
116. A Visual Query Interface for Assisting in Decision Support of Tumors Using Image Findings Structured by Bayesian Networks
Hsu W.1*; Dordoni A.1; El-Saden S.2; Bui A.2; 1. Biomedical Engineering IDP, University of California, Los Angeles, Los Angeles, CA; 2. Radiology, UCLA Geffen School of Medicine, Los Angeles, CA.
Address correspondence to W. Hsu (willhsu{at}mii.ucla.edu)
Objective: Diagnostic imaging plays an important role in characterizing solid tumors. Presence of certain image features such as contrast enhancement, edema, and multifocality has been shown to correlate with patient outcome and survival rate. However, physicians lack tools to effectively employ these imaging features in assessing the diagnosis, prognosis and outcome of a patient. This project presents an interface that allows visual querying of an underlying causal disease model through the manipulation of graphical metaphors. The model uses a Bayesian belief network (BBN) to structure evidence variables derived from image features appearing in MR scans of an abnormal region.
Materials and Methods: Evidence variables are organized semantically into different levels (e.g. organism, organ, region of interest [ROI]). These levels specify the system's schema and dictate when and how a metaphor for a particular variable is presented to the user. An iconic manipulation toolkit provides the necessary functionality to overlay a tumor with user-defined features on a canonical image of the organ of interest. The toolkit is context-sensitive, adapting metaphors based on the defined level. Descriptors defined in the visual query such as spatial, morphological, and geometrical relationships are automatically extracted and translated into a maximum a posteriori (MAP), maximum probability explanation (MPE), or counterfactual query. The BBN results in turn influence the generation of a simulated MR image that depicts how the lesion, surrounding structures, and prognosis would change given the hypothesized characteristics.
Results: A preliminary prototype has been developed and applied to answering questions related to prognosis for a brain tumor patient given a set of parameters. Visual metaphors present an effective method for conveying spatial and morphological information in comparison to a textual query format. The system methodology encourages a mode of interaction that appeals to the visual nature of human thinking.
Conclusion: A visual interface enhances the utility of diagnostic imaging by improving the ability of physicians to understand, analyze, and act rationally upon collected data. This project presents a generalized framework for exploiting implicit morphological and geometrical descriptors in graphical metaphors to query a complex model comprised of image features. Not only is this framework applicable to solid tumors, but any disease specified using the proposed semantic organization.
117. Design of an Automated Follow-up Application for Radiology
Boonn W.W.1*; Vandermeer P.3; Siddiqui K.M.2; Siegel E.L.3; Langlotz C.P.3; 1. Department of Radiology, University of Pennsylvania Health System, Philadelphia, PA; 2. Department of Radiology, VA Maryland Health Care System, Baltimore, MD; 3. Department of Radiology, University of Maryland Medical Center, Baltimore, MD.
Address correspondence to W.W. Boonn (william.boonn{at}uphs.upenn.edu)
Objective: The results of radiologist's recommendations, including validation of imaging findings, are essential for continuing education and performance improvement. Verification of receipt of notification of critical or unexpected findings is both crucial to patient care and an integral part of standards of practice. The purpose of this study was to evaluate the needs for, design, and implement a prototype automated, Web-based application to track radiology cases for follow-up.
Materials and Methods: (1) A Web-based survey was validated and distributed to radiologists and radiology fellows and residents at our institution, with 59 responses received. The results after data analysis confirmed that follow-up is performed manually and on an ad hoc basis (if at all). Data indicated that 42% of radiologists perform follow-up on "all questionable or interesting cases," whereas 46% perform follow-up only "when I remember and have time." Handwritten lists are kept by 4%, whereas 26% use PDAs and 14% use a PC database. Manual look-up of reports for each patient was reported to discourage radiologists from routinely pursuing feedback. (2) We developed an automated, Web-based prototype application to track cases for follow-up. Underlying the system is a Web portal system that provides clinical history to the radiologist at the time of interpretation, linking the radiology Web server, the RIS and HIS, and multiple clinical information systems. A trigger in this application is defined as a clinical event, imaging study, or laboratory/pathology report. Elapsed time also can be used as a catch-all trigger. Notification can be made within the radiology portal or by e-mail or pager, and an HL7 listener monitors transactions across the hospital network. This prototype system was activated for trial use.
Results: The prototype system, with a simple user interface, has met with initial enthusiasm. At the time of image interpretation, the radiologist can select a case for follow-up, along with triggers, notification, and comment options. The patient is then added to the radiologist's automated follow-up list. Initial results indicate that > 95% of users prefer such a system to previous methods and that it is effective in alerting them to studies that remain to be followed.
Conclusion: With increasing requirements for communication of findings to clinicians, an automated application such as this prototype may help to "close the loop," ensure that recommended studies have been performed, and provide a valuable follow-up record.
118. Wiki-Based Collaborative Web page for Residents
Garg N.1*; Chang A.2; 1. Radiology, St. Vincent Hospital, Worcester, MA; 2. Medical Student, University of Massachusetts Medical School, Worcester, MA.
Address correspondence to N. Garg (naveen.garg{at}gmail.com)
Objective: The web has become an important tool in the education of radiology residents. However, on most radiology portals used, many links are broken or outdated, or are simply irrelevant to what you are looking for. We have created a collaborative online webpage for radiology residents that allows residents to modify and contribute to it. It will serve as a central repository of hyperlinks, a database of radiological references. Users of the site are allowed to modify both the organization of the data and the content itself. As a result, the wiki will constantly represent the latest and most relevant links through dynamic contribution.
Materials and Methods: We have used "PHP Wiki," an open source tool to create and moderate wikirad. It contains hyperlinks to educational resources categorized in two ways: one by ACR coding standards by organ system, and another by modality. A few links were seeded by the wiki moderators, and many more links were then submitted by other residents in the program and on the world wide web. The address of the wiki is http://www.wikirad.svhrad.com. We compare it to two traditional radiology web portals: 1. http://brighamrad.harvard.edu/Links.html - top link in a google search for "radiology department links." 2. http://www.auntminnie.com/index.asp?Sec=lin&Sub=def - (aunt minnie is the top link in BrighamRad under "Portals and E-Publications.") Criteria compared are: a. Organ systems and modalities represented and categorized in a useful manner. b. Number of broken links c. user satisfaction among radiology residents.
Results: The wiki had organ systems represented by anatomical location and they were numbered according to usefulness. BrighamRad had no sub-categorization under educational resources by anatomical area. Aunt Minnie was categorized by "subspecialty." But it was more than 2 link depths away from the main portal, and they were not numbered by usefulness. Broken links were quickly fixed by users of the wiki, whereas BrighamRad had 2 out of 15 links that were broken and were not fixed. Aunt Minnie had too many irrelevant links that were as bad as broken links. Radiology residents reported higher satisfaction levels the wiki.
Conclusion: Wiki is a better way for radiology residents to organize their collective educational efforts on the web. In the future, with the minimum amount of resources and maintenance, one can consolidate the open source radiology educational content on the web with a wiki.
119. Bayesian-Filtering for the Categorization of Radiology Reports
Pyrros A.T.1; Nikolaidis P.1*; Yaghmai V.1; Zivin S.1; Flanders A.2; Tracy J.I.3; 1. Department of Radiology, Northwestern University Medical School, Chicago, IL; 2. Department of Radiology, Thomas Jefferson University, Philadelphia, PA; 3. Department of Neurology, Thomas Jefferson University, Philadelphia, PA.
Address correspondence to A.T. Pyrros (apyrros{at}radiology.northwestern.edu)
Objective: The objective of this study was to develop a Bayesian-filter that could distinguish positive computed tomography radiology reports of appendicitis from negative reports of appendicitis.
Materials and Methods: Standard unstructured electronic text radiology reports containing the keyword appendicitis were obtained using a java based text search engine from a hospital PACS system. A total of 500 reports were selected, from multiple radiologists, and then manually categorized and merged into two separate text files, 250 positive reports and 250 negative findings of appendicitis. The two text files were then processed by the freely available UNIX based software dbacl 1.9, a digramic Bayesian classifier for text recognition, on a Linux based Pentium II system. The software was then trained on the two separate merged text files categories of positive and negative appendicitis. The ability of the Bayesian filter to discriminate a priori between reports of negative and positive appendicitis reports was then tested on 100 randomly selected reports of appendicitis, 50 positive cases and 50 negative cases.
Results: The training time for the Bayesian filter was approximately two seconds. The Bayesian filter subsequently was able to categorize 50 out of 50 positive reports of appendicitis and 50 out of 50 reports of negative appendicitis, in under 10 seconds.
Conclusion: A Bayesian-filter base system can be employed to quickly categorize radiology report findings and automatically determine after training, with a high degree of accuracy, whether the reports have text findings of a specific diagnosis. The Bayesian-filter can potentially be applied to any type of radiologic report finding and any relevant category.
* Will present paper
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
