| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Computers in Radiology |
Department of Radiology, Indiana University, 714 N Senate Ave., Ste. 100, Indianapolis, IN 46202.
Received March 26, 2004;
accepted after revision July 1, 2004.
Address correspondence to M. S. Frank.
Abstract
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionConclusionReferences |
|---|
CONCLUSION. We believe this approach provides a feasible solution for converting radiologists' educational 35-mm slides into well-organized, high-quality digital media suitable for both online education and speaker-led presentations.
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionConclusionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionConclusionReferences |
|---|
|
Typically, we digitize 20-40 slides at a time. After considerable
experimentation that included digitization at up to 2,500 x 1,600
pixels—with the results empirically assessed by two radiologists, our
production assistant, and our information technology manager, we chose to
digitize at a resolution of 1,250 x 850 pixels (resolution, 
907
pixels/inch) rather than higher-resolution alternatives. The digitizer
requires approximately 30 sec per slide at this resolution. Digitizing a
single batch of slides can therefore take up to 20 min. We found that
optimizing contrast resolution was more important than spatial resolution when
using a digitization strategy. Sixteen-bit gray-scale TIFF was chosen because
of its gray-scale range and the capabilities of the accompanying
image-processing software (NikonScan, Nikon) to effectively manipulate such
images. Digital images are thus initially generated as 1,250 x 850
pixel, 16-bit gray-scale TIFF files. These files are individually reviewed
and, when necessary, are manipulated by our production assistant using Nikon
Scan software, and then are batch-converted into JPEG format using a quality
factor of 90 or higher. The adjusted 16-bit TIFF images, each approximately 2
MB in size, are then archived to optical media for long-term safekeeping. We
do not digitize text slides.
We found that the Nikon Scan software has some ideal features for quickly adjusting the original TIFF images. The production assistant uses this software on her desktop computer, which has a brighter and higher-resolution display than does the acquisition computer. While she is reviewing and manipulating one batch of newly digitized slides via network connectivity to the acquisition computer, it can simultaneously be digitizing the next batch. Some images initially produced by the digitizer are either too dark or too light, but these images usually can be adjusted with the Nikon Scan software in a matter of seconds to produce a suitable image. The software provides a graphical contrast-adjustment curve that can be manipulated in real time with a click-and-drag operation. Usually, this feature is effective at quickly improving an undesirable gray-scale distribution in newly digitized images.
After a batch of 35-mm slides is converted to 1,250 x 850 pixel images in JPEG format, the production assistant enters these images and any available descriptive or diagnosis-related information into our enterprise teaching file system [1]. These minimally compressed JPEG images serve as "golden copies" from the perspective of the radiologist. If available, the history, findings, and diagnosis for each "case" represented by one or more 35-mm slides are entered into the system. Usually, however, only the diagnosis for a case is available, typically having been written somewhere on the slide frame. The teaching-file software facilitates batch input of image files using conventional drag-and-drop and copy-paste operations, and it also has a built-in tool for quickly designating an American College of Radiology diagnostic code for a case. Among the system's image-manipulation features we use are batch-mode generation of ancillary JPEG images that are of more appropriate dimensions and file size for use on the World Wide Web; automated placement of a copyright notice on these down-sized images; and when requested by the slide owner, automated generation of a PowerPoint (Microsoft) presentation from the digitized slides.
We store each person's digitized slides in his or her personal teaching-file repository on our enterprise teaching-file server. The server hosts several dozen such repositories, each Web-enabled. The cases can then be searched and their images retrieved by their owner using a Web browser via password-protected access to his or her personal teaching-file on our Intranet. Several radiologists also permit the inclusion of their content in a department-wide consolidated index—still password-protected, but permitting all radiologists to search, which is also built into our teaching-file system. Once content has been entered into a personal digital repository in our system, the owner of that repository and any authorized designees can use the system's content-authoring software to update, embellish, and access and use any content originally entered by the production assistant.
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionConclusionReferences |
|---|
Probably the largest hurdle initially was to understand the different ways in which the digitizer can be configured and how to optimize its operation to consistently and efficiently produce good digital images from radiologic slides. It quickly became evident that digitizing at a resolution of greater than approximately 1,000 x 1,000 pixels did little to improve image quality yet substantially increased the time required for digitization and the noise in the digitized result. Findings consistent with our observations have been reported when digitizing analogue radiographs for conversion into 35-mm slides [3].
A greater challenge for us instead lay in obtaining the best possible contrast resolution. Fortunately, the digitizer provides an adequate dynamic range of optical density when in 16-bit gray-scale mode to generate a TIFF image with contrast resolution reasonably true to that of the original slide. Our tests revealed an actual optical density range of approximately 3 (corresponding to 10 bits of depth). This range was not always well represented in the digitizer's automated generation of its output image. However, we found that if the 35-mm slide appeared to be of reasonable quality, the Nikon Scan software could almost always be used to quickly transform the original image to an entity suitable for conversion to a good to excellent JPEG image. We consider the Nikon Scan software to be a critical component in the project's overall success.
We found that having a flexible work-flow model is helpful for accommodating the different types of slide collections and the ways in which they are stored, catalogued, and submitted by their owners. For example, some radiologists file their slides in case-centric collections, whereas others organize slides primarily as lecture-based collections. Some radiologists use transparent plastic mounting sheets in a threering notebook, others use drawer-based filing systems with their slides essentially in stacks, and others use slide carousels.
Many radiologists write the diagnosis pertaining to a slide (often cryptically) on the slide mount with no other information made available. Sometimes, only one in a series of several related slides will be labeled. Although this is usually no problem for a radiologist, it can at times be vexing for the production assistant. Some radiologists submit entire lectures to each be digitized and filed as single entries (i.e., sets of perhaps 40-80 images) to be culled and organized by the radiologist later. Regardless of how the slides are submitted to the production assistant, we try to restore them to their original location, order, and orientation. This is not always easy once they are removed from their original container, rotated when necessary, and loaded appropriately for batch-feeding through the digitizer. Because the digitizer requires that each portrait-oriented slide be rotated by 90° within its batch to be digitized without undesirable cropping, considerable attention is required to later return slides to their correct order and orientation in their original container. The situation is further complicated by the need for the production assistant to track and "de-rotate" such images once in digital format.
We chose the Nikon digitizer for two main reasons. It has the best range of optical density we could find in an acceptably priced digitizer, and we made it a requirement that our system have an automated batch-mode slide feeder. In general, the batch-feeding mechanism on the Nikon digitizer performs acceptably, but it is prone to jam in some circumstances. For example, warped slides can cause a jam. Slide collections that are composed of slides with mounts of widely varying thickness are more likely to jam. Also problematic are the slides on which a stickon label has been placed. Over time, these labels become brittle, the adhesive deteriorated; sometimes the mere act of picking up the slide will cause the label to partially or completely separate from the slide. A partially separated label creates an unwanted edge that substantially increases the risk of a jam.
We found that a quick and informal "rub test" will usually weed out slides that are likely to be problematic: If rubbing the edge of the label causes separation in even the slightest amount, the label will be completely removed before digitization. Because removing the label also removes the slide's identifying information, we experimented and developed several commonsense techniques to track this information during the digitization process. Most often, simply writing the same information directly onto the slide frame after removing the label is satisfactory to both the production assistant and the slide owner.
The overriding factor that has made this project worthwhile is that good to excellent digital images can be produced and catalogued reasonably efficiently. Several thousand valuable educational images have been salvaged from an archaic medium and transformed into an enduring and highly available format. Most of our radiologists think that the images are more than adequate for lectures, conferences, and online use and, in many cases, are also suitable for publishing in journals and textbooks. Of the 11 radiologists who have submitted at least 100 slides, one said that she expected higher image quality, and the others expressed satisfaction (and in some instances, glee) with their content now in digital form.
Our PACS is now our primary source of new digital educational content, and it is likely that images from it will gradually supersede most digitized hard-copy images (derived from both radiographic film and 35-mm slides) in our digital teaching file system today. However, this transition will take years, and we believe that considerable value remains in the thousands of familiar, meticulously produced 35-mm slides already in the hands of our radiologists. Many slides are excellent examples of a specific entity, and some depict rare findings. Many of the digitized 35-mm images are good enough to obviate recollecting similar cases from the PACS, thus saving our radiologists' time.
We have learned that not all 35-mm slides are good candidates for digitization. For example, some slides seem to be indelibly grimy. Others have been hopelessly scratched. Still others have an extremely limited contrast latitude. We found it essential to remove dust from slides before digitizing them. The adage "garbage in, garbage out" applies well here. If a slide looks marginal or poor when projected onto a screen, it typically is not a good candidate for digitization. Even so, some slides that represent a once-in-a-lifetime example are still prized in digital format by the radiologist even if image quality is not ideal.
Digitized images are stored in an enterprise database system and are therefore conveniently available for many purposes. For example, our teaching-file system facilitates creation of content suitable for independent study such as interactive case-based exercises and thematic digital curricula. The system also facilitates speaker-led conferences by enabling the radiologist to automatically create PowerPoint slides in a variety of formats using images in the database.
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionConclusionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionConclusionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  G. Gunabushanam and S. Sharma Simple Method to Eliminate Camera Reflections in Digital Photographs of Single-Emulsion Films Am. J. Roentgenol., January 1, 2007; 188(1): W96 - W97. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
