Bittorf@derma.med.uni-erlangen.de, http://www.uni-erlangen.de/docs/derma/personen/bittorf.htm
N.C. Krejci-Papa, Department of Dermatology, School of Medicine, University of Erlangen, Germany
mfdk01@rrze.uni-erlangen.de, http://www.uni-erlangen.de/docs/derma/personen/krejci.htm
T. Ray, Hospital and Clinics, University of Iowa, USA
thomas-ray@uiowa.edu
A. Huntley, Department of Dermatology, School of Medicine, UC Davis; visiting scientist: Computing Center, Weizmann Institute of Science, Rehovot, Israel
ahuntley@wicc.weizmann.ac.il, http://huntley.weizmann.ac.il/about.html
D.C. Collison, Section of Dermatology, Dartmouth-Hitchcock Medical Center, Lebanon, NH, USA
dwc@dartmouth.edu
R. Drugge, Stamford, Connecticut, USA
rdrugge@netaxis.com
D. Hiller, Department of Dermatology, School of Medicine, University of Erlangen, Germany
M. Bergmann, Department of Dermatology, School of Medicine, University of Erlangen, Germany
Bergmann@derma.med.uni-erlangen.de, http://www.uni-erlangen.de/docs/derma/personen/bergmann.htm
T.L. Diepgen, Department of Dermatology, School of Medicine, University of Erlangen, Germany
Diepgen@derma.med.uni-erlangen.de
The work described in this paper is financially supported by the German Research Net Association (DFN-Verein) under the grant no. TK558 - RTB 03.1 - 3.5.
Slides from the archives of the Departments of Dermatology, University of Erlangen and University of California at Davis were digitized on Kodak Photo-CD (PCD). This source data was stored as the 'original image' in the JPEG File Interchange Format (JFIF) at a resolution of 768*512*24 and as 'thumbnail' 30*40 or 45*60 on hard disc. Different quality factors of the U-Lead JFIF implementation have been compared. The images were described with hierarchical keys. The information provided for every image was searched with a PERL script. A simple HTML forms-based query interface was developed using the Common-Gateway-Interface (CGI). Queries based on the same back-end have been included into 'generic' lecture scripts as push buttons. Different hardware platforms (Mac, HP, PC) and browsers (Mosaic and Netscape) were used. Performance tests have been done with users in Germany, Israel and US.
The PCD proved to be an inexpensive medium for input and archiving of images. It demonstrated to be superior in color and resolution to other methods we tested like a 1850 dpi 24bit-colors slide scanner or video. We compared various images at different quality factors on the screen as well as with a projection of slides produced by an exposer. We decided for a quality factor of 60 that corresponds to a compression of about 1:35 for clinical (n=200) and 1:40 for histological images (n=33). In most cases hardly any differences between the original and the compressed image could be detected. However, there are few images that appeared notably distorted even at a compression ratio of 1:15. Resolution had a strong impact on the compression factor achieved at a given quality level. At the quality factor 60 we achieved a compression factor of approximately 41 at 1536*1024*24 and approximately 35 at 768*512*24. The resolution of 768*512*24 proved to be sufficient, but the accurate display of colors is a specific problem to dermatology as they are very important for diagnosis. Large images were displayed using an image processing program as viewer that did the necessary gamma correction.
Hierarchical keys for Diagnosis (ICD9-compatible), localization of the lesion and others developed in Erlangen were used for the description of the images and stored linked to each image. Hierarchical keys proved to be convenient as the user can determine the required degree of accuracy. The uniform treatment of input fields provided by the CGI allowed the usage of a common PERL-backend for both the query interface and the dynamic lecture script. The dynamic lecture script was implemented using several forms in a document. It provided the search parameters for each query with the 'input type=hidden' HTML elements. The display times for images ranged from 4 seconds in a in-town high-speed network with a 486DX2-66 PC up to 90 seconds across the Atlantic Ocean with a Mac IIci at peak usage time. In the latter case a large amount of the time was needed for the reduction of the color depth and decompression; the transfer times accounted with up to 1s/Kb.
The objective of the project is to establish a dermatologic image database which allows multicentered storage of the images. International standards of dermatologic image-descriptions will allow to search and retrieve from the whole database. For improved transmission speeds European data might be mirrored at a center in the USA and vice versa. Online available dermatologic texts might use the links of the database for illustration.