Section 3 -

Virtual Microscopy as Future Tool in the Workflow of Diagnostic Pathology Manfred Dietel, K. Schluns, K. Saeger, T. Schrader, P. Hufnagl Institute of Pathology - "Rudolf-Virchow-Haus" Charité Medical Faculty, Humboldt University Berlin D-10098 Berlin

The routine use of digital images for diagnosis is largely performed in radiology using diagnostic instruments with digital image output and PACS systems. Digital routine pathology with a PACS system and a virtual microscope as user front-end also has the possibility to bring decisive advantages in functionality for the diagnosis, workflow, time management, cost effectiveness and quality assurance [1, 2]. Thinking of functionality, a critical view on the light microscope bares several constraints, such as: Only a clipping of the glass slide is visible, orientation in the slide is difficult. Only selected, lens-dependant magnifications are available. Different stains can only be visualized one after the other; overlay to compare reac-tions cell by cell is not possible. Generally only one, at the maximum a few local users time can read the slides at the same. Annotations on the microscopic layer itself are difficult and unprecise. Glass slides are unique and can not be copied. No integration is possible into the digital workflow. In accordance conventional telepathology based on static images has several fundamental limitations in particular in the approach of primary diagnostic work where the pathologist should be able to continuously view the whole slide at different magnifications similar to a normal microscope. Thus the development of digital or virtual microscopy is gaining more and more interest and attraction. Past approaches to develop software for histology diagnosis on a monitor (virtual microscope [3]) can be divided into two main groups: They are either based on a remote-controlled robotic microscope including a digital camera and transmittance of live images over the internet. This technique originates from telepathology projects. One example is the early work by Wolf et al. [4]. The other approach utilizes digital histological images, resulting from partially scanned glass slides, mostly used for teaching [5, 6]. Making completely digitized histological slides available over internet is still technically highly advanced. Such a slide at 400x magnification results in up to 6 GB of image data. Total images of that size can not be transmitted to a client over the Internet [7]. The solution is a process termed "image streaming" allowing real time visualization of large images via the Internet without transmitting the whole image to the client [8]. We designed and developed a virtual microscope system that fulfils the high demands on digital routine pathology. Requirements for acceptance of such a system are image visualization in real-time, functionality superior to a light microscope, acceleration of the diagnosis. Our approach is termed "DVM" for "digital virtual microscope". For this challenge the diagnostic process in histopathological diagnosis is analyzed, and the important figures of a virtual slide, a digital microscope and the individual case are outlined as well as the necessary hardware and software. According to this analysis, the possibilities of the new technology are promising, and obvious disadvantages will become negligible with increasing bandwidth, processor speed and memory space. The use of digital images for diagnosis, as known from radiology, can bring substantial improvements for efficiency and functionality in routine histology. We developed a system that allows diagnosis with a virtual microscope using digital images from complete slides under highest magnification. The major technical challenges were the large data volumes and the necessary visualization speed. Using image streaming technologies, we developed a system that solves those problems. We present our system and its advantages. Fast access to reference cases, easy integration of telepathology into hospital information systems and an uninterrupted digital workflow may, in the long term, convince pathologists of the DVM as a realistic alternative to the light microscope. The Digital Virtual Microscope (DVM) System The virtual microscope system consists of three tiers: A slide scanner that digitises whole glass slides in highest magnification (at least 400x), a PACS system or database that can handle the virtual slides, which are up to 6 GB large (uncompressed), and the user front-end, the DVM itself. We mainly scoped on the development of the Virtual Microscope. DVM was written in HTML, ASP, JavaScript and VBScript. The development platform was Microsoft® Visual Interdev 6.0. The virtual microscope user interface (client) runs within Microsoft® Internet Explorer. DVM Software Figure 1 illustrates the DVM system. Glass slides are scanned with a slide scanner and stored on a PACS server. A pathologist can visualise and diagnose the virtual slides with the web-based DVM software, which accesses the virtual slides using image streaming. For the user front-end we integrated a specialised internet based visualisation tool from the remote sensing industry, the "Image Web Server" (Earth Resource Mapping, Perth, Australia). This software and its file format ECW (Enhanced Compressed Wavelet) support high compression and a very fast image data retrieval. Combining it with the functionality needed for a Virtual Microscope we achieved a tool that allows very fast image access. The technical advantage lies in the sole transmission of the actually displayed sub-image (in magnification and region) without transmitting the whole image. Digitalization For testing our DVM user front-end, we scanned histological slides using a Leica® DM RXA microscope. A Hitachi 3 CCD colour camera HV-C20M was instrumental in order to scan sub-images. We developed a software using Microsoft® Visual C++ as development platform to control the microscope stage movements using the Leica® DM SDK and the camera using Matrox® Meteor II framegrabber. Sub images were scanned by moving the microscope stage in meander lines and scanning the images one by one until the whole specimen had been covered. The software integrates the "ER Mapper" software by "Earth Resource Mapping", Perth, Australia, to defragment the sub-images and compress the image into ECW format. PACS The PACS for the digital histology routine diagnosis [9] also has presumably higher demands than in radiology. The images are stored on a developing server containing two Intel Pentium III 1.5 GHz processors and two 80 GB hard drives. The server runs the Microsoft® Windows 2000 Server software. For the database we use a Microsoft® SQL server database. Images for open access within the hospital can also be stored on the Charité PACS system "Marvin", which enables delivering images by DICOM or HTTP. For an average pathology institute with 5 pathologists, 35.000 cases per year, 5 glass slides per case and 135 MB file size per average virtual slide (80% biopsies, 20% ectomies) a server with 450 GB memory could handle the data of five days. The virtual microscope solution would therefore imply a 2 TB server with short access times for actual cases of one month and a 250 TB server with low access times as long term archive for 10 years. while only the sub-images (in region and magnification) are being transmitted. Data Volume The data volume of a single virtual slide depends mostly on tissue size and magnification. Table 1 shows the files sizes for virtual slides, uncompressed. We calculated the file sizes for a target resolution of five pixels per mm in each direction and a colour depth of three bytes per pixel. An average histological slide (500 mm²) with a maximum magnification of 400x results in a file size of 5.5 gigabytes, if compressed with a ratio of 1:15 in 380 megabytes. Image Streaming The images are transmitted from server to client using the image streaming technology included in the "ecw" protocol by the "Image Web Server" (Earth Resource Mapping, Perth, Australia). The principle of image streaming is that the entire image remains on the server. Table 1: Data volumes of virtual slides in megabytes, uncompressed tissue size (mm2) Magnification 10 25 50 100 200 400 100 0,72 4,47 17,88 72 286 1144 200 1,43 8,94 35,76 143 572 2289 300 2,15 13,41 53,64 215 858 3433 400 2,86 17,88 71,53 286 1144 4578 500 3,58 22,35 89,41 358 1431 5722 600 4,29 26,82 107,29 429 1717 6866 700 5,01 31,29 125,17 501 2003 8011 800 5,72 35,76 143,05 572 2289 9155 900 6,44 40,23 160,93 644 2575 10300 1000 7,15 44,70 178,81 715 2861 11444 DVM Functionality Figure 2 shows the DVM user front-end. The upper left image acts as an overview image, indicating the actual region visualized on the large image. The lower left screen area consists of various control tools and information screens. The DVM offers real time browsing and zooming of complete virtual slides with a high variability, multilayer visualization of a slide in several different stains and the setting and using of annotation markers. Furthermore the DVM system facilitates the archiving and enables fast access to reference cases. Applications in Education For integrated web-based, digital histology teaching an advanced variant of the "virtual microscope" is necessary. For the University of Berlin a Client-Server Software consisting of digitalisation facility and server software integrating image streaming technology as well as user front-end ("Digital Virtual Microscope") was designed, called Meducase [14]. Among others problem-based learning was one of the aims to be reached. The digitized histological glass slides (6 GB compressed and an average of 225 MB) with a targeted magnification of 400-fold were processed. These images can be visualized close to "real-time" via internet using image streaming. The "digital virtual microscope" (DVM) as the user front-end described above allows for different types of didactics and enables the user to easily handle highly complex and large histology images supplemented with clinical data. The DVM offers new functionalities like an "observation path" and an UMLS-based (Unified Medical Language System) context search. Discussion The Virtual Microscope overcomes many of the known restrictions of light microscopy and of static telepathology. It additionally offers a great deal of functionality supporting the complex diagnostic workflow. The DVM may therefore increase the efficiency and facilitate the work of histopathologists. It further will contribute to the international standardization of the diagnostic process since it opens access to a discussion platform for more or less all pathologists around the world. This might become important in particular for pathologists from the developing countries as difficult cases can easily be exchanged in a cheap and time saving manner. An already world wide functioning system has been introduced by the International Union Against Cancer (UICC) in cooperation with the Institute of Pathology in Berlin [15]. A critical aspect for the employment in routine diagnostic is a fast slide scanner. We estimate a scanning time of one minute per slide as being sufficient in order to handle the daily throughput of an average pathology institute. Slide scanners are more and more available on the market [10, 11]. The actual scanning times for 400 x magnifications from known slide scanning systems are around 20 minutes, but the technical evolution is proceeding. Although the computer might primarily be an uncommon tool for routine diagnosis, many experiences make us optimistic that the advanced functionality will convince pathologists in the long term. The possibility to access a reference case data base very quickly in order to compare cases and the uninterrupted digital workflow are desired improvements. Additionally, telepathology is, in consequence of the web based solution, always possible without additional prerequisites [12, 13]. All cases are always accessible and the archive can act as case database for research and teaching [9]. Furthermore the DVM is capable to improve handling and efficacy of virtual slides for web-based histology teaching. Although the virtual microscope can not and should not substitute a light microscope, it proves to be a valuable tool for teaching that is highly accepted by students [14]. A major challenge of VM technology is the integration into the information systems of institutes of pathology and hospitals. Thus standardization of the software will become a crucial issue to make the technology compatible with the "digital environment" often varying greatly from institution to institution. In particular in so called specialized interdisciplinary centers, like breast centers (Fig. 3) or colon centers, the linkage with other specialties such as gynecology, radiology, surgery etc. will contribute to establish faster and more efficient work flows of diagnosis and treatment. References: Barbareschi M, Demichelis F, Forti S, Dalla Palma P: Digital Pathology: Science Fiction? Int J Surg Pathol, 8, 261-63, Oct 2000 Furness PN: The use of digital images in Pathology. J Pathol, 183, 253-63, 1997 Ferreira R, Moon B, Humphries J, Sussmann A; Saltz J, Miller R, Demarzo A: The Virtual Microscope. Proceedings of the AMIA Annual Fall Symposium, 449-53, 1997 Wolf G, Petersen D, Dietel M, Petersen I. Telemicroscopy via Internet. Nature, 1998;391;613-14 Heidger PM, Dee F, Consoer D, Leaven T, Duncan J, Kreiter C. Integrated Approach to Teaching and Testing in Histology With Real and Virtual Imaging. Anat Rec. 2002;269;107-12 Klatt EC, Dennis SE. Web-based Pathology Education. Arch Pathol Lab Med. 1998;122;475-79 Hamza S, Anderson P, Reddy V, Siegal, GP. Use of the Internet in Pathology Training and Education. Anat Path. 2001;8; 290-97 Harris T, Leaven T, Heidger P, Kreiter C, Duncan J, Dick F. Comparison of a Virtual Microscope Laboratory to a Regular Microscope Laboratory for Teaching Histology. Anat Rec. 2001; 265; 10-14. Demichelis F, Della Mea V, Forti S, Dalla Palma P, Beltrami CA : Digital storage of glass slides for quality assurance in histopathology and cytopathology. Journal of Telemedicine and Telecare, 8, 138-142, 2002. [Ref ID 30} Felten CL, Strauss JS, Okada DH, Marchevsky AD: Virtual Microscopy: High Resolution Digital Photomicrography as a tool for Light Microscopy Simulation. Hum Pathol 30 477-83, 1999 Leong FJ, McGee J: Automated complete slide digitization: a medium for simultaneous viewing by multiple pathologists. J of Pathology, 195, 508-514, 2001 Petersen I, Wolf G, Roth K, Schlüns K: Telepathology by the Internet. J Pathol, 191, 8-14, 2000 Okada DH, Binder SW, Felten CL, Strauss JS, Marchevsky AM. „Virtual Microscopy"and the Internet as Telepathology Consultation Tools. American Journal of Dermatopathology, 21, 525-531, 1999. The Meducase project. URL: http://www.meducase.de, 2003 Dietel M, Ngyuyen-Dobinsky TN, Hufnagl P: The UICC-Telepathology Consultation Center . Cancer 2000, 89:187-191.