Nick Hawkins, Lance Smith, July 1995
Abstract
Introduction
Materials and Methods
Discussion
Acknowledgments
References
We describe the design and implementation of a large interactive database of images of diseased tissues. Images of macroscopic, microscopic, radiographic and clinical aspects of human disease have been acquired, converted and stored for use in the database. Image identifiers, together with relevant textual information, have been indexed in a relational database format, to facilitate effective and flexible retrieval of data. A graphical user interface has been constructed to simplify data access. A formative evaluation of the project is presented, together with an outline of ways in which the database may be further developed as a means of improving undergraduate education in the discipline of pathology.
The discipline of Pathology is concerned with the understanding of disease and disease processes, based largely on the analysis and interpretation of visual images. These images, either of diseased organs (macroscopic pathology) or thin slices of diseased tissue viewed under the microscope (microscopic pathology, histopathology), play a vital role in the teaching of this discipline. The School of Pathology at the University of NSW, like many other institutions involved in the teaching of pathology, has actively sought to collect, catalogue and display a large number of macroscopic specimens in its pathology museum, which now contains over 2500 specimens. The School also possesses microscopic sections of a wide variety of diseases, as well as an extensive 35 mm slide collection which highlights the radiographic and clinical features of disease. These various images represent the visual teaching resources of the School, and are central to the learning process in pathology.
Over the past decade there have been several reports of the compilation and use of libraries of pathological images (1-6). With one exception (5), they have involved the use of laser discs and players, which are stand alone analogue devices for the storage of "television" quality images. Frequently, they have consisted of collections of images that are specific for certain computer assisted learning modules, although at least one extensive database of images has been produced (2). Recent advances in microcomputer power and availability, and in particular the use of graphical user interfaces, has meant that it is now practical to use microcomputers for the storage and retrieval of these visual teaching resources in digital format. Importantly, this computer technology offer the possibility of greatly enhancing the educational benefits of this information. In this paper, we describe the establishment an extensive image database based on these visual teaching resources, and the software we have developed to enhance its use as an educational tool.
Museum specimens were photographed under suitable lighting conditions using a Nikon F801 camera and Kodak Ektachrome 100 transparency film. Transparencies of suitable quality were transferred to PhotoCD format using a commercially available transfer service (Kodak Australia).
Individual PhotoCD images were converted to three separate image sizes using the Photostyler 2.0 image manipulation program (Aldus). Firstly, each image was cropped to a maximum dimension of either 1024 pixel (width) or 768 pixels (height), depending on its shape, and stored as a 24 bit colour image in Windows bitmap format (high resolution image). The images were then transferred to a CD-ROM disk using a CD-ROM writer (Philips CDD521) for long term storage. In the second stage, each cropped image was placed on a black 1024 x 768 pixel background, and the whole image then resized to 640 x 480 pixels and compressed using a JPEG algorithm (1:15) to produce a working image. Finally, the working images were converted to 120 x 96 pixel postage stamp images. The last two image conversions were amenable to automation, and thus could be performed through the use of keyboard macros, thus significantly reducing the time involved in image manipulation. The working and postage stamp images were stored on a networked hard disk, and later on a CD-ROM, for use with the database.
Where the colour or focus of an image was considered suboptimal, appropriate corrections were made. In general this was necessary only with histopathological images, and with surgical and clinical images from existing 35 mm transparencies.
The database was constructed with Access 2.0 (Microsoft Australia), a relational database program running under the Windows operating environment. It was designed to reflect the centrality of images to the program. Text-based clinical, macroscopic, and microscopic information regarding the museum specimens was imported from existing museum catalogue files. For each image, links were established to the one or more diseases apparent in the image, and to information regarding museum specimens where appropriate (Figure 1).
The interface for the program was designed using Visual Basic 3.0 (Microsoft Australia), and was based on the principle of using visual objects, rather than text, to access data. An important part of the interface is the use of the array function of Visual basic to allow the presentation of the results of a search as clickable objects on screen. This is useful in maintaining the visual and realistic nature of the database.
Following completion of the first phase of database construction, a series of brief demonstrations of the program was made to 25 students in small groups. Issues regarding the program raised in subsequent interviews were used as the basis for the construction of a formal questionnaire, which sought responses regarding the ease of use of the database, the best routes for the access of stored material, the ways in which the database would best be used, and the usefulness of certain materials that may be included.
Sixteen students were selected at random from each of the two major groups likely to use the program, namely the fourth year medical students based primarily at distant teaching hospitals, and third year campus-based students. Five of the six academic members of staff within the School of Pathology were also asked to answer an appropriately modified questionnaire, so that discrepancies between staff and student expectations of use could be identified.
Each respondent was allowed 10-20 minutes free use of the program, and this was followed by a brief discussion with the authors regarding the potential uses of the database and the nature of additional materials that could be provided. The respondents were then asked to complete the questionnaire. Statistical analysis of questionnaire data was performed using Excel 5.0 (Microsoft Australia).
A diagram of the current design of the database is shown in Figure 1. Information is stored in separate tables, reflecting the logical arrangement of that material. The Images table contains information on each electronic image, including its file name, the type of image (X-ray, histopathological section, close-up or back view of a museum specimen etc.), and its source. Information on specimens from the School of Pathology museum, including museum catalogue and bay numbers, as well as clinical details and macroscopic and microscopic descriptions, were stored in the Museum Details table. Over 770 of the 2100 images present in the Museum have been recorded in the database, along with 1.4 megabyte (300 pages) of relevant textual information from the museum catalogue.
The diseases table contains information on separate diseases, including name, organ, system and pathological process. At present there are 507 separate disease entities involving 89 organs and 14 systems. These three tables are linked by a fourth table, the Image Path Link table, allowing great flexibility in the indexing of images. For instance, several diseases can be present in the one museum specimen, and be represented in separate images showing back, front and close-up views of the specimen. Another table provides a list of synonyms for diseases, allowing greater recognition of specific diseases. There are 888 synonyms presently recorded for the 438 specific disease entries.
Several different approaches to the graphical user interface have been attempted over the past 18 months, and modifications have been made on the basis of informal assessment by potential users. The current interface has been designed to emphasise the visual nature of the material, and has been constructed with the needs of the end user firmly in mind.
All of the 37 respondents to the questionnaire found the program either very easy to use (72%) or fairly easy to use (28%). There was no differences seen between each of the three respondent groups. A recurrent theme noticeable in comments made by respondents was the request for navigation tools such as Next/Back buttons, menus, as well as help buttons. These requests were noted, and are either available in the current program, or will be implemented in future versions.
The opening form provides the user with a choice of four routes by which information in the database can be accessed (Figure 2). One access path takes the form of a metaphor of the actual School of Pathology Museum, providing an interactive floor plan on which students may click to gain access to a visual display of specimens. Students may also access specimens directly through selection from a list of disease synonyms, chosen from an alphabetical listing (Figure 3). In this case, all diseases related to that synonym are shown on an image of a cardfile, and the specific disease of interest is then chosen from those listed. Finally, access to individual specimens can be made by a listing of either specific organs/tissues, or pathological process. When a given image is displayed in full on the screen, small icons showing relevant associated images and textual details are also provided, further facilitating student exploration of the images and their relationships (Figure 4).
As part of the formative assessment, respondents were asked to rank these different methods of access, together with a fifth option for a random display of diseases. In general, the approach favoured by a student appeared to relate closely to the purpose for which they were undertaking study. The most favoured approaches were through pathological process, and to a lesser extent through organ, whilst symptom and random approaches were considered less favourably. The low number of staff data points made comparison unreliable.
There are currently 2705 images which have been converted to 28 PhotoCDs and subsequently resized and adjusted for use in the database. When stored in the JPEG compressed format of working (640 x 480 pixel) and postage stamp (120 x 90 pixel) images, this represents a total image file space of approximately 280 megabytes.
Images recorded to date include 2279 images taken from photographs of 770 specimens, 104 X-ray or CT-scan images from 70 diseases, and 148 microscopic images. A further 336 images have been obtained from the School of Pathology 35 mm transparency teaching collection, with preference being given to those images which show fresh surgical or autopsy material (Table I). The 640 x 480 pixel images were considered to be always (65%, 24/37) or mostly (35%, 13/37) of sufficient quality to allow recognition of relevant disease details.
Responses were analysed regarding the potential utility of the program in a variety of settings. Interestingly, there was little enthusiasm regarding the proposition that the database might replace the existing pathology museum entirely, with 29% agreeing, 21% disagreeing and 50% undecided.
There have been several reports in the literature outlining the use of image databases in the teaching of Pathology. The UKPATH series of laser videodiscs is perhaps the prototypic example (2)Mercer. We have described the in-house development, over a period of 18 months, of an extensive image database relevant to the needs of our students. The database is now in use by students as a tool for private study and revision, and at the same times continues to undergo further enhancement.
We believe that this electronic image database greatly improves the current use of the visual teaching resources in the teaching of Pathology, and hence has the potential to improve the quality of student learning. Firstly, it provides a means of delivering these resources to our students, both within the School and at remote sites. Images are available over the campus network, and the program can therefore be used on personal computers throughout the university. The database has also been made available in CD-ROM format, in order to facilitate its use in associated teaching hospitals, and by students at home.
Secondly, by a variety of mechanisms, the program significantly improves the educational value of these resources. We have found that the quality of the images, particularly with magnification and appropriate lighting, is often superior to the actual specimen. We are currently developing a hot spot facility which will allow the labelling and display of discrete lesions within each specimen. The use of a database allows the images to be displayed together with information regarding the clinical history, and with detailed descriptions of the pathological changes present, thus enriching the students' appreciation of the disease process. It also allows better use of the specimens within the museum, as it permits cross-referencing of the many different points of interest within individual specimens. The existing physical layout of the museum, where specimens are grouped in separate areas based on organ systems, mitigates against effective cross-referencing of specimens in this manner.
We consider that the program provides the unsupervised student with the ability to rapidly and interactively browse through the image collection. They will be able to correlate the macroscopic appearance of a specific disease with its microscopic, radiographic and clinical appearances, in the manner and pace which best suits their needs. The integration of these features lies at the heart of teaching in pathology, yet at present is difficult to achieve except in programmed tutorials and within the setting of practical classes.
In the longer term, the database is viewed as an important resource, which can be used in the development of more precisely designed and targeted interactive teaching modules, both in Pathology and within other disciplines.
The authors would like to acknowledge the work of Dr Grace Higgins, Curator of the Pathology museum, Ms Mirella Fabbri for her photographic contributions, and Prof Athol Lykke for his enthusiastic support of the project. This work was supported by a grant from the Committee for the Advancement of University Teaching, Australia.
1. Woods JW, Jones RR, Schoultz TW, Kuenz M, Moore RL. Teaching pathology in the 21st century. An experimental automated curriculum delivery system for basic pathology. Arch Pathol Lab Med 1988; 112:852-856.
2. Mercer J, Pringle JH, Rae MJ, Harkin PJ, Lauder I. How do we teach pathology? The laser videodisc and computer-assisted learning. J Pathol 1988; 156:83-89.
3. Jones RR, Woods JW, Towbin JA, Schoultz TW. Use of the interactive videodisc to teach pathology: a preliminary report. J Biocomm 1986; 13:22-27.
4. Kumar K, Hodgins M. Use of interactive videodisc for teaching of pathology laboratory cases. J Pathol 1990; 160:145-149.
5. Hoffman HM, Irwin AE, Ligon RG. PathPics: an image-based instructional program used in the pathology and histology curriculum and transmitted over a wide area network. Proc Ann Symp Comp Applic Med Care 1992; 1992:796-797.
6. Kumar K, Khilnany M. Innovative interactive videodisc units for teaching of pathology laboratory cases. Five years' experience. Am J Clin Path 1994; 101:665-670.