Vincenzo Della Mea, PhD
Institute of Pathology, University of Udine, Italy
Published as:
Della Mea V. Store-and-forward telepathology. In: European Telemedicine
1998/99. Hernandez B, Wootton R (eds.). EHTO/RSM Press/Kensington Publications,
London, 1999.
Introduction
In the recent years, there is a growing interest for low-cost telemedicine tools in all areas, including telepathology.
Briefly, the main recognised applications of telepathology include the usual telemedicine applications, i.e. telediagnosis, distant learning and teaching, remote image and data processing, and quality control. However, telediagnosis is the most prominent application, and in the case of pathology it could involve two distinct diagnostic tasks: intraoperative frozen section service and remote expert consultation. The former case occurs during surgery, and may be done by telepathology, provided that a real-time system is available. The latter case regards the request for a second opinion to a distant expert, when a difficult case is encountered during daily practice; this is currently carried out by sending to the consultant the glass slides or the paraffin blocks through courier or postal mail. It is immediately clear that this is a classical situation where a store-and-forward method could be thought; however, there are problems related to the peculiar nature of the images needed by the pathologist for doing a diagnosis that makes it not so immediate the adoption of such kind of systems.
This paper will first briefly describe the process of diagnosis and consultation in pathology, then analysing the store-and-forward solutions for doing this, including image acquisition, storage and delivery.
Diagnosis and consultation in pathology
Anatomic Pathology aims at morphologically studying lesions induced by diseases, and in turn at providing the useful information to obtain a correct diagnosis and to establish the prognosis. Anatomic Pathology is an image-based discipline, in which a diagnosis is achieved by examining visually perceivable features from images obtained in many ways, first of all from a light microscope; images thus represent the main source of knowledge.
The pathologist scans the glass slides through the microscope at different magnifications, looking for diagnostic features. When something is found, it is related to the available clinical details in order to identify one or more possible diagnoses. Sometimes, the difficulty of the case makes necessary to request a so-called second opinion to another physician, possibly an expert in such specific field. Two levels of consultation can be distinguished: one occurs inside the same department between colleagues, the other involves an external consultant, expert in the specific field of the case. Both levels of second opinion consultation are a very important issue in pathology; the latter may involve the communication with an expert located everywhere in the world.
Consultation through store-and-forward telepathology
The basic idea for introducing the store-and-forward paradigm in telepathology is to select and acquire some diagnostically representative images from the glass slide, and to send them to the remote consultant by means of some telematic channel. The complete preparation of a case thus includes image selection and acquisition, storage in an appropriate format, aggregation with clinical data, and transmission.
Image acquisition
Currently there are two ways of acquiring images from the microscope: by means of analogue or digital cameras. The former need also a framegrabber card into the computer, which converts the analogic signal coming from the camera into digital values; the usually available devices are based either on a single input chip (monoCCD) or on three separated chips for red, green and blue colour components (3CCD), which are preferable for pathology due to the higher quality. Spatial resolution is up to 600 lines with colour resolution of 24 bit, and images are acquired at up to 30 frames per second. The latter devices have a digital output, and are usually connected to the computer through a direct connection to the SCSI bus or other high speed connection; digital cameras and photoscanners reach an higher spatial and colour resolution (up to 5000x5000 pixels), in spite of acquisition speed, making its adoption adequate mainly for store-and-forward telepathology.
Generally speaking, still image acquisition allows for higher resolution than dynamic real-time methods, because the need for real-time makes not possible to adopt very high resolution images, which occupy a large amount of memory. However, the optical resolution of a microscope depends on the objective magnification, and thus at higher magnifications (starting from 20x) the resolution obtainable with 3CCD cameras is adequate for capturing all the information coming from microscope, while at low magnification, greater resolution is needed to reach the same quality level. Very high-resolution devices could be used for acquiring low magnification images of whole biopsies, making thus not necessary the selection of fields from them.
Storage and compression issues
An image with 640x480 pixels and 24 bit colour (e.g., coming from a 3CCD camera) needs 900 Kbytes of storage space; for a 3000x2000 image (typical size of a high-end digital camera), the value grows up to 17.6 Mbytes.
Being transmission time dependent on image file size, methods for reducing the amount of memory needed for storage are highly recommended. The main way to obtain this is to apply some compression algorithm, which may be of two kinds: lossless (i.e. completely reversible) and lossy (i.e., with a loss of information such that a compressed and decompressed image is different from the original one). Former methods (including Huffman and LZW) allow for compression ratios up to 1:3, depending on the image features, while the latter ones are more efficient, with a selectable level of quality and thus compression (which could be up to 100-200:1). The most known lossy algorithm is JPEG (Joint Photographic Expert Group), developed to exploit eye and camera limitations for discharging apparently unuseful visual information from images; in store-and-forward static telepathology, JPEG has been widely applied at compression ratios up to 30:1, with a recent study and formalisation of guidelines for its use (1).
Another way of reducing storage size is the colour reduction from 24 to 8 bits, which means that the number of available different colours decreases from millions to 256, while the file size becomes one third of the original one; on such files, a lossless compression method could be further applied. Doolittle et al. claim the diagnostic equivalence between 24 and 8 bit images, when the colour is reduced by a specific adaptive algorithm (2).
Another important issue is that of file formats: in order to allow the effective interchange of images, they should be stored using some standard. Currently, the GIF format is being used for 8 bit, lossless compressed images; TIFF is adequate for 24 bit lossless compressed images, and the JFIF format (usually called JPEG) hosts JPEG compressed images; other formats should be agreed upon. On the other hand, no real standardisation is available for the whole case, i.e. the aggregation of clinical data and images; current implementations use proprietary formats, or HTML as a document framework (but with the limitations due to its currently poor semantic structuration, which will be overtaken by the upcoming XML standard).
Delivery of cases
Once images have been acquired and aggregated with clinical data into a multimedia case description, the transmission can finally occur, through some telematic channel, exactly as occurs in other store-and-forward telemedicine tasks. Although in principle any method is appropriate, a correct implementation should consider all steps for making reliable store-and-forward communications, i.e., by dealing with errors and abnormal situations, and possibly considering security and privacy of data too. Internet e-mail, which has been used in different trials as a mean for doing store-and-forward telepathology, can be considered an adequate protocol for this application, because it correctly approaches the problems above by means of the client/server SMTP protocol for message transmission and the MIME message format suitable for multimedia messaging with privacy-enhanced capabilities. Some experimentation has been carried out using telepathology workstations directly connected through modems; however, this implies that both workstations are active at the same time, which is not the case when using client/server systems such as e-mail.
The following figure shows the typical aspect of a store-and-forward telepathology software based on Internet e-mail.
Discussion
Differently from other telemedicine fields, the process of pre-recording images in pathology implies the extraction of only part of the information usually available to the referring pathologist (which observes the slide directly at the microscope), due to the huge amount of data needed to record a whole glass slide at different magnifications. However, the most part of such information may be considered unuseful, but there is no certainty that all images needed for diagnosis are really selected by the referring pathologist. In fact, the selection of images is the most discussed issue of this approach, because it is in charge to the referring pathologist, which could lead the consultant pathologist towards a diagnostic misinterpretation, by subjectively and incompletely characterising the case through images selected basing on his own diagnostic hypothesis. This problem is pointed out by Weinstein to demonstrate the inadequacy of static telepathology for diagnosis (3), and is also the major difficulty recognised by pathologists as hindering diagnostic teleconsultations (4), even from a legal point of view. However, there are also several works demonstrating the diagnostic efficacy of store-and-forward static telepathology in various pathology fields (5-9) as well as papers with either sceptical or negative results (3,10), so perhaps is too early for the definitive acceptance or rejection of such method. As examples, Halliday et al. (10) evaluated to 6.3% the error rate due to image selection, whilst Nordrum et al. (9) obtained a 100% agreement on a preliminary study; several authors recognised the relationship between referring pathologistí expertise and accuracy of telediagnoses.
As a matter of fact, it seems interesting to continue the research on this field at least because it allows the asynchronous consultation of an expert, without imposing the availability of both pathologists at the same time, particularly difficult when they are located in different time zones. In addition, the quality of images in static telepathology is usually higher than that obtainable in dynamic telepathology, due the bandwidth limits for real-time transmission of large image files.
The use of the Internet for store-and-forward telepathology (but also for other telemedicine specialities) introduces new problems, related mainly to the quality of service from the security and performance points of view; such problems could be partially eliminated by means of Intranets, i.e. a sort of private networks where Internet protocols are adopted. However, the privacy issues of telemedicine are almost the same of electronic commerce, thus the solutions currently studied and adopted for it could represent the way secure communications can be carried out also for telemedicine applications.
Current trends in telepathology software implementation include the availability of a store-and-forward module inside a wider system, often provided with real-time or near-real-time capabilities, either dynamic or not. This way, the pathologist can choose time by time the most adequate solution for his diagnostic problems, provided that diagnostic quality is always ensured.
References