Telepathology diagnosis by means of digital still images: an international validation study

Routine surgical telepathology in the Department of Veterans Affairs: experience-related improvements in pathologist performance in 2200 cases

OBJECTIVE

To determine whether diagnostic concordance, case deferral rate, and/or time required to review slides changed significantly as telepathologists gained additional experience using a hybrid dynamic/store-and-forward (HDSF) telepathology (TP) system on the 2000 cases following an initial 200 consecutive surgical cases, previously reported.

MATERIALS AND METHODS

Gross surgical pathology specimens were prepared by specially trained personnel in Iron Mountain, Michigan. For TP, glass slides were placed on the stage of a robotic microscope at the Iron Mountain VAMC (remote site); control of the motorized microscope was then transferred to a pathologist located 220 miles away at the Milwaukee, Wisconsin, VAMC (host site). For each case, a telepathologist had the option of either rendering a diagnosis or deferring the case for later analysis by conventional light microscopy (LM). After the slides were read by TP and a surgical pathology report had been generated (for nondeferred cases), the slides were transported to Milwaukee, where they were reexamined by the same pathologist, now using LM. When there was disagreement between the TP and LM diagnosis, a supplemental or revised report was issued, and the referring physician was notified by telephone immediately. All supplemental and revised reports were reviewed by a third pathologist in the group. The slides were then reviewed by the pathology group practice or, when there was no consensus, by the Armed Forces Institute of Pathology to establish a "truth" diagnosis. To determine changes in telepathologist performance with experience after the initial start-up of the service, their performance in handling 10 consecutive sets of 200 surgical pathology cases was analyzed.

RESULTS

Concordance rates for clinically significant TP and LM diagnoses were high for all 10 sets, ranging from 99% to 100%. Comparing the first set (Cases 201-400) with the last set (Cases 2001-2200), viewing times per case were reduced from 10.26 min to 3. 58 min. Viewing times per slide were reduced from 3.44 min to 1.13 min per slide, comparing the first and last sets. Case turnaround times (TAT) decreased from 2.46 days to < or =1.5 days.

CONCLUSION

Thes results demonstrate that improvements in TP services occur over time as the result of additional experience using the TP system. The high diagnostic concordance and low rate of case deferral lend additional support to the proposal that a host-site pathologist using HDSF TP can substitute effectively for an on-site pathologist as a service provider.

Detection of retinal lesions after telemedicine transmission of digital images

PURPOSE/BACKGROUND

To assess whether loss of image resolution or colour and subsequent telemedicine transmission of digital images affects the accuracy of retinal lesion detection by ophthalmologists when compared with the original transparencies.

METHODS

Fifteen ophthalmologists of different experience independently scored 11 retinal images for pathological signs. The images were presented as either transparencies or colour and monochrome digital images, which had been transmitted via telephone lines to a geographically remote location. One patient's eye was also imaged using scanning laser ophthalmosocopy (SLO) which produced a dynamic black and white digital image. ANOVA analysis was performed.

RESULTS

Total scores were higher for transparencies than colour (p = 0.0003) or black and white digital images (p = 0.00006). Expert observers (n = 5) considered separately showed no significant difference of accuracy between transparencies and either colour digital (p = 0.09) or monochrome digital images (p = 0.11). Experts were better than trainees at detecting pathology from less familiar images: total score (p = 0.02), colour digital (p = 0.03), monochrome digital (p = 0.02) and SLO images (p = 0.004).

CONCLUSION

Experienced observers can identify sight-threatening retinal pathology from poorer-resolution digital images that have been transmitted by telemedicine. They can also adapt to viewing less familiar images such as black and white digital or SLO images.

Telepathology: a diagnostic tool for the millennium?

Many developments in science have their origins in science fiction and telepathology is no exception. The concept was first illustrated in 1924 in the magazine 'Radio News'. It was not until 1980, however, that the first working telepathology system was demonstrated. Although the system was shown to work, it required special hardware, dedicated software and special microwave transmission links to be installed. Little interest was shown worldwide because of the very high cost and the inability of many people to replicate such a system. Ten years later, the personal computer (PC) was able to provide more than adequate performance at low cost for both image display quality and speed, and the development of video technology had resulted in high quality images being produced by television cameras that were now easily affordable. Microscopes were also relatively cheaper. Thus, by 1993 or 1994, all the hardware necessary to produce a telepathology system was available at reasonable cost. Telepathology can now be used for remote primary diagnosis, remote referral to a specialist pathologist, remote teaching, remote presentation of post-mortem or microscopic findings, quality assurance image circulation and feedback, and consensus diagnosis for pathological review in clinical trials. There are two residual problems. The first concerns the speed of data transmission, commonly referred to as the bandwidth. The second is that the software provided by most of the manufacturers and suppliers of these systems is not entirely suitable to the task and the systems are not interoperable. It is clear that the approach of the manufacturers is at present unlikely to produce telepathology systems which pathologists feel comfortable in using. A somewhat different approach is illustrated by the accompanying article in this issue from the Berlin group, where a relatively simple Java-based applet and the Internet are used to allow single or multiple users to view slides on a robotic microscope. This could form the basis for a truly useful system, but still needs modification for some applications.

"Virtual microscopy" and the internet as telepathology consultation tools: diagnostic accuracy in evaluating melanocytic skin lesions

The Internet offers a widely available, inexpensive tool for telepathology consultations. It allows the transfer of image and text files through electronic mail (e-mail) or file transfer protocols (FTP), using a variety of microcomputer platforms. We studied the use of the Internet and "virtual microscopy" tools for the diagnosis of 35 skin biopsies, including a variety of benign and malignant melanocytic lesions. Digitized images from these lesions were obtained at 40x and 100x optical magnification, using a high resolution digital camera (Microlumina, Leaf Systems, Southborough, MA), a light microscope with a phototube adapter and a microcomputer with a Pentium 166 MHz microprocessor. Two to four images of each case were arranged on a "canvas" to represent the majority or an entire biopsy level, using Photoshop software (Adobe Systems Inc., San Jose, CA). The images were compressed using Joint Photographers Expert Group (JPEG) format. The images were then viewed on a computer video monitor in a manner that closely resembles light microscopy, including scrolling by using the "hand tool" of Photoshop and changing magnification digitally up to 4 times without visible image degradation. The image files, ranging in size from 700 kilobytes to 2.1 megabytes (average 1.6 megabytes) were attached to e-mail messages that contained clinical information, using standard Multipurpose Internet Mail Extension (MIME) protocols and sent through the Internet, for interpretation by a dermatopathologist. The consultant could open the images from the e-mail message, using Microsoft Outlook Express (Microsoft Corp., Redmond, WA) and Photoshop software, scroll them, change magnification and render a diagnosis in a manner that closely simulates light microscopy. One hundred percent concordance was obtained between the telepathology and traditional hematoxylin and eosin slide diagnoses. The Internet and relatively inexpensive "virtual microscopy" tools offer a novel technology for dermatopathology consultations. Potential applications of this technology to pathology and technical problems posed by the use of an open, widely distributed network to share sensitive medical information are discussed.

Image sampling in static telepathology for frozen section diagnosis

BACKGROUND

A frozen section diagnostic service is often not directly available in small rural or mountain hospitals. In these cases, it could be possible to provide frozen section diagnosis through telepathology systems. Telepathology is based on two main methods: static and dynamic. The former is less expensive, but involves the crucial problem of image sampling.

AIMS

To characterise the differences in image sampling for static telepathology when undertaken by pathologists with different experience.

METHODS

As a test field, a previously studied telepathology method based on multimedia email was adopted. Using this method, three pathologists with different levels of experience sampled images from 155 routine frozen sections and sent them to a distant pathology institute, where diagnoses were made on digital images. After the telepathology diagnoses, the glass slides of both the frozen sections and the definitive sections were sent to the remote pathologists for review.

RESULTS

Four of 155 transmissions were considered inadequate by the remote pathologist. In the remaining 151 cases, the telepathology diagnosis agreed with the gold standard in 146 (96.7%). There was no significant divergence between the three pathologists in their sampling of the images. Each case comprised five images on average, acquired in four minutes. The overall time for transmission was about 19 minutes.

CONCLUSIONS

The results suggest that in routine frozen section diagnosis an inexperienced pathologist can sample images sufficiently well to permit remote diagnosis. However, as expected, the internet is too unreliable for such a time dependent task. An improvement in the system would involve integrated real time features, so that there could be interaction between the two pathologists.

Telemedicine applications in otolaryngology

Under suitable technical and clinical conditions, remote interactive fiber-optic NPL can be used to evaluate a range of commonly occurring pathologies with a high degree of reliability. A clinical protocol appropriate for interactive and store-and-forward fiber-optic NPL was proposed for further evaluation. Additional applications of telemedicine in otolaryngology were outlined, including otoscopy, intra-oral examination, and evaluation of external facial pathology. We envisage tele-otolaryngology taking place in a variety of ways: Interactions between rural-based PCPs and specialists (routine evaluation of hoarseness, dysphagia), using store-and-forward techniques. Consults from emergency medicine physicians at a general service hospital to a specialist (laryngeal trauma, acute peritonsillar abscess, TM perforations) using interactive means. Case discussions between specialist and sub-specialist using a combination of store-and-forward and interactive technologies. Potentially, there are at least three significant benefits from widespread acceptance of telemedicine in the field of otolaryngology, or indeed in any of the medical specialties: Saved lives and reduced medical costs due to early detection of serious pathology (in this case, head and neck cancers). Reduced unnecessary referrals to specialists, and consequent savings to the patient and health-care insurer, accompanied by more efficient usage of specialist time. Enhanced level of medical education and interaction, as the link between the referring and consulting physician is more immediate and direct [27]. For these reasons, combined with the high number of visits at the primary-care level related to issues in otolaryngology-head and neck surgery, tele-otolaryngology is poised to be a leading telemedicine application within the next few years.

Dermatopathology via a still-image telemedicine system: diagnostic concordance with direct microscopy

OBJECTIVE

To determine the concordance of dermatopathology diagnosis by still-image telemedicine technology and direct microscopy.

MATERIALS AND METHODS

Skin specimens (N = 79) were examined by a dermatopathologist using a still-image phone system, and the diagnoses were compared with those made by the same dermatopathologist 1 year earlier by direct microscopy. The telemedical diagnoses were reached first without, and then with, patient histories.

RESULTS

When the patient history was available, identical diagnoses were made in 66 of the 79 cases (84% concordance rate). Without patient history, the concordance rate was 80%. The diagnostic concordance rate for the diagnosis of benign nevocytic nevi, inflammatory diseases, and benign and malignant non-squamous cell carcinoma neoplasms was statistically significantly greater than the concordance rate for the diagnosis of squamous cell carcinoma and squamous cell carcinoma in situ (P = 0.005).

CONCLUSIONS

The diagnostic concordance rate achieved by teledermatopathology using a still-image phone system fell short of the 99% intraobserver diagnostic concordance rate using direct microscopy.

Virtual microscopy: high resolution digital photomicrography as a tool for light microscopy simulation

Recent advances in microcomputers and high resolution digital video cameras provide pathologists the opportunity to combine precision optics with digital imaging technology and develop new educational and research tools. We review recent advances in virtual microscopy and describe techniques for viewing digital images using a microcomputer-based workstation to simulate light microscopic examination, including scanning at low power to select features of interest and zooming to increase magnification. Hardware and software components necessary to acquire digital images of histological and cytological slides, and closely simulate their examination under a light microscope are discussed. The workstation is composed of a MicroLumina digital scanning camera (Leaf Systems, Southborough, MA), light microscope (Olympus Optical Co., Lake Success, NY), Pentium (Intel Corp., Santa Clara, CA) 166 MHz microcomputer configured with 64 megabytes of random access memory (RAM), a MGA Millenium Powerdesk graphics card (Matrox Graphics, Inc., Montreal, Canada) and Photoshop software (Adobe Systems Inc., San Jose, CA) running in a Windows 95 (Microsoft Corp., Redmond, WA) environment. Images with spatial resolutions of up to 2700 x 3400 pixels in 36-bit color, can be displayed simultaneously as distinct images in a montage, or merged into a single composite image file to highlight significant features of a histological or cytological slide. These image files are saved in Joint Photographers Experts Group (JPEG) format using compression ratios of up to 80:1 without detectable visual degradation. The advantages and technical limitations of various workstation components are addressed and applications of this technology for pathology education, proficiency testing, telepathology, and database development are discussed.

Teleconsultation in otolaryngology: live versus store and forward consultations

OBJECTIVE

To evaluate the relative strengths and weaknesses of interactive and delayed teleconsultations in otolaryngology.

SETTING

Ambulatory clinic at an urban tertiary care facility.

SUBJECTS

Forty-five adult patients with known or suspected upper aerodigestive tract pathology.

INTERVENTION

Patients were interviewed by an otolaryngology chief resident (CR) using a standardized protocol; the results were presented to a board-certified otolaryngologist present locally (LBCO) and a remote physician viewing the encounter by video-conferencing elsewhere in the hospital (RBCO). The CR performed a complete otolaryngologic examination, including fiberoptic nasopharyngolaryngoscopy. The CR and LBCO viewed the examination on a video monitor; the RBCO viewed the same image on the video-conferencing monitor. Each physician independently recorded findings and rendered a diagnosis. A third board-certified otolaryngologist, who reviewed the stored data file (text and stored images) in a delayed fashion (DBCO), documented his findings and made a diagnosis.

RESULTS

The CR and LBCO agreed on diagnosis in 92% (36 of 39) of cases. The LBCO and RBCO arrived at the same diagnosis in 29 of 34 (85%) cases. The DBCO agreed with the LBCO for 18 of 28 (64%) diagnoses. Agreement on management recommendations between the LBCO/DBCO pair were also lower than for the LBCO/RBCO pair.

CONCLUSIONS

Both interactive and delayed techniques can be used to provide relatively accurate clinical consultations in otolaryngology. Telemedicine can be applied for subspecialty consultations, screening programs, remote emergency triage, second opinions, and resident education.

Image selection in static telepathology through the Internet

A telepathology study was carried out to examine the differences occurring when the images were selected by an experienced pathologist, a junior pathologist and a first-year resident. One hundred and fifty-five consecutive frozen-section pathology cases were collected and sent for consultation to a remote experienced pathologist using multimedia email. Local diagnoses (as reported in the files of the Institute, not from the image selector) and remote diagnoses (based on the images) were compared with those performed on paraffin-embedded sections. Acquisition time and number of selected images were recorded for each case and used to compare the different behaviour of the three local pathologists. Of the 155 cases sent by telepathology, four were considered insufficient for a diagnosis by the remote pathologist and thus the diagnosis was postponed. In the remaining 151 cases, the overall diagnostic agreement between remote and definitive diagnosis was 96.7%. The results indicate that in the routine diagnostic work of a frozen-section service, an inexperienced pathologist can select images which are sufficiently informative for a remote diagnosis, in a sufficiently short time.

Transaction costs economics as a conceptual framework for the analysis of barriers to the diffusion of telemedicine

Telemedecine has been talked about for more than 20 years, without it entering daily use with any success. Based on transaction costs economics, the present analysis of the exchange relationships between health care producers highlights certain characteristics of the current technical and legislative context, which leads to transaction costs. It also demonstrates that the introduction of telemedicine shifts the costs associated with agents' opportunism from patients to health-care producers themselves. All these costs may be considered nowadays to thwart the use of telemedicine. It is argued here that the Public Authorities and professionals of health care could act upon telemedicine in two fields: (1) intervention in the institutional environment aims notably at better defining the property rights of telemedicine, and so constitutes an unavoidable means of encouraging health-care producers to invest in new technology; and (2) implementation of organisational forms and mechanisms susceptible to regulating such telemedical relationships between health care producers-given the present institutional environment-constitutes an essential means for overcoming the immediate barriers blocking the diffusion of telemedicine.

Diagnostic accuracy of a rural live video telepathology system

Accuracy of diagnoses rendered using a live video telepathology network was assessed for permanent sections of surgical pathology specimens. To determine accuracy, telepathology diagnoses were compared with those obtained by directly viewing the glass slide using a standard microscope. A total of 294 cases were read via both telepathology and glass slide by attending pathologists at a tertiary care medical center. Overall accuracy was defined as exact concordance between diagnoses. Clinically insignificant differences in diagnoses were excluded to determine clinically significant accuracy. For the 285 cases with complete data, the overall accuracy for telepathology was 0.912 (95% confidence interval [CI], 0.872-0.941), whereas the overall accuracy for glass slide readings was 0.968 (95% CI, 0.939-0.985). This difference is statistically significant (p = 0.009). When focusing on clinically significant discrepancies, where the difference in diagnosis might affect therapeutic decisions, the video accuracy was only slightly less than the glass slide accuracy (0.965 [95% CI, 0.934-0.982] vs. 0.982 [95% CI, 0.957-0.994], respectively), but this difference is not statistically significant (p = 0.302). Most of the cases with clinically significant differences involved lesions with inherently high interobserver variation. Certainty of diagnosis did not differ between video and glass slide readings (p = 0.911), but there was an association between certainty of diagnosis and diagnostic accuracy for video (p = 0.003 for clinically significant accuracies). Based on these findings, we recommend when using this telepathology system that only preliminary diagnoses should be given in the following situations: for diagnostic areas with known high interobserver variability; when the consultant has any degree of uncertainty about the presence or absence of the lesion in question; and when there is insufficient experience using telepathology as a diagnostic medium.

Dynamic-robotic telepathology: Department of Veterans Affairs feasibility study

In this retrospective study, we assess the accuracy, confidence levels, and viewing times of two generalist pathologists using both dynamic-robotic telepathology and conventional light microscopy (LM) to render diagnoses on a test set of 100 consecutive routine surgical pathology cases. The objective is to determine whether telepathology will allow a pathology group practice at a diagnostic hub to provide routine diagnostic services to a remote hospital without an on-site pathologist. For TP, glass slides were placed on the motorized stage of the robotic microscope of a telepathology system by a senior laboratory technologist in Iron Mountain, MI. Real-time control of the motorized microscope was then transferred to a pathologist in Milwaukee, WI, who viewed images of the glass slides on a video monitor. The telepathologists deferred rendering a diagnosis in 1.5% of cases. Clinically important concordance between the individual diagnoses rendered by telepathology and the "truth" diagnoses established by rereview of glass slides was 98.5%. In the telepathology mode, there were five incorrect diagnoses out of a total of 197 diagnoses. In four cases in which the telepathology diagnosis was incorrect, the pathologist's diagnosis by LM was identical to that rendered by telepathology. These represent errors of interpretation and cannot be ascribed to telepathology. The certainty of the pathologists with respect to their diagnoses was evaluated over time. Results for the first 50 cases served as baseline data. For the second 50 cases, confidence in rendering a diagnosis in the telepathology mode was essentially identical to that of making a diagnosis in the LM viewing mode. Viewing times in the telepathology mode also improved with more experience using the telepathology system. These results support the concept that an off-site pathologist using dynamic-robotic telepathology can substitute for an on-site pathologist as a service provider.

Diagnostic accuracy of an international static-imaging telepathology consultation service

Static-image and dynamic- (real-time) image telepathology are competing technologies. Although some studies suggest that the diagnostic accuracy of the dynamic-image telepathology approaches the accuracy of light microscopy, few reports have documented the diagnostic accuracy of static-image telepathology as used in the setting of an actual surgical pathology consultation practice. We report the results of an analysis of 171 telepathology consultation cases submitted to the Arizona-International Telemedicine Network (AITN). Digital images were submitted by pathologists from six participating institutions in Arizona, Mexico, and China. Telepathologists could render a telepathology diagnosis (TP) or defer rendering a diagnosis to obtain additional video images, glass slides for detailed analysis, or to obtain tissue blocks for special studies such as immunohistochemistry. The telepathologists rendered diagnoses for 144 cases and deferred 27 cases. Two pathologists retrospectively evaluated-glass slides from each case and rendered a consensus glass slide (GS) "truth" diagnosis. There was 88.2% concordance between TP and GS diagnoses (127 of 144 diagnoses). Concordance of 96.5% was achieved for clinically important diagnoses (139 of 144 diagnoses). Telepathologists deferred making a diagnosis to obtain glass slides for conventional light microscopy in 14 cases (8.1%) and for results of immunohistochemistry studies in 13 cases (7.6%). Thus, correct diagnoses were rendered by static-image telepathology in 127 of 171 cases (74.3%) at the time of telepathology diagnostic sessions. Inappropriate field selection and sampling biases of referring pathologists, as well as a tendency of static-image telepathologists to underestimate the complexity of some cases, may reduce the value of consultations based on the viewing of static images.