Face-to-Face Diagnosis vs Telediagnosis of Pigmented Skin Tumors:  A Teledermoscopic Study FREE

THE DIAGNOSIS of pigmented skin tumors is a continuing challenge, especially for primary care physicians, but also for dermatologists. Dermoscopy (dermatoscopy, epiluminescence microscopy, surface microscopy) has been introduced as an additional tool for the diagnosis of pigmented skin lesions, allowing the clinician to visualize features that are not discernible by naked-eye examination.^1- 13 In recent years, dermoscopic features have been proved to increase diagnostic accuracy when interpreted by expert dermatologists. Recently, progress in informatic sciences and in telematic technologies led to a new methodological approach called telemedicine that includes telecontrol, teleconsultation, and teleassistance using televisual and audiovisual systems.^14- 22

The first applications of telemedicine in dermatology were described by Perednia and Brown²³ in 1995. These authors conducted a study to demonstrate the utility of teledermatology in some rural areas of Oregon, because there were only 2 practicing dermatologists in the entire eastern part of this state. In this preliminary study, however, no detailed results concerning diagnostic validity were presented. In 1997, Zelickson and Homan²⁴ described the utility of dermatologic consultations using telematic network in a nursing home in Minneapolis, Minn. A nurse sent digital images of 30 cutaneous disorders of 29 patients, using a standard telephone line, to several dermatologists, who independently made a preliminary diagnosis on the monitor. These telediagnoses obtained by the "store-and-forward" method were compared with those made by the on-site dermatologist and revealed a diagnostic concordance in a high percentage of cases (88%). In the last 2 years, several studies have been performed demonstrating that teledermatology represents a useful diagnostic tool, especially to support underserved communities with a competent dermatologic service.^25- 30

In 1998, Provost et al³¹ reported a high clinical and dermoscopic concordance in the diagnosis of atypical (dysplastic) melanocytic nevi and early malignant melanoma by 4 observers when comparing rearview-projected conventional transparency slides with transmitted, compressed, digitized images. Remarkably, to our knowledge no further studies have been performed focusing on the validity of teleconsulting for the diagnosis of pigmented skin lesions.

Our purposes were to verify the diagnostic concordance of pigmented skin lesions using dermoscopic devices between 2 groups of expert dermatologists, one of which examined the patients on a face-to-face basis and the other evaluated exclusively digital images on the monitor (telediagnosis); to compare face-to-face diagnoses and telediagnoses with histopathologic diagnoses; and to compare the diagnostic accuracy obtained by telediagnosis with the level of diagnostic difficulty of a given pigmented skin tumor and with the photographic quality of images.

Sixty-six pigmented skin lesions, including melanocytic (n = 57) and nonmelanocytic (n = 9) lesions, which were subsequently excised, were selected for study during a 3-week period. In detail, the histopathologic diagnoses were as follows: Clark nevus (n = 50), Clark nevus and angioma (n = 2), blue nevus (n = 1), dermal nevus (n = 1), Reed nevus (n = 1), Spitz nevus (n = 1), malignant melanoma (n = 1); basal cell carcinoma (n = 4), angiokeratoma (n = 1), epidermis cyst (n = 1), dermatofibroma (n = 1), seborrheic keratosis (n = 1), and tattoo (n = 1). The histopathologic diagnosis of Clark nevus has been made according to the criteria of Ackerman et al.³² We included 34 female and 32 male patients (mean age, 41.2 years; range, 8-82 years). Lesions were situated on the back (n = 24), lower extremities (n = 11), abdomen (n = 8), buttocks (n = 5), chest (n = 4), upper extremities (n = 3), thorax (n = 3), thigh (n = 3), head (n = 2), neck (n = 1), nose (n = 1), and ear (n = 1).

The equipment used for dermoscopic analysis is composed of a stereomicroscope with various magnifications (×6, ×10, ×16, ×25, and ×40) (Wild-Heerburg M-650; Leica Microscopy Systems Ltd, Heerbrugg, Switzerland), a high-resolution, 3 CCD (charged-couple device) video camera (DXC 930P; Sony Corporation, Tokyo, Japan), a personal computer with a 120-MHz processor (Pentium; Intel Corporation, Santa Clara, Calif) and 16-megabyte RAM (Deskpro 4000; Compaq Computer Corporation, Houston, Tex), a high-resolution 50.8-cm color monitor (PVM2053MD; Sony Corporation), commercially available software (Digital Epi Microscopy Melanoma Image Processing Software [DEM MIPS]; Biomips SRL, Florence, Italy), a modem at 28,800 bauds per second (Trust Communicator; AASHIMA Distribution, Witham Essex, England), and an Internet e-mail browser (Eudora Pro; Qualcomm, San Diego, Calif).

Each clinical and dermoscopic image was captured with the stereomicroscope connected to a high-resolution video camera and a personal computer in Truevision advanced graphic array format file (Targa; 768 × 576 pixels, 870 kilobyte) and converted successively into a Joint Photographic Expert Group (JPEG) format file (78% < Quality < 85%, 40-80 kilobyte), with the DEM MIPS software. The compression of images was necessary to decrease their size and to reach a high-speed transmission using telematic networks. Interestingly, the image quality of a given pigmented skin lesion did not decrease substantially by conversion and compression to JPEG format.

All pigmented skin lesions were examined first clinically and then dermoscopically by 2 of us (K.P., S.C.) at the Department of Dermatology, University of L'Aquila, L'Aquila, Italy; the final diagnosis was based on clinical features and dermoscopic criteria (epiluminescence microscopy pattern analysis). Only 1 clinical and 1 dermoscopic image of a given pigmented skin lesion, after the conversion in JPEG format file, were codified with an anonymous code number and were sent via e-mail as attached files to the Department of Dermatology, University of Graz, Graz, Austria, together with essential but anonymous clinical data such as age and sex of the patients and site of the lesions (eg, patient 3: 82-year-old man with pigmented lesion on the back). All clinical and dermoscopic images were taken at magnifications of 16- and 25-fold, but only images with 16-fold magnification were chosen for teleconsultation.

After informed consents were obtained from the patients, all lesions were excised surgically at University of L'Aquila and diagnosed histopathologically by a dermatopathologist (H.K.) at Graz University.

In each of the 66 cases, 1 clinical and 1 dermoscopic image was evaluated on a 38.1-cm standard-resolution color monitor (800 × 600 pixels; 65,536 colors; 75 Hz) (Philips) by a dermatologist (H.P.S.) in Graz. In addition to the images, only the basic clinical data of the patients (ie, age, sex, and location of the pigmented skin lesion) were available. The level of diagnostic difficulty for each individual pigmented skin lesion was judged by another dermatologist (I.H.W.) as low, medium, or high based on the concurrent knowledge of histopathologic diagnoses. This investigator was not aware of the face-to-face diagnosis or the telediagnosis.

The quality of the images was further analyzed on the same standard-resolution color monitor on which the telediagnosis was performed by a professional photographer in Graz (W.S.) and classified as excellent, good, or sufficient.

Statistical analysis was performed using the Statistical Product and Service Solutions Software for MS Windows (SPSS Inc, Sunnyvale, Calif). First, the concordance between the face-to-face diagnosis and the telediagnosis was calculated using κ statistics. Then, the concordance between the face-to-face diagnosis and the histopathologic diagnosis as well as between the telediagnosis and the histopathologic diagnosis was calculated. Further, the level of diagnostic difficulty of a given pigmented lesion was compared with the face-to-face diagnosis and with the telediagnosis. Also, the quality of images was correlated to the correctness of telediagnosis.

There was a concordance in 60 (91%) of 66 cases between the face-to-face diagnosis and the telediagnosis. Concordance between the face-to-face diagnosis and the histopathologic diagnosis was observed in 61 (92%) of 66 cases, whereas the concordance between the telediagnosis and the histopathologic diagnosis was lower, with 57 (86%) of 66 cases. Although the number of correct telediagnoses was lower than that of face-to-face diagnoses, the difference was not statistically significant (Wilcoxon test, P = .10). Remarkably, when the face-to-face diagnosis was missed, the correct telediagnosis could not be achieved (κ, P<.001).

Concordance between face-to-face diagnosis and telediagnosis did not depend on whether the face-to-face diagnosis was correct, and the concordance of the diagnosis was not influenced by whether the lesion was a nevus or not.

The level of diagnostic difficulty of a given pigmented skin lesion was valued as high in 16 cases (24%), medium in 35 (53%), and low in 15 (23%). In the 6 cases (9%) in which there was a discordance between the face-to-face diagnosis and the telediagnosis, the level of diagnostic difficulty was high in 4 and medium in 2. Four of these 6 cases (patients 3, 11, 59, and 66) were diagnosed correctly face-to-face as Clark nevi, but overdiagnosed as melanoma (patients 3 and 59), melanoma in situ (patient 11), and melanoma within a nevus (patient 66) by telediagnosis. In a case of tattoo (patient 62), the face-to-face diagnosis was correct, whereas the telediagnosis of blue nevus was made. The last of these 6 cases (patient 15), a heavily pigmented Clark nevus, was interpreted as seborrheic keratosis in face-to-face diagnosis, whereas the telediagnosis was pigmented spindle cell nevus (Reed nevus).

The quality of photographs as assessed by a professional photographer on a standard-resolution color monitor was defined as excellent in 16 cases (24%), good in 39 cases (59%), and sufficient in 11 (17%).

Remarkably, the concordance of the face-to-face diagnosis with the telediagnosis did not depend on the quality of images but was influenced significantly by the level of diagnostic difficulty of a given pigmented skin lesion (Spearman correlation, P = .01). The correctness of the face-to-face diagnosis depended on whether the lesion was a nevus or any other diagnosis (χ² test, P = .03) and depended significantly on the level of diagnostic difficulty (Spearman correlation, P = .007). The correctness of the telediagnosis also depended highly on the level of diagnostic difficulty (Spearman correlation, P = .002), but it did not depend whether the lesion was a nevus or any other diagnosis (χ² test, P = .07). The correctness of the telediagnosis did not depend on the quality of images (Spearman correlation, P = .83). Detailed data on the discordance between face-to-face diagnosis and histopathologic diagnosis as well as on the discordance between telediagnosis and histopathologic diagnosis are provided in Table 1 and Table 2.

In this study of 66 pigmented skin lesions, we have demonstrated a high level (91%) of concordance between the clinical and dermoscopic diagnosis on a face-to-face basis and the telediagnosis on a standard-resolution color monitor. All images were sent to Graz based on the store-and-forward method via e-mail. Although the number of correct telediagnoses was somewhat lower than that of face-to-face diagnoses, this difference was not statistically significant (Wilcoxon test, P = .10). The lower rate of correct telediagnoses was not influenced by the photographic quality of the images that basically was rather good in all cases, but the diagnostic accuracy was related in both locations to the level of diagnostic difficulty of the pigmented skin lesions (Spearman correlation, P = .01). Our results with a nearly similar diagnostic accuracy are not surprising, because the level of competence in regard to the diagnostic ability of pigmented skin lesions basically is the same for both groups of investigators. A higher diagnostic accuracy for the telediagnostic procedure presumably can be obtained only when the face-to-face diagnosis is made by general physicians or other staff members (nurses, photographers) not especially trained in diagnosing pigmented skin lesion. However, this assumption has not yet been proven by a well-designed, large-scale, prospective study comparing face-to-face diagnosis by general physicians with telediagnosis by experts in pigmented skin lesions. Since the accuracy of telediagnosis was only marginally worse than face-to-face diagnosis, a patient could be referred to telediagnosis when face-to-face counseling by a dermatologist is not possible for geographical or other reasons.

The phenomenon that some pigmented skin lesions are more difficult to diagnose than others is well known by dermatologists and is reflected by the relative low sensitivity in the clinical diagnosis of melanoma.³³ Moreover, our findings are in accordance with those of Kvedar et al,²⁶ who demonstrated that the level of diagnostic certainty of the office-based clinician had a greater impact on diagnostic accuracy than photographic quality. With regard to the photographic quality of the images analyzed by a professional photographer, in 3 of 6 cases with discordant diagnoses the quality was excellent, and in the other 3 cases good. So, it is reasonable to state that the telediagnosis was not influenced by the photographic quality of the images.

In 4 of 6 cases in which there was a discordance between diagnoses, the differential diagnosis between nevus or melanoma was very difficult as highlighted by the high level of diagnostic difficulty. Nevertheless, these 4 cases were diagnosed correctly face to face as Clark nevi, but telediagnosed as melanoma, probably because the lesions were dark brown to black and, therefore, dermoscopic structures were not clearly discernible. In a case of tattoo, the face-to-face diagnosis was correct, but the telediagnosis was not, presumably because the history of the patient was not available in Graz. The last of these 6 cases, a heavily pigmented Clark nevus, was interpreted face-to-face as melanoma, whereas only the particular type of nevus was not recognized correctly by telediagnosis, and the lesion was interpreted as pigmented spindle cell nevus (Reed nevus).

Disagreement between face-to-face clinical and histopathologic diagnosis occurred in 5 (8%) of 66 cases. In 2 of these (patients 2 and 61), the face-to-face diagnosis of melanoma was based on dermoscopic features of melanoma, ie, asymmetry due to nonhomogeneous coloration, pseudopods, and irregular extensions. The histopathologic findings, however, revealed an atypical Spitz nevus and a Clark nevus, respectively.

In our study, no malignant lesions (1 melanoma and 4 basal cell carcinomas) on a face-to-face basis as well as that of telediagnosis were interpreted as being benign. The only melanoma included in this study (patient 55) was diagnosed correctly using both methods. The low number of melanomas in our study (1 [2%] of 66 skin tumors) reflects the normal prevalence of melanoma in a pigmented skin lesion clinic and represents a realistic approach to the daily routine.

The technical equipment used for this study is already easily available and certainly will become less expensive in the future. It is not necessary to have a specialized technical education to use these telematic technologies. The compression of the images in JPEG format file is necessary to reach a high-speed transmission. Nevertheless, the quality of the compressed images is sufficient to make a correct diagnosis.³⁴ The cost of teleconsultation using the Internet is not very expensive. With more advanced communication technologies, such as ISDN (Integrated Services Digital Network), optical fibers, and satellites, it will be possible to send images of even greater quality at a high speed of transmission.

Based on our experience, we believe that for pigmented skin tumors, the store-and-forward method is better than interactive teleconsultation. We demonstrated that a dermatologist specialized in dermoscopy can make a correct dermoscopic diagnosis based on static images of pigmented skin lesions on a standard-resolution color monitor.

Based on the results of our study of 66 pigmented skin lesions with a diagnostic concordance of 91% between the face-to-face diagnosis and the telediagnosis using the store-and-forward method, we conclude that teleconsultation of clinical and dermoscopic images via e-mail represents a valuable tool for the diagnosis of pigmented skin lesions when expert counseling is not available for face-to-face diagnosis.

Accepted for publication June 18, 1999.

We are grateful to Werner Stieber, medical photographer at the Department of Dermatology, University of Graz, Graz, Austria, for the careful assessment of the quality of the photographic images.

Corresponding author: H. Peter Soyer, MD, Department of Dermatology, University of Graz, Auenbruggerplatz 8, A-8036 Graz, Austria (e-mail: peter.soyer@kfunigraz.ac.at).