Differences in Dermoscopic Images from Non-polarized Dermoscope and Polarized Dermoscope Influence the Diagnostic Accuracy and Confidence

Objective: To evaluate whether the differences in colors and structures observed in dermoscopic images from non-polarized dermoscopes (NPD) and polarized dermoscopes (PD) can impact physicians’ diagnostic ability and their confidence levels. Participants: 100 dermatologists who attended a one-day course on the fundamental of dermoscopy course at Memorial Sloan-Kettering Cancer Center. Design: Twenty five pigmented lesions were chosen, which consisted of 7 seborrheic keratoses, 3 basal cell carcinomas, 2 atypical nevi, 5 melanomas, 3 dermatofibromas, 3 blue nevi and 2 hemangiomas. Two images of each lesion (one NPD and one PD) for a total of 50 lesions were included in the image presentation. Participants were not informed of the study design and were not told that they would be viewing the same lesions under 2 different imaging modalities. Statistical analysis examining the participants’ responses was performed using the McNemar’s test and Paired t test. Main Outcome Measures: The main outcomes included the assessment of the diagnostic accuracy and confidence level for clinicians viewing lesions via NPD and PD. Results: Ninty-one participants completed the study. Statistically significant differences in the diagnoses were observed in the seborrheic keratosis, atypical nevus and melanoma groups. For seborrehic keratosis, 75% and 59% of the final participants correctly diagnosed SK when presented with the NPD and PD images, respectively. For atypical nevi, 19% and 33% had the correct diagnoses when presented with NPD and PD images, respectively. For melanomas, 23% and 34% had correct diagnoses with the NPD and PD images, respectively. For the categories of seborrehic keratosis and atypical nevus, participants were statistically more confident in their diagnoses when presented with the NPD images than with the PD images. For the category of basal cell carcinoma, participants were more confident in their diagnosis when viewing the PD images compared to the NPD images. Conclusion: There are observed differences between NPD and PD in term of the color and structure visualized. In some cases, physicians diagnostic accuracy and confidence are affected by the differences seen with the different dermoscopes. NPD and PD appears to provide different but complementary information. INTRODUCTION Dermoscopy is a valuable tool for the diagnosis of pigmented and non-pigmented skin lesions . In the hands of experienced users, the device helps to improve clinical diagnostic accuracy , and increase physicians’ confidence in their clinical diagnoses. The standard dermoscope (NPD) uses non-polarized, halogen or incadenscent light sources. These dermoscopes require the application of immersion liquids to enhance the penetration of light through the stratum corneum, thereby allowing the observer to see deeper structures within the skin. These were the only type of devices available in 1990s. As a result, nearly all of the dermoscopic structures, patterns, and diagnostic algorithms that have been described thus far are based on NPDs technology. Furthermore, dermoscopic images shown in most textbooks and in many lectures and courses are taken with cameras coupled to NPDs. Over the past several years, polarized dermoscopes 8 9 (PD) have emerged on the market. These dermoscopes use the properties of cross-polarized light to view deeper skin structures, not visible to the unaided eye. They are smaller in size, and they do not require a liquid interface. They offer the capability of viewing the skin with or without direct skin contact. The use of PD is becoming more prevalent among dermatologists, especially the residents. It was generally thought that PD and NPD are similar and the dermscopic images obtained with NPD and PD were comparable in quality. However, studies have demonstrated some striking differences in the colors and dermoscopic structures observed with NPDs and PDs . In this study, we evaluated whether these differences can impact physicians’ diagnostic accuracy and confidence levels in examining pigmented skin lesions. METHODS Subjects: Dermatologists and dermatology residents with a beginner level of experience in dermoscopy attended a day long course on the fundamentals of dermoscopy at Memorial Sloan-Kettering Cancer Center. After the morning lectures, 100 of the registered physicians participated in a study to assess their ability to diagnose lesions based on dermoscopic images. A short survey was administered to all study participants to assess their level of familiarity and/or expertise with dermoscopy. For this study, twenty five lesions were randomly chosen from a database of pigmented skin lesions with clinical images, NPD and PD images, and histologic confirmation. Only lesions with good image quality were chosen. Lesion selections were made by one of the study dermatologists (AAM). Seven categories of pigmented lesions were included in the study: 7 seborrheic keratoses, 3 basal cell carcinomas, 2 atypical nevi, 5 melanomas, 3 dermatofibromas, 3 blue nevi and 2 hemangiomas. Two images of each lesion (one standard non-polarized dermoscopic, and one polarized non-contact dermoscopic) for a total of 50 lesions were included in the image presentation. The order in which the study images were presented was randomized. The orientation (rotation) of the study images was different from NPD to PD in an effort to make the lesions less familiar. Participants were not informed of the study design and were not told that they would be viewing the same lesions under 2 different imaging modalities. Study lesions were presented to all of the participants, in a darkened lecture hall. Participants were given a wireless hand-held audience response keypad on which to record their responses. For each image, the participants were asked, the following questions: (1) “What is your diagnosis of this lesion?” (See Table 1 for the choices) and (2)“How confident are you in you diagnoses on a scale of 1-5,” with 1 indicating very confident and 5 indicating not confident at all. All 50 study lesions were presented in the same manner. Statistical Considerations: Distributional characteristics of all study variables were examined. Descriptive frequencies, means and medians were used to describe keypad response data. McNemar’s test was used to assess differences in diagnosis between NPD and PD. Paired t-tests were used to assess differences in confidence in diagnosis between NPD and PD evaluation. A general estimating equations approach was used to explore differences in physician confidence between NPD and PD. Separate regression models were created for each diagnostic lesion category. All analyses were performed with Stata SE v.9.1, College Station, TX. RESULTS: A total of 100 physicians participated in the study. Technical difficulties with the audience response system rendered 5 respondents’ data unusable. Four respondents had very incomplete data, >75% missing, so their responses were omitted from the analysis. The final sample size included 91 participants. The level of dermoscopy experience of the participants varied, and majority of the participants were novices. On average, for each lesion pair, 85% of the observers provided a diagnosis. The percentages of participants with the correct diagnoses for each category of lesions are shown in Table 2. There were statistically significant differences in the diagnoses of seborrheic keratosis, atypical nevus and melanoma. In the seborrheic keratosis group (figure 1), 75% of the final participants had correctly diagnosed SK when presented with the NPD images, and 59% had the correct diagnoses with the PD images. Sixteen percent of the responses were misdiagnosed as MM when presented with the NPD images, and 28% misdiagnosed as MM with the PD images. In the atypical nevus group (figure 2), 19% and 33% had the correct response when presented with NPD and PD images, respectively. In the melanoma group (figure 3), 23% and 34% had correct diagnoses with the NPD and PD images, respectively. There was no statistical difference for the BCC (figure 4), blue nevus, DF, and hemangioma group (Table 2). The confidence level of the participants is shown on Table 3. There was no statistical difference in the confidence levels for the blue nevus, DF, hemangioma, and melanoma groups. For the categories of SK and atypical nevus, participants were more confident with their diagnoses when presented with the NPD images than with the PD images. For the categories of BCC, the participants were more confident with their diagnoses when presented with the PD images.