A and B, Baseline and corresponding 1-year follow-up reflectance confocal microscopy (RCM) images (0.5 × 0.5 mm), respectively, at dermal-epidermal junction level showing a constellation of 3 adjacent dermal papillae (asterisks). C and D, Baseline follow-up RCM images (0.5 × 0.5 mm), respectively, showing nests in adjacent dermal papillae (arrowheads). E and F, Baseline and follow-up RCM images (0.5 × 0.5 mm), respectively showing corresponding arrangement of glyphics (arrowheads) and surface depressions (asterisks).
A and B, Reflectance confocal microscopy images at the dermal-epidermal junction level (0.5 × 0.5 mm) from a lower back nevus of a woman in her 40s. An hourglass shape made of 2 dermal papillae from baseline (A, asterisks and circle) can be recognized (at a slight rotation) in 1 of 4 follow-up images (B, asterisks and circle).
A and B, Reflectance confocal microscopy images at the dermal-epidermal junction level from an upper back nevus of a woman in her 60s. A. Baseline mosaic image (5 × 5 mm). Two independent readers' annotations (squares) deviate by 2%. B, Follow-up image (1.25 × 1.25 mm). The same tissue anchor composed of a constellation of 6 dermal papillae (asterisks) can be recognized in both images.
Scope A, Selinger L, Oliviero M, Farnetani F, Moscarella E, Longo C, Rabinovitz HS, Pellacani G. Precise Longitudinal Tracking of Microscopic Structures in Melanocytic Nevi Using Reflectance Confocal MicroscopyA Feasibility Study. JAMA Dermatol. 2016;152(3):299–304. doi:10.1001/jamadermatol.2015.4993
Reflectance confocal microscopy (RCM), a cellular-level, in vivo imaging technique, may be potentially used for monitoring melanocytic neoplasms for microscopic stability vs changes over time.
To test feasibility of using RCM to track specific microscopic structures within nevi over 1 year.
Design, Setting, and Participants
This was an observational study, a review of prospectively acquired RCM images, performed at a tertiary academic medical center. Seventeen patients were enrolled from adult patients presenting to pigmented lesion clinic; from each participant, 3 confirmed benign nevi were randomly selected from the upper and lower back and from the lower extremity.
Nevi underwent standardized RCM imaging at baseline and after 1 year.
Main Outcomes and Measures
We tested interobserver reproducibility in recognition of tissue anchors, RCM structures that can be identified at 2 time points. We used 2 tests to measure concordance between independent readers: (1) In the multiple choice matching test (n = 43 nevi), readers were shown a tissue anchor in a baseline RCM image (≤1 × 1-mm field-of-view) and asked to identify the same structure in 1 of 4 equally sized RCM images obtained from the same nevus at follow-up. (2) In the annotation test (n = 29 nevi), readers were shown a tissue anchor in a follow-up RCM image (≤1× 1-mm field-of-view) and asked to annotate the corresponding location of this structure in the baseline RCM mosaic image (≤5 × 5-mm field-of-view) from the same nevus; good agreement was defined as annotations deviant by less than 10% of the mosaic's width.
In total, 17 patients (mean age, 45 years [range, 28-70 years]; 10 [59%] were women) contributed a total of 51 nevi, of which 44 nevi (86%) were used for the study. Images from 7 nevi (14%) were suboptimal in quality. Tissue anchors were identified at both time points in all 44 nevi. Selected tissue anchors were located at a mean depth of 54.3 µm; the most commonly selected anchors (37 of 44 images [84.1%]) were dermal papillae. In the multiple choice matching test, compared with a reference reader, 2 readers correctly matched baseline to follow-up tissue anchors in 40 of 43 nevi (93%; P < .01) and 42 of 43 nevi (98%; P < .01), respectively. In the annotation test, there was good agreement between 2 readers in all 29 cases (100%); the mean deviation was 2% (range, 0%-7.5%).
Conclusions and Relevance
Precise longitudinal tracking of microscopic structures in melanocytic nevi using RCM is feasible.
Monitoring of melanocytic neoplasms for change is a sensitive strategy to identify melanoma.1- 3 However, melanocytic nevi also change frequently, particularly in the first 4 decades of life.4,5 Thus, a more in-depth study of the natural evolution of nevi and the expected changes in nevi over time may lead to the development of metrics for identifying the distinctive changes that occur in melanoma.
Reflectance confocal microscopy (RCM) is a technique that enables identification of tissue structures at cellular-level magnification and resolution. We have previously shown that RCM allows for classification of nevi based on their microscopic tissue patterns.6,7 We also used RCM to evaluate whether the overall pattern of nevi was stable or changing over time.7 However, to our knowledge, RCM had not been previously used to systematically track individual microscopic structures over time. We reasoned that precise tracking of microscopic structures is feasible because RCM allows for accurate navigation of the focal point of imaging within the lesion, using the x-, y-, and z-axis coordinates.
To this end, the aim of the present study was to test the feasibility of using RCM to track specific microscopic structures within nevi over a 1-year period. The present study is part of a research project that uses RCM to classify the pattern of nevi and follow the evolution of their pattern over time.
The institutional review board at Sheba Medical Center approved the study. Written informed consent was obtained from study participants, who did not receive a stipend for participation. Consenting study participants were prospectively enrolled from the population of adult patients 18 years or older presenting to a pigmented lesion clinic at a tertiary academic medical center for skin examination. To be included in the study, patients had to have at least 3 confirmed benign nevi (based on clinical and dermoscopic examination), 1 on the upper back, 1 on the lower back, and 1 on the lower extremity (excluding the foot). The nevi included in the study were randomly selected at each anatomic site. Exclusion criteria for the study included age younger than younger than 18 years, pregnancy, and being unwilling or medically unable to lie down for at least 30 minutes.
All included nevi were subject to standardized overview and close-up clinical and dermoscopic imaging. Clinical imaging was obtained using a Canon PowerShot G11 camera (Canon Inc). Dermoscopic imaging was acquired using DermLite Foto (3Gen Inc).
We performed RCM imaging using a previously described, commercially available microscope (Vivascope 1500; Caliber Imaging & Diagnostics Inc) with a 830-nm diode laser.8 Briefly, a drop of index-matching oil was applied to the skin as an immersion fluid. Next, a metal tissue ring, attached to a disposable, optically clear, polycarbonate adhesive window, was centered on the nevus and affixed to the skin. Third, a water-based gel was applied over the adhesive window as an immersion fluid. Finally, the RCM objective lens, magnetically coupled to the tissue ring, was immersed into the gel. At the initiation of RCM scanning, we identified the surface of the skin and registered it as zero depth. Individual RCM images (500 × 500 µm) were sequentially collected in the horizontal x-y plane using an automated stepper and stitched together by dedicated software to create a mosaic image with a field-of-view of up to 5 × 5 mm. For each nevus, we obtained at least 5 horizontal sections that would include the suprabasal epidermis (spinous-granular layers), the basal epidermis, the upper and deeper dermal-epidermal junction (DEJ) and the superficial dermis.
The same 3 nevi were reimaged after a period of 1 year. Variations in the follow-up period stemmed from the date of clinical follow-up visit and patient availability for RCM imaging session. At the follow-up imaging session, we used the following procedure: First, we identified the 3 nevi that were imaged at baseline using the overview clinical imaging. Second, we acquired the follow-up visit’s clinical close-up and dermoscopic images. Third, we placed the RCM's tissue ring with intention to use the same orientation as the baseline imaging. Again, at the initiation of RCM scanning, we identified the surface of the skin and registered it as zero depth. Mosaic RCM images were acquired at the same z-depths as the baseline mosaics.
Two tests were performed to ascertain that (1) the same RCM structures can be recognized in the follow-up and baseline RCM images (multiple choice matching test) and to determine whether (2) the same specific structures that are extracted from the follow-up images can be reproducibility identified in the baseline mosaic (annotation test). Reproducible identification of the same structure in the follow-up and baseline RCM images and the identification of these structures within the larger mosaic RCM image would demonstrate the potential for using RCM for monitoring melanocytic lesions for microscopic stability vs changes over time.
First, we tested whether different observers can reproducibly recognize the same tissue structures, from baseline and follow-up RCM images of the same nevus. We regard microscopic structures that can be potentially recognized at 2 time points as tissue anchors; examples of tissue anchors include dermal papillae of peculiar shape or nests of melanocytes that stand out (Figure 1). We cropped out 0.5 × 0.5-mm to 1 × 1-mm RCM images of tissue anchors from the baseline 5 × 5-mm mosaic. From the follow-up mosaic, 4 RCM images precisely corresponding in size (0.5 × 0.5 mm to 1 × 1 mm) and in imaging depth to the baseline RCM image were also cropped out, including an RCM image containing the same tissue anchor as the cropped-out baseline image, and 3 additional, randomly obtained RCM images from the same nevus. An investigator (A.S.) confirmed that the same registration anchors were present in the baseline and follow-up RCM images. As the testing interface, we prepared a PowerPoint presentation (Microsoft Inc), in which the baseline RCM image was presented side by side with a 2 × 2 tile of the 4 follow-up RCM images. The follow-up image that contained the tissue anchor corresponding to the baseline image was randomly placed within the 2 × 2-image tile. This testing interface will henceforth be referred to as the multiple choice matching test (Figure 2). Two readers (M.O., E.M.) who had not previously seen the baseline or follow-up mosaic images and were blinded to the purpose of the study were invited to participate in the study. Readers were asked to identify which one of the images from the 2 × 2 tile (follow-up images) contained the same tissue structure as the single (baseline) image. Readers were made aware of the possibility that slight variations in z-depth and xy-plane orientation of tissue structures may have occurred between the 2 matching images. Readers were also blinded to the clinical and dermoscopic images of the nevi. For each comparison, only 1 correct answer was possible. Readers were also asked to note if the comparison was difficult (eg, if they were not confident they had correctly identified the matching follow-up image). One set of baseline and follow-up images was used to explain the interface to the readers, and an additional 43 sets of images were used for testing.
Second, we tested whether readers could identify tissue anchors within large mosaic images. We cropped out 0.5 × 0.5-mm to 1.5 × 1.5-mm RCM images, which contained tissue anchors, from the follow-up mosaic. As the testing interface, we prepared a PowerPoint presentation, in which the cropped follow-up RCM image was presented side by side with the corresponding baseline mosaic image. In this test, the corresponding baseline and follow-up images were matched based on the morphologic appearance of the tissue anchor and not based on matching the precise registered imaging depth. Thus, there could be some variation in z-depth between the baseline and follow-up image. To optimize matching of the images, if there was rotation of more than 10° between the images, then the cropped follow-up image was rotated to match the alignment of the baseline mosaic. An investigator (A.S.) identified the same tissue anchor within the baseline RCM mosaic; the tissue anchor was annotated by placing a square in the corresponding location of the baseline mosaic, which was calibrated to match the size of the cropped follow-up RCM image. This testing interface will henceforth be referred to as the annotation test (Figure 3). Next, a blinded reader (M.O.) was sent the nonannotated PowerPoint interface and asked to annotate the location of the tissue anchors within the baseline mosaic; the appropriately sized squares were provided at the corner of each PowerPoint slide. Again, the readers were blinded to the clinical and dermoscopic images of the nevi. For measuring agreement between the investigators, we tested deviation in the placement of the annotation squares between the 2 readers. For this test, 31 randomly selected nevi from the 44 nevi included in the study were used. Two paired images were used to train the blinded reader, and 29 additional sets of images were used for testing.
Descriptive statistics were used to describe the study participants' demographics and the clinical and dermoscopic attributes of the included nevi. For the multiple choice matching test, a χ2 test was used to analyze the reader's frequency of correctly matching of baseline RCM image to the corresponding follow-up RCM compared with fortuitous matching the correct images by 0.25 chance; P < .05 was considered significant. The paired t test was used to compare the z-depth of the baseline and follow-up images selected for the annotation test; P < .05 was considered significant. For the annotation test, we decided a priori that good agreement between readers would be considered if the deviation was less than 10% of the breadth of the mosaic (eg, a deviation of up to 0.5 mm in the case of a mosaic 5 × 5 mm in size). Statistical evaluation was performed using Excel software (Microsoft Inc).
In total, 17 patients (mean age, 45 years [range, 28-70 years]; 10 [59%] were women) contributed a total of 51 nevi, of which 44 nevi (86%) were used for the study; the remaining 7 lesions (14%) were not used because the RCM mosaics at one of the time points was suboptimally stitched or there were artifacts that obscured the RCM image. Of the 44 nevi used for the study, 14 (32%) were from the upper back, 14 (32%) from the lower back, and 16 (36%) from the lower extremities; 29 (66%) were macules and 34% were papules; the mean diameter of the nevi was 4.7 mm. Distribution of the 44 nevi by dermoscopic pattern included 32 with a reticular pattern (72.7%); globular, 4 (9.1%); homogenous, 6 (13.6%); and complex (reticular-globular), 2 (4.5%). All lesions were imaged at baseline and after a mean follow-up interval of 12.5 months (range, 9-16 months).
Tissue anchors (Figure 1) were identified in both baseline and follow-up images in all 44 nevi. For the multiple choice matching test (Figure 2), the RCM images from 44 nevi containing the selected tissue anchors were located at a mean depth of 54.3 µm (range, 25.0-84.8 µm). The tissue anchors in the images were based on the following anatomic structures: dermal papillae (a constellation of adjacent papillae or uniquely shaped papilla), 37 of 44 images (84.1%); rete ridges (a constellation of adjacent retes or uniquely shaped retes), 9 (20.5%); dermal nests (a constellation of adjacent nests or uniquely shaped nests), 8 (18.2%); and skin glyphics (presenting as dark, elongated linear structures), 2 (4.5%); in 33 of the images (75%) only 1 type of tissue anchor was identified, and in 11 (25%) multiple types of anchors were identified. The 2 blinded readers correctly identified the matching follow-up image in 40 of 43 (93%; P < .01) and 42 of 43 (98%; P < .01) of the test set images, respectively. For both readers, the images that were incorrectly classified contained only 1 type of tissue anchor per image. The first reader noted 12 of 43 cases (28%) as difficult to match, including the 3 cases in which the selection was incorrect. The second reader noted 5 of 43 cases (12%) as difficult to match, all of which were correctly answered, while the case that was incorrectly matched was not perceived as difficult.
For the annotation test (Figure 3), we matched the tissue anchors at the 2 time points based on the morphological appearance. The mean z-depth for the tissue anchors used was 62.9 µm (range, 30.9-164.7 µm) for the baseline images and 62.8 µm (range, 44.9-109.8 µm) for the follow-up images (P = .99). In 20 of 29 images (69%), the paired images were obtained at precisely the same z-depth and in 9 of 29 (31%), there was an average z-depth difference of 32.3 µm (range, 6.0-94.5 µm). Comparing the placement of the annotating square between the 2 readers, there was good agreement (deviation < 10%) in all 29 cases (100%); the mean deviation in annotations was 2% (range, 0%-7.5%), and in 90% of cases, the deviation was less than 5%.
A promising application of RCM is the ability to noninvasively monitor tissue microscopic structures over time for stability vs change. Herein, we describe a method for precise monitoring of RCM patterns in nevi using microscopic structures as tissue anchors. After 1 year of follow-up, we were able to identify consistently the same individual tissue anchors in both the baseline and follow-up images. Variations in the shape, size, and arrangement of tissue structures, such as dermal papillae, epidermal rete ridges, nests of melanocytes, and deep surface glyphics (skin furrows), allow for their specific recognition over time.
While the investigators performing the real-time imaging have acquired substantial experience in identifying tissue anchors, we were able to demonstrate that blinded RCM readers can also readily recognize the tissue anchors. In the multiple choice matching test, blinded readers accurately recognized tissue structures imaged at baseline, from a tile of multiple follow-up images, even if the orientation of the follow-up image was at a slight angle. Furthermore, the annotation test demonstrated that independent readers are able to identify and annotate tissue anchors within large mosaic images with minimal interobserver deviation.
Our follow-up RCM imaging protocol uses the same registered z-depths as the baseline images. In 70% of the nevi, there was a very good match between the registered z-depth of imaging and the morphologically determined anatomic level of imaging. However, side-by-side comparison of the baseline and follow-up images obtained at the same registered z-depths showed that in 30% of nevi, mosaics obtained at different registered z-depths (with mean difference of 30 µm) showed a better morphologic match. The variations in depth between matching by registered z-depth vs by anatomic morphologic characteristics likely result from breathing or other movements during RCM imaging. These findings suggest that multiple horizontal sections, probably at z-intervals of 20 µm, should be obtained during follow-up imaging, to compensate for any motion artifacts.
The precise RCM monitoring technique described herein opens new and exciting avenues for research and clinical applications. First, longitudinal tracking of nevi at the microscopic level will enhance our understanding of nevogenesis.5 For example, we will be able to study whether previously theorized processes of nevus migration between the epidermis and dermis, for example, abtrupfung (downward migration) or hochsteigrung (upward migration), exist.9,10 Second, change vs stability of pattern in melanocytic neoplasms monitored via RCM will likely become apparent after a shorter follow-up interval. At present, we use dermoscopy to monitor melanocytic neoplasms for a period of 3 to 4 months to 1 year to detect change suspicious for melanoma.11 This strategy runs the risk of missing timely diagnosis in a subset of initially unrecognized melanomas, owing to patients' failure to arrive to the follow-up appointment scheduled in several months,12- 14 or to a rapid change within these months in the rare subset of fast-growing melanomas.15 Kittler and Binder16 have shown that monitoring may decrease sensitivity of melanoma diagnosis if the patient is not compliant with the follow-up appointment. Interestingly, Argenziano et al12 have shown that patient compliance with dermoscopic monitoring increases with shorter intervals; there was 84% compliance with 3-month follow-up, 63% with 6-month appointment, and only 30% with annual follow-up. To this end, we anticipate that using RCM would require much shorter intervals than dermoscopy to detect change in melanoma; such ultrashort RCM monitoring can offer added safety—clinically significant changes in melanoma will be less likely and patient compliance will probably increase. Finally, longitudinal tracking of nevi will allow for a more precise classification into biologically distinct subsets.5,6 For example, nevi that are presently categorized into different subclasses, based on their cross-sectional morphologic characteristics, may prove to be different faces of the same biologic subset over the evolution of the nevus.17
Our study has limitations. First, included lesions were clinically and dermoscopically banal nevi and hence less likely to exhibit marked change over a period of 1 year. This stability facilitates the precise recognition of tissue anchors at 2 time points. It is quite possible that in growing or more rapidly changing nevi, precise registration of microscopic structures over time will be challenging and require more frequent sequential RCM imaging. Second, the study included a limited number of nevi; it is possible that in a larger series with more light-colored nevi, the scarcity of pigmentation will make specific identification of structures over time more difficult. That said, previous RCM studies have shown that even slightly pigmented tissue structures can be detected on RCM images, as in the case of amelanotic to light colored melanomas18,19; in addition, tissue anchors, such as the relative location of skin glyphics or follicular infundibula, may allow for anchoring even in changing lesions.
Precise RCM-based longitudinal tracking of microscopic structures in melanocytic nevi using tissue anchors is feasible. This technique opens new opportunities for RCM monitoring of lesions in vivo for both clinical and research purposes. More studies are needed to show the reproducibility of recognition of tissue anchors over time, in a larger set of stable and changing melanocytic neoplasms.
Corresponding Author: Alon Scope, MD, Department of Dermatology, Sheba Medical Center, Tel Hashomer 52621, Israel (firstname.lastname@example.org).
Accepted for Publication: October 19, 2015.
Published Online: January 6, 2016. doi:10.1001/jamadermatol.2015.4993.
Author Contributions: Dr Scope had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Scope.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Scope, Farnetani, Pellacani,
Critical revision of the manuscript for important intellectual content: Selinger,Oliviero, Moscarella, Longo, Rabinovitz.
Statistical analysis: Scope, Selinger.
Obtained funding: Scope.
Administrative, technical, or material support: Selinger.
Study supervision: Scope.
Conflict of Interest Disclosures: Ms Oliviero reports serving as consultant for Caliber ID, manufacturer of a commercial confocal microscope, and Dr Rabinovitz reports serving as consultant and speaker for Caliber ID. No other disclosures are reported.
Funding/Support: The research was funded by the European Commission Marie Curie FP7 Reintegration Grant (PIRG07-GA-2010-268359; principal investigator, Dr Scope).
Role of the Sponsors: The sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of data; or in the preparation, review, or approval of the manuscript.