Search tips
Search criteria 


Logo of jcadLink to Publisher's site
J Clin Aesthet Dermatol. 2016 March; 9(3): 36–38.
Published online 2016 March 1.
PMCID: PMC4896819

Correlation Between the Evaluation of Pigmented Lesions by a Multi-spectral Digital Skin Lesion Analysis Device and the Clinical and Histological Features of Melanoma


Objective: To correlate Multi-spectral Digital Skin Lesion Analysis classifier scores with histopathological severity of pigmented lesions and clinical features of melanoma. Design: Classifier scores were computed for 1,632 skin lesions. Dermatologists evaluated the same lesions for Asymmetry, Border Irregularity, Color variegation, Diameter >6mm, Evolution, Patient’s Concern, Regression, and/or “Ugly Duckling” sign. Classifier scores were correlated to the number of clinical risk features and for six histopathological severity levels of pigmented lesions. Measurements: Average classifier score, Welch’s t-test, and chi-square analysis. Results: Melanomas had higher mean classifier scores (3.5) than high-grade dysplastic nevi (2.7, p=0.002), low-grade dysplastic nevi (1.7, p<0.0001), non-dysplastic nevi (1.6, p<0.0001), and benign non-melanocytic lesions (2.0, p<0.0001). Classifier score and the number of clinical risk characteristics directly correlated (Pearson coefficient 0.32, p<0.0001). Conclusion: Correlation of classifier scores to clinical and histological melanoma features supports the effectiveness of Multi-spectral Digital Skin Lesion Analysis in assessing the risk of pigmented lesions requiring biopsy. Optimizing outcomes of dermatologist decisions to biopsy suspicious pigmented lesions may be enhanced utilizing Multi-spectral Digital Skin Lesion Analysis.

The incidence of malignant melanoma has increased an average of four percent per year in the United States since the early 1970s.1 Today, over one American dies every hour from melanoma.2 Detected in its earliest stage, melanoma is almost completely curable with simple excision. However, the five-year survival rate drops to <10 percent with stage IV disease. Therefore, the key challenge for early diagnosis is to determine which atypical pigmented skin lesions require biopsy. Technology has been shown to augment the biopsy decision accuracy of dermatologists evaluating suspicious pigmented lesions.3

Multi-spectral Digital Skin Lesion Analysis (MSDSLA, MELA Sciences Inc, Irvington, New York) uses narrow bands of visual and near-infrared light (430-950nm) to capture data from the skin surface down to 2.5mm in depth. At a 20-micron resolution (~diameter of 3 melanocytes), individual cellular atypia is not evaluable by MSDSLA. Instead, the device uses 75 analytical parameters to determine the level of morphological disorder within an atypical pigmented skin lesion to generate a classifier score (CS).3 Using the CS, MSDSLA provides objective data that can be integrated into the dermatologist’s biopsy decision. Knowledge of the CS was shown to improve the average biopsy sensitivity for melanoma of 179 dermatologists from 69 to 94 percent (p<0.001).4

The purpose of this study was to further characterize the MSDSLA CS for levels of histologic severity of atypical and banal pigmented lesions and its level of correlation to independently evaluated clinical features of the same lesions.


Data were collected from 1,632 skin lesions originally used to establish the safety and effectiveness of MSDSLA.5 CSs from this series ranged from -5.2 to +9.0. The raw melanoma/high-grade lesion/other lesion ratio was then calculated (Table 1). A logistic regression model was applied to the data to model the increase in probability of a lesion being melanoma/high-grade lesion as a function of an increase in CS using the formula:

Frequency of melanomas/high-grade lesions and nonmelanomas/high-grade lesions by CS

In [Ps/(1 - Ps)] = b0 + bl S

where Ps is the probability of a lesion being a melanoma or high-grade for a given CS (S).

The same 1,632 lesions were also interpreted by a clinical dermatologist panel. Eight clinical characteristics were included in their analysis—Asymmetry, Border Irregularity, Color variegation, Diameter greater than 6mm, Evolution, Patient’s Concern, Regression, and/or Ugly Duckling (ABCDEPRU). CS data were evaluated using Welch’s t-test, Pearson correlation, and chi-square analysis. Overall, the melanomas in the study were considered early by accepted criteria. Forty-five percent of the melanomas evaluated were in situ and the invasive melanomas had a median Breslow thickness of 0.36mm.


The probability of a lesion being a melanoma directly correlated with the CS (Table 1). The model demonstrated that an increase in CS of 1 increased the odds ratio of being melanoma/high-grade dysplastic lesion by 1.66 (p<0.0001). Melanomas were found to have significantly higher mean CSs than high-grade lesions (3.5 vs. 2.7, p=0.002). Melanomas also had significantly higher mean CSs than low-grade dysplastic nevi (DN) (1.7, p<0.0001), non-DN (1.6, p<0.0001), and benign non-melanocytic lesions (2.0, p<0.0001). High-grade lesions had significantly higher mean CSs than low-grade DN and non-DN. The mean CS for low-grade DN was similar to that for non-DN, partly attributable to inclusion in the latter group of relatively high mean CS intradermal and Spitz nevi (~1.8) and low mean CS junctional nevi (~1.3). Interestingly, the mean CS for melanomas in situ (n=57) was 3.9 versus 3.1 (n=70) for invasive melanomas, although this difference was not statistically significant. This finding suggests that melanoma in situ may contain similar morphologic characteristics to the superficial component of invasive melanoma.

The MSDSLA CS also correlated with level of morphologic disorganization. Mean CSs were computed for six histological severity levels of lesions: melanomas, high grade lesions (high grade DN, atypical melanocytic proliferation/hyperplasia), low grade DN, non-DN (intradermal nevi, Spitz nevi and junctional nevi), nonmelanoma skin cancers, and other non-nevus benign lesions. Of the 15 possible pair wise comparisons of the six histological variants, six statistically significant comparisons were evident using Welch’s t-test (Table 2). Data also revealed a direct correlation (Pearson coefficient 0.32, p<0.0001) between CS and the number of ABCDEPRU characteristics (Table 3).

Comparison of mean CSs between lesion types
CS by number of clinical/historical characteristics


This retrospective analysis sought to determine how the level of morphologic disorganization objectively measured by MSDSLA relates to the clinically and histologically subjective features of suspicious pigmented lesions. Higher CSs were significantly associated with more advanced histopathology. Despite the discordance that can exist among dermatopathologists when reviewing the same pigmented skin lesion,6 a significant correlation with the severity of the subjective histological description and the objectively generated MSDSLA CS was found. In addition, CS was shown to directly correlate with the number of ABCDEPRU characteristics of melanoma clinically present in the studied lesions.

MSDSLA has been shown to improve biopsy accuracy of suspicious pigmented lesions while reducing the number of unnecessary biopsies.4,5,7-9 This study demonstrates that the algorithms used to objectively generate the CS by MSDSLA correlate well with the subjective clinical and histological features that are typically used to assess the likelihood that a pigmented lesion might be melanoma.


DISCLOSURE:Dr. Winkelmann’s fellowship is partially funded by MELA Sciences Inc. Drs. Rigel, Ferris, and Cockerell serve as consultants for MELA Sciences Inc. Dr. Sober is a former consultant for MELA Sciences Inc. Ms. Tucker is an employee of MELA Sciences Inc.


1. [March 9, 2015]. “Health Effects of UV Radiation.” Ultraviolet radiation and the INTERSUN Programme. 2015. World Health Organization. Available at:
2. [March 9, 2015.]. http://www.cancer.orgacs/groups/content/@epidemiologysurveil ance/documents/document/acspc-031941.pdf American Cancer Society Cancer Facts and Figures. Available at:
3. Gutkowicz-Krusin D, Elbaum M, Jacobs A, et al. Precision of automatic measurements of pigmented skin lesion parameters with a MelaFind multispectral digital dermoscope. Melanoma Res. 2000;10(6):563–570-E7. [PubMed]
4. Rigel DS, Roy M, Yoo J, et al. Impact of guidance from a computer-aided multispectral digital skin lesion analysis device on decision to biopsy lesions clinically suggestive of melanoma. Arch Dermatol. 2012;148(4):541–543. [PubMed]
5. Monheit G, Cognetta AB, Ferris L, et al. The performance of MelaFind: a prospective multicenter study. Arch Dermatol. 2010;147(2):188–194. [PubMed]
6. Braun RP, Gutkowicz-Krusin D, Rabinovitz H, et al. Agreement of dermatopathologists in the evaluation of clinically difficult melanocytic lesions: how golden is the “gold standard”? Dermatology. 2012;224(1):51–58. [PubMed]
7. Yoo J, Roy M, White R, Rigel DS. Impact of guidance from a computer-aided multispectral digital skin lesion analysis device on dermatology residents’ decision to biopsy lesions clinically suggestive of melanoma. J Am Acad Dermatol. 2013;68(4):s1–AB152.. [PubMed]
8. Hauschild A, Chen SC, Weichenthal M, et al. To excise or not: impact of MelaFind on German dermatologists’ decisions to biopsy atypical lesions. J Dtsch Dermatol Ges. 2014;12(7):606–614. [PubMed]
9. Wells R, Gutkowicz-Krusin D, Veledar E, et al. Comparison of diagnostic and management sensitivity to melanoma between dermatologists and MelaFind: a pilot study. Arch Dermatol. 2012;148(9):1083–1084. [PubMed]

Articles from The Journal of Clinical and Aesthetic Dermatology are provided here courtesy of Matrix Medical Communications