Diagnostic accuracy of dermatoscopy for melanocytic
and nonmelanocytic pigmented lesions
Cliff Rosendahl, MB BS,aPhilipp Tschandl, Cand med,bAlan Cameron, MB BS,aand Harald Kittler, MDb
Brisbane, Australia, and Vienna, Austria
Background: It is unknown whether dermatoscopy improves the diagnostic accuracy for all types of
pigmented skin lesions or only for those that are melanocytic.
Objective: We sought to assess if the addition of dermatoscopy to clinical examination with the unaided
eye improves diagnostic accuracy for all types of pigmented lesions.
Methods: We analyzed 463 consecutively excised pigmented skin lesions collected during a period of
30 months in a primary care skin cancer practice in Queensland, Australia.
Results: Of 463 lesions, 217 (46.9%) were nonmelanocytic. Overall 30% (n = 138) were malignant
including 29 melanomas, 72 basal cell carcinomas, and 37 squamous cell carcinomas. The diagnostic
accuracy for malignant neoplasms measured as area under receiver operating characteristic curves was 0.89
with dermatoscopy and 0.83 without it (P \ .001). Given a fixed specificity of 80%, the corresponding
sensitivity was 82.6% with dermatoscopy and 70.5% without it. The improvement achieved by dermatos-
copy was higher for nonmelanocytic lesions than for melanocytic lesions. A short algorithm based on
pattern analysis reached a sensitivity of 98.6% for basal cell carcinoma, 86.5% for pigmented squamous cell
carcinoma, and 79.3% for melanoma. Among benign conditions, the highest false-positive rate (90.5%) was
observed for lichen planuselike keratosis.
Limitations: Estimates of diagnostic accuracy are influenced by verification bias.
Conclusions: Dermatoscopy improves the diagnostic accuracy for nonmelanocytic lesions. A simple
algorithm based on pattern analysis is suitable for the detection of melanoma and nonmelanoma skin
cancer. ( J Am Acad Dermatol 2011;64:1068-73.)
Key words: basal cell carcinoma; dermatoscopy; melanocytic nevi; melanoma; nonmelanoma skin cancer;
pigmented skin lesions; squamous cell carcinoma.
lesions in comparison with examination with the
unaided eye.1-3However, all studies focused on
battery of studies including 3 meta-analyses
demonstrates that dermatoscopy improves
the diagnostic accuracy for pigmented skin
the differentiation between melanoma and melano-
cytic nevi, without adequately considering the fact
that many pigmented lesions examined in routine
clinical practice are not of melanocytic origin.
Nonmelanocytic pigmented lesions frequently ex-
amined by dermatoscopy are seborrheic keratoses,
solar lentigines, lichen planuselike keratoses (LPLK)
(also known as benign lichenoid keratoses), dermat-
ofibromas, basal cell carcinomas (BCC), pigmented
Bowen disease, pigmented actinic keratoses, and
pigmented inflammatory conditions.4-14It is still
AUC: area under the receiver operating
basal cell carcinoma
LPLK: lichen planuselike keratoses
From the School of Medicine, University of Queensland, Brisbanea;
and Department of Dermatology, Division of General Derma-
tology, Medical University of Vienna.b
Funding sources: None.
Conflicts of interest: None declared.
Accepted for publication March 31, 2010.
Reprint requests: Harald Kittler, MD, Department of Dermatology,
Division of General Dermatology, Medical University of Vienna,
Wa ¨hringer Gu ¨rtel 18-20, 1090 Vienna, Austria. E-mail: harald.
Published online March 28, 2011.
ª 2010 by the American Academy of Dermatology, Inc.
the diagnostic accuracy for these lesions.
The primary aims of this study were to: (1) deter-
mine the proportions of equivocal melanocytic and
of both melanoma and nonmelanoma skin cancer,
using a consecutive series comprising all pigmented
skin lesions submitted for
histology; and (2) analyze
whether the diagnostic accur-
acy for all pigmented malig-
nant neoplasms increases if
dermatoscopy is used in ad-
dition to inspection with the
The cases collected were
from a consecutive series of
from the primary care skin
C. R. As all lesions scheduled
for biopsy in this practice are
routinely photographed, clini-
cal and dermatoscopy images
were available for all lesions.
Clinical images (overview and close-up) were taken
with Canon EOS digital single lens reflex (SLR) cam-
eras (Canon, Tokyo, Japan). The close-up was taken
using a macro lens (60-mm f2.8 macro, Canon) with
diffuse illumination at a constant reproduction ratio
determined by a custom-fabricated spacer. The de-
gree of magnification of the close-up images was
similar to that of the dermatoscopy images.
preferentially using the Dermlite Fluid device (3 Gen,
San Juan Capistrano, CA); alternatively Dermlite Foto
(custom nonpolarized) (3 Gen) and Heine Delta 20
devices (Heine, Optotechnic GmbH, Herrsching,
Germany) were used for large and inaccessible le-
sions, respectively. Dermatoscopic photographs were
taken with Canon EOS SLR cameras (Canon). Images
were presented to the assessors as PowerPoint slides
(Microsoft Corporation, Redmond, WA).
were biopsied or excised and 24% (n = 466) of these
were pigmented. Three were excluded from the study
because of poor image quality leaving 463 suitable for
study. The study was approved by the local ethics
committee of the University of Queensland, Australia.
Histopathologic diagnosis and classification of
All lesions were excised or biopsied and subjected
to standard histopathologic examination. Histopatho-
logy was regarded as the diagnostic gold standard. If
the histopathologic diagnosis was not in line with the
clinical or dermatoscopic diagnosis, the pathologist
was asked to re-examine the specimen. Although
additional deeper cuts were requested on several
specimens, this did not result in any changes of
We categorized the cases
into benign and malignant
lesions. The malignant group
included melanomas, BCC,
and squamous cell carcino-
mas. All other lesions were
benign. We classified actinic
keratoses (n = 14) as a super-
ficial variant of cutaneous
thus malignant. Melanocytic
nevi and melanomas were
classified as melanocytic; all
other lesions were classified
analysis is suitable for the detection of
melanoma and nonmelanoma skin
Presentation of images
For each lesion a triplet
images, a pair of clinical images (overview and
close-up), followed by a dermatoscopic image was
presented to a blinded observer on a computer
had to give a diagnosis (only one diagnosis was
allowed) and was asked to indicate the level of
confidence that the lesion presented was benign or
continuous rating scale ranging from 0 (absolute
certainty that the lesion is benign) to 100 (absolute
asymmetry of structure and color (‘‘chaos’’) and for
clues to malignancy. Asymmetry of structure and
color were defined according to the basic principles
of pattern analysis as revised by Kittler.15Clues to
malignancy included: eccentric structureless zone
(any color except skin color), gray or blue structures,
peripheral black dots or clods, segmental radial lines
or pseudopods, polymorphous vessels, white lines,
ridges (acral lesions).
Continuous variables are given as mean and SD.
false-positive rates (FPR), and true-negative rates
d It is unknown whether dermatoscopy
improves the diagnostic accuracy for all
types of pigmented skin lesions or only
for those that are melanocytic.
d In this series of 463 cases including 217
(46.9%) nonmelanocytic lesions we show
that the addition of dermatoscopy to
clinical examination particularly
improves the diagnostic accuracy for
nonmelanocytic pigmented lesions.
d A simple algorithm based on pattern
J AM ACAD DERMATOL
VOLUME 64, NUMBER 6
Rosendahl et al 1069
(specificity) according to standard formulas. We
applied the McNemar test for the comparison of
paired proportions. We used the level of confidence
reported by the observers to obtain receiver operat-
ing characteristic curves and calculated area under
the receiver operating characteristic curve (AUC) to
compare the diagnostic accuracy with and without
dermatoscopy. ROCKIT software (Kurt Rossmann
Chicago, Chicago, IL) was used for all calculations
involving receiver operating characteristic curves
and SPSS 16.0 software (SPSS, Chicago, IL) was
used for all other statistical calculations.16All given
P values are 2-tailed and a P value less than .05
indicates statistical significance.
The final sample consisted of 463 pigmented skin
age of 57years (SD: 17 years).The majority of lesions
were located on the trunk (n = 241, 52.1%) followed
by the extremities (n = 128, 27.6%), head and face
(n = 82, 17.7%), and palms and soles (n = 10, 2.2%).
Histopathologically, 246 (53.1%) pigmented lesions
turned out to be melanocytic and 217 (46.9%) were
of nonmelanocytic origin. Of 246 melanocytic le-
sions, 29 (11.8%) were diagnosed as melanomas
histopathologically and 217 (88.2%) as melanocytic
nevi. Twenty (68.9%) melanomas were in situ and 9
were invasive, of which only one had a tumor
thickness of more than 1 mm. Nonmelanocytic
lesions (n = 217) were benign in 49.8% (n = 108)
and malignant in 50.2% (n = 109). The frequencies of
all histopathologic diagnoses are given in detail in
Diagnostic accuracy with and without
The observer’s single best diagnosis matched the
histopathologic diagnosis in 80.1% (n = 375) of cases
with dermatoscopy and in 69.1% (n = 320) without
malignant neoplasms (including melanoma and non-
melanoma skin cancer) was measured as AUC. The
AUC achieved with dermatoscopy was 0.89 and
significantly higher than without dermatoscopy
(AUC = 0.83, P \.001) (Fig 1). Given a fixed true-
negative rate (specificity) of 80% the corresponding
true-positive rate (sensitivity) was 82.6% with derma-
toscopy, and 70.5% without it. When only nonmela-
nocytic lesions were considered the AUC was 0.91
with dermatoscopy and 0.84 without it (P \ .001).
With regard to melanocytic lesions the AUC was 0.76
with dermatoscopy and 0.71 without. This difference
was statistically not significant (P = .22).
Accuracy of a short algorithm based on pattern
The overall sensitivity of the proposed short
algorithm to detect any type of malignancy (mela-
nomas and nonmelanoma skin cancer) was 90.6%
with a specificity of 62.7%. The algorithm reached a
sensitivity of 98.6% for BCC, 86.5% for pigmented
squamous cell carcinoma, and 79.3% for melanoma
(Table II). With regard to benign conditions the
Table I. Frequencies of cases according to
Type of lesionN (%)
Basal cell carcinoma
Squamous cell carcinoma
Lichen planuselike keratosis
Ink spot lentigo
Fig 1. Receiver operating characteristic curves for diag-
nostic accuracy achieved with dermatoscopy (solid line)
and naked eye examination (dotted line).
J AM ACAD DERMATOL
1070 Rosendahl et al
highest FPR was observed for LPLK (n = 21, FPR:
90.5%). The FPR for melanocytic nevi was 25.8%,
which corresponds to a specificity of 74.2%.
Although previous studies have demonstrated
that dermatoscopy improves the diagnostic accuracy
for pigmented melanocytic lesions we have not
discovered any prior study where dermatoscopy
has been tested for the stated intention of assessing
pigmented skin lesions (both melanocytic and non-
melanocytic) for malignancy of any type, whether
melanoma, pigmented BCC, or pigmented squa-
mous cell carcinoma. Most previous studies focused
exclusively on the differentiation between mela-
nonmelanocytic skin lesions.17-19
analysis by Kittler et al,214 of 27 studies did not
include nonmelanocytic lesions and if nonmelano-
cytic lesion were included their number was low. Of
those that did include nonmelanocytic lesions, the
studies evaluated diagnostic accuracy with and with-
out dermatoscopy with respect to the diagnosis of
melanoma only and excluded malignant nonmela-
nocytic lesions from calculations of sensitivity and
Vestergaard et al1and by Bafounta et al3diagnostic
accuracy was calculated for melanomas only.
In this study of 463 lesions we have shown that
dermatoscopy improves diagnostic accuracy for all
types of pigmented lesions, including nonmelano-
cytic lesions (Fig 1). Interestingly, although diagnos-
naked-eye assessment for both melanocytic and
nonmelanocytic lesions, this was statistically signif-
icant only for nonmelanocytic lesions. In other
words, dermatoscopy adds more useful information
when it comes to the diagnosis of nonmelanocytic
lesions. We think that several reasons contributed to
this result. First, there is a higher similarity in the
In the meta-
dermatoscopic appearance of melanocytic nevi and
early melanoma than for BCC or pigmented squa-
mous cell carcinoma and benign nonmelanocytic
lesions. For this reason it is easier to differentiate a
BCC from a seborrheic keratosis than a melanocytic
nevus from a melanoma in situ by dermatoscopy.
Second, selection bias, that is the distortion of the
estimates of the diagnostic accuracy by selection of
lesions for dermatoscopic examination, and verifi-
cation bias, that is the distortion of the estimates of
diagnostic accuracy by selection of lesions for bi-
opsy, may have influenced the results obtained for
melanocytic and nonmelanocytic lesions in different
ways. Third, thehighfrequencyof insitumelanomas
(69%) reflects the high diagnostic difficulty of the
melanomas included in the sample. It has been
shown that dermatoscopy has limitations regarding
the diagnosis of some early melanomas when clues
to melanoma are absent. In this respect information
beyond dermatoscopy, for example the clinical ap-
pearance in comparison with other lesions nearby
(ugly duckling sign) and information with regard to
anatomic site (site of lesion in comparison with
distribution of similar lesions) may provide more
useful information for the diagnosis of melanoma
than for the diagnosis of malignant nonmelanocytic
lesions. Fourth, the relative low prevalence of mel-
anomas in comparison with BCC and squamous cell
carcinomas may be responsible for the lack of
statistical significance of the improvement in the
diagnostic accuracy achieved with dermatoscopy.
Finally, the Australian population is characterized by
a high incidence of heavily sun-damaged skin on
people of Celtic origin. Frequently this leads to a
dense covering of solar lentigines. In regression
these are LPLK and the presence of gray dots
dermatoscopically gives a differential diagnosis that
In this study clinical images were prepared in a
standardized fashion by one operator with careful
attention to quality. The clinical images had a similar
magnification as the dermatoscopic images (Fig 2).
The authors speculate that this may have improved
diagnostic accuracy based on clinical images com-
pared with previous studies and that without this
factor the improvement in diagnostic accuracy with
dermatoscopy may have been even greater. Another
previous studies have been almost exclusively based
on referral centers. Of the 36 studies in the 3 major
meta-analyses only one, by Argenziano et al,20drew
specimens from a primary care setting.
It may be considered that specimens from a
primary care center are more representative for the
Table II. Accuracy of short algorithm based on
pattern analysis according to lesion type
Basal cell carcinoma
Squamous cell carcinoma
Lichen planuselike keratosis
J AM ACAD DERMATOL
VOLUME 64, NUMBER 6
Rosendahl et al1071
dermatologist. In this respect it is very relevant that in
our study of sequential lesions, 46.9% were non-
melanocytic. This underlines the importance of the
lesions, including those that are nonmelanocytic.
The diagnostic system we used for dermatoscopic
examination in this study is pattern analysis.15We
referred to pattern analysis because it does not
differentiate between melanocytic and nonmelano-
every pigmented lesion including those that are
nonmelanocytic. We also tested ashortand relatively
simple andeasily appliedalgorithmbasedonpattern
analysis. The short algorithm assesses each lesion for
‘‘chaos’’ (asymmetry of color or structure) and only
when chaos is recognized are ‘‘clues’’ searched for. It
is similar to the rule of Menzies et al21that has been
used for melanocytic lesions. We found that this
system has arelativelyhigh sensitivity of90.6% when
applied in this way to diagnose malignancy of any
type. Because of verification bias, which occurs
because only suspicious lesions are selected for
biopsy, the specificity is low for some types of
nonmelanocytic lesions such as LPLK. This is also
because of the fact that not every aspect of derma-
toscopy can be captured with a simple short algo-
rithm. To fully master a diagnostic procedure such as
dermatoscopy, a complete diagnostic system, such
as pattern analysis, is needed to make specific
diagnoses and not just to differentiate between
benign and malignant. In this regard we also dem-
onstrated that dermatoscopy is more specific than
examination with the unaided eye. The single best
diagnosis given by dermatoscopy matched the path-
ologic diagnosis in a higher percentage of cases than
with naked eye examination.
This study demonstrates that dermatoscopy
improves the diagnostic accuracy for malignancy in
a series that included both melanocytic and non-
melanocytic lesions and that the improvement was
even more pronounced for nonmelanocytic lesions.
A short algorithm based on pattern analysis (chaos
and clues) is easy to apply and achieves a high
sensitivity in diagnosing malignancy of any type.
This algorithm may be used as an entry-level system
for novices in dermatoscopy but we recommend the
use of a complete diagnostic system such as pattern
analysis that allows specific diagnoses as the opti-
1. Vestergaard ME, Macaskill P, Holt PE, Menzies SW. Dermo-
scopy compared with naked eye examination for the diagno-
sis of primary melanoma: a meta-analysis of studies performed
in a clinical setting. Br J Dermatol 2008;159:669-76.
2. Kittler H, Pehamberger H, Wolff K, Binder M. Diagnostic
accuracy of dermoscopy. Lancet Oncol 2002;3:159-65.
3. Bafounta ML, Beauchet A, Aegerter P, Saiag P. Is dermoscopy
(epiluminescence microscopy) useful for the diagnosis of
melanoma? Results of a meta-analysis using techniques
adapted to the evaluation of diagnostic tests. Arch Dermatol
4. Cameron A, Rosendahl C, Tschandl P, Riedl E, Kittler H.
Dermatoscopy of pigmented Bowen’s disease. J Am Acad
5. Altamura D, Menzies SW, Argenziano G, Zalaudek I, Soyer HP,
Sera F, et al. Dermatoscopy of basal cell carcinoma: morpho-
logic variability of global and local features and accuracy of
diagnosis. J Am Acad Dermatol 2010;62:67-75.
Fig 2. Set of images of pigmented basal cell carcinoma: clinical overview (A), clinical close-up
(B), and dermatoscopy (C).
J AM ACAD DERMATOL
1072 Rosendahl et al
6. Menzies SW. Dermoscopy of pigmented basal cell carcinoma.
Clin Dermatol 2002;20:268-9.
7. Zalaudek I, Ferrara G, Leinweber B, Mercogliano A, D’Ambrosio
A, Argenziano G. Pitfalls in the clinical and dermoscopic
diagnosis of pigmented actinic keratosis. J Am Acad Dermatol
8. Braun RP, Rabinovitz HS, Krischer J, Kreusch J, Oliviero M, Naldi
L, et al. Dermoscopy of pigmented seborrheic keratosis:
a morphological study. Arch Dermatol 2002;138:1556-60.
9. Argenziano G, Soyer HP, Chimenti S, Talamini R, Corona R, Sera
F, et al. Dermoscopy of pigmented skin lesions: results of a
consensus meeting via the Internet. J Am Acad Dermatol
10. Tschandl P, Argenziano G, Bakos R, Gourhant JY, Hofmann-
Wellenhof R, Kittler H, et al. Dermoscopy and entomology
(entomodermoscopy). J Dtsch Dermatol Ges 2009;7:589-96.
11. Zaballos P, Daufi C, Puig S, Argenziano G, Moreno-Ramirez D,
Cabo H, et al. Dermoscopy of solitary angiokeratomas: a
morphological study. Arch Dermatol 2007;143:318-25.
12. Zaballos P, Puig S, Llambrich A, Malvehy J. Dermoscopy of
dermatofibromas: a prospective morphological study of 412
cases. Arch Dermatol 2008;144:75-83.
13. Kilinc KI, Gencoglan G, Akalin T, Ozdemir F. Different dermo-
scopic faces of dermatofibromas. J Am Acad Dermatol 2007;
14. Puig S, Romero D, Zaballos P, Malvehy J. Dermoscopy of
dermatofibroma. Arch Dermatol 2005;141:122.
15. Kittler H. Dermatoscopy: introduction of a new algorithmic
method based on pattern analysis for diagnosis of pigmented
skin lesions. Dermatopathology:
16. Metz CE, Herman BA, Roe CA. Statistical comparison of two
ROC-curve estimates obtained from partially-paired datasets.
Med Decis Making 1998;18:110-21.
17. Nachbar F, Stolz W, Merkle T, Cognetta AB, Vogt T, Landthaler
M, et al. The ABCD rule of dermatoscopy: high prospective
value in the diagnosis of doubtful melanocytic skin lesions.
J Am Acad Dermatol 1994;30:551-9.
18. Argenziano G, Fabbrocini G, Carli P, De Giorgi V, Sam-
marco E, Delfino M. Epiluminescence microscopy for the
diagnosis of doubtful melanocytic skin lesions: comparison
of the ABCD rule of dermatoscopy and a new 7-point
checklist based on pattern analysis. Arch Dermatol 1998;
19. Zalaudek I, Argenziano G, Soyer HP, Corona R, Sera F, Blum A,
et al. Three-point checklist of dermoscopy: an open Internet
study. Br J Dermatol 2006;154:431-7.
20. Argenziano G, Puig S, Zalaudek I, Sera F, Corona R, Alsina M,
et al. Dermoscopy improves accuracy of primary care physi-
cians to triage lesions suggestive of skin cancer. J Clin Oncol
21. Menzies SW, Ingvar C, McCarthy WH. A sensitivity and spec-
ificity analysis of the surface microscopy features of invasive
melanoma. Melanoma Res 1996;6:55-62.
J AM ACAD DERMATOL
VOLUME 64, NUMBER 6
Rosendahl et al1073