Vol. 88 No.

2 August 1999

ORAL SURGERY
ORAL MEDICINE
ORAL PATHOLOGY

ORAL AND MAXILLOFACIAL RADIOLOGY

Editor: Sharon L. Brooks

Statistical study for sonographic differential diagnosis of

tumorous lesions in the parotid gland
Mayumi Shimizu, DDS, PhD,a Jrgen Ussmller, Dr Med,b Joerg Hartwein, Dr Med,c Karl
Donath, Dr Med, Dr HC,d and Naoko Kinukawa, MS,e Fukuoka, Japan, and Pforzheim and
Hamburg, Germany
KYUSHU UNIVERSITY, SILOAH HOSPITAL, AND HAMBURG UNIVERSITY

Objective. The purpose of this study was to clarify characteristic sonomorphologic features of parotid lesions statistically and
to propose new criteria for the differential diagnosis.

Study design. Eighty-six tumorous lesions were analyzed with regard to the following sonomorphologic features: boundary,
shape, echo intensity level, distribution of internal echoes, and acoustic enhancement. Stepwise polychotomous logistic
regression analysis was performed to assess characteristic sonographic features. As dependent variables, we used pleomorphic
adenoma, Warthin tumor, malignant tumors and other benign lesions; as predictor variables, we used the aforementioned sonomorphologic features. Proportion of the occurrence of each dependent variable was calculated.
Results. Lobular shape and homogeneous internal echoes predicted pleomorphic adenoma. A lesion with multiple anechoic
areas would be Warthin tumor with very high sensitivity. Malignant tumors showed either heterogeneous internal echoes
without characteristic structures or polygonal shape.
Conclusions. These sonomorphologic features should be observed to make more exact differential diagnoses for operation
and therapy planning.

(Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1999;88:226-33)

Many kinds of imaging techniques have been developed for diagnosing salivary gland diseases, including
plain radiography, sialography, scintigraphy, computed
tomography (CT), CT-sialography, magnetic resonance
imaging (MRI), and ultrasonography. In recent years,
ultrasonography has become one of the most important
imaging techniques because of its ease of use and
absence of ionizing radiation. Ultrasonography, when
This work was supported in part by DAAD (Deutscher
Akademischer Austausch-dienst) and carried out at University
Hospital Eppendorf, Hamburg University.
aInstructor, Department of Oral & Maxillofacial Radiology, Faculty
of Dentistry, Kyushu University.
bPrivatdozent and Oberarzt, Ear-Nose-Throat Clinic, University
Hospital Eppendorf, Hamburg University.
cProfessor and Chairman, Ear-Nose-Throat Clinic, Siloah Hospital.
dProfessor and Chairman, Department of Oral Pathology, Institute
for Pathology, Hamburg University.
eInstructor, Department of Medical Information Science, Faculty of
Medicine, Kyushu University.
Received for publication Nov 2, 1998; returned for revision Dec 19,
1998; accepted for publication Feb 27, 1999.
Copyright 1999 by Mosby, Inc.
1079-2104/99/$8.00 + 0 7/16/99050

226

used with high-frequency sonic waves, is reported to

be superior to other modalities such as CT and sialography not only in detecting tumorous lesions1-5 but also
in describing the structure and even the vascularity of
such lesions.6
There are many reports about the characteristic sonographic features of pleomorphic adenoma, Warthin
tumor, lipoma, cyst, and lymphadenitis.1,3,7-11 Some
authors have reported that it is possible to make correct
diagnoses to some extent by means of pathognomonic
findings on sonograms alone.2,7 However, there are 4 or
5 features for each kind of lesion; because none of them
are specific, we have to combine them to make diagnoses. If 1 or 2 features are not typical for the lesion,
there is often controversy over which feature should be
used for diagnosing. Moreover, the difficulty in distinguishing pleomorphic adenoma from lymphadenitis has
been pointed out,7,8 and some malignant tumors may be
falsely diagnosed as benign tumors.4,7,9,12,13 Some
authors state that there are only characteristic, not
specific, sonomorphologic findings.9,10,14 Thus far, it
remains unclear how differential diagnoses of parotid

Shimizu et al 227

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

Volume 88, Number 2

Table I. Features on sonograms

Features

Classes

Boundary

Very clear
Relatively clear
Partially unclear
Oval
Lobular
Polygonal
Glandular parenchymal
slightly hypoechoic
Very hypoechoic
Homogeneous
Multiple anechoic areas
Heterogeneous with characteristic structures
Heterogeneous without characteristic structures
Enhanced
Unchanged
Attenuated

Shape

Echo intensity level

Distribution of internal echoes

Acoustic enhancement
(posterior echoes)

lesions can be made by means of ultrasonography and

how exact the differential diagnoses may be. Therefore,
we performed our study to establish simple and reliable
criteria using statistical analysis, which has not been
done thus far.
There are many kinds of tumorous lesions in the
parotid region. Pleomorphic adenomas have a tendency
to recur if the capsules are damaged at surgery; for that
reason, partial parotidectomy is suitable for such
tumors. On the other hand, enucleation is sufficient for
Warthin tumors. Lipomas and lymphadenitis can be
observed without further treatment. Thus, the operation
and therapy planning depend on the diagnosis. We have
to make more exact differential diagnoses, not just
differentiate malignant lesions from benign ones. If
which sonographic features should be used can be clarified, it becomes possible to make more exact diagnoses. For this reason, we used stepwise polychotomous logistic regression analysis, even though the
sample sizes for most lesion types were small.
The purposes of this retrospective study were (1) to
clarify characteristic sonomorphologic features of tumorous lesions of the parotid gland by statistical analysis on
sonographic findings and (2) to propose certain criteria
for making sonographic differential diagnoses.

MATERIALS AND METHODS

Patients
Eighty-six tumorous parotid gland lesions in 84
patients (at the Ear-Nose-Throat Clinic of Hamburg
University from November 1993 to August 1995) were
analyzed retrospectively. Two separate lesions were

Table II. Histopathologic diagnoses and numbers of

cases
Diagnosis
Benign lesions
Pleomorphic adenoma
Warthin tumor
Other benign lesions
Lipoma
Basal cell adenoma
Cyst
Lymphadenitis
Miscellaneous
Myoepithelial parotitis
Epithelioid cellular parotitis
Angio-lymphoid hyperplasia
Cystic lymphoid hyperplasia in AIDS
Malignant tumors
Acinic cell carcinoma
Cystadenocarcinoma
Mucoepidermoid carcinoma
Salivary duct carcinoma
Fibrosarcoma
Malignant lymphoma
Metastatic tumor

No. of cases
72
22
30
20
4
2
4
5
5
1
1
1
2
14
2
2
2
1
1
2
4

found in each of 2 patients. The patients included 50

males and 34 females ranging in age from 13 to 83
years (mean, 50.6 years).

Sonographic features
Before surgical intervention, ultrasonography was
performed with a linear small-parts electronic sonographic scanner (7.5 MHz center frequency; Quantum
2000, Siemens, Erlangen, Germany). Both transversal
and longitudinal scan sections were obtained at the
greatest dimensions of the lesions. The contralateral
(normal) parotid glands were also examined.
Sonographic features of the lesions are listed in Table
I. With regard to boundaries, if a lesion had either a
thin hyperechoic line on the anterior side or a capsulelike structure, it was categorized as very clear. If a
contour showed any interruption, it was classified as
partially unclear. For echo intensity level, if a lesion
showed approximately the same echo intensity level as
glandular parenchyma, it was termed glandular
parenchymal level. If anechoic areas in a lesion
accounted for less than 50% of the lesion, it was categorized as slightly hypoechoic; if more than 50%, as
very hypoechoic. For the distribution of internal
echoes, we categorized multiple anechoic areas as a
specific class. We categorized other characteristic findings of internal echoes, such as a hilus and regularly
distributed hyperechoic lines, as heterogeneous with
characteristic structures. Acoustic enhancement was

228 Shimizu et al

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

August 1999

Fig 1. Ultrasonography of pleomorphic adenoma. Lobular

shape and homogeneous internal echoes are observed.
Predictive proportion for pleomorphic adenoma was 73.9%.

Fig 3. Ultrasonography of malignant tumor (mucoepidermoid

carcinoma, poorly differentiated type). Polygonal shape
(pointed portion of arrow) and heterogeneous internal echoes
without characteristic structures are observed. Predictive
proportion for malignant tumor was 100%.

Fig 2. Ultrasonography of Warthin tumor. Lesion shows oval

shape and multiple anechoic areas (arrows). Predictive
proportion for Warthin tumor was 94.7%.

estimated by comparison with the echo intensity level

in the anterior part of the normal parotid gland. If there
was an internal shadow that caused partial attenuation
of the posterior echoes, we classified the case as attenuated (that is, attenuated posterior echoes).

Statistical methods
Stepwise polychotomous logistic regression analysis
was performed to assess characteristic sonographic features of the various parotid diseases to determine which
could be used as the criteria for sonographic differential
diagnoses. We used a BMDP PR module (BMDP Statistical Software Inc, Los Angeles, Calif) for data analysis.
The dependent variable, Y, had 4 groups: pleomorphic
adenoma (Pleo), Warthin tumor (Warth), malignant
tumors (Malig), and other benign lesions (Others).
As candidates for independent (predictor) variables, we
used the 5 features on sonograms listed in Table I:
boundary, shape, echo intensity level, distribution of
internal echoes, and acoustic enhancement. Categories of
independent variables were also made on the basis of the
classes in Table I. The design variables were generated
from independent variables by means of the partial
method in the BMDP PR. The reference categories for
the 5 variables were very clear, oval, glandular
parenchymal slightly hypoechoic, homogeneous, and
enhanced, respectively. Therefore, the exponentiated
value of the coefficient for each category could be
considered an odds ratio to the reference category.
We analyzed 78 cases for this statistical analysis. We
excluded 2 cases that could be diagnosed as malignant
tumors only by attenuated posterior echoes (including
internal shadow)15 and another 6 cases whose shape or
enhancement we could not estimate.

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

Volume 88, Number 2

Shimizu et al 229

We defined p as proportion of the occurrence for

each group of a dependent variable Y, which is either
Pleo, Warth, or Malig; p(Y) is calculated as follows:

whereas 21 (70.0%) of 30 cases of Warthin tumor

showed very clear boundaries. Malignant tumors
showed partially unclear boundaries in 7 (50.0%) of 14
cases. The only other case that had an unclear
boundary was one Warthin tumor.
Most of the pleomorphic adenomas (20 of 22 cases;
90.9%) and lipomas (3 of 4 cases; 75.0%) and 50% of
the malignant tumors (7 of 14 cases) had lobular
shapes. On the other hand, Warthin tumors (19 of 30
cases; 63.3%), cysts (3 of 4 cases; 75.0%) and lymphadenitis cases (all 5 cases; 100%) had oval shapes.
Polygonal shape was observed in 4 (28.6%) of 14 cases
of malignant tumors and in 2 (6.7%) of 30 cases of
Warthin tumor.
Very hypoechoic internal echoes were observed in 18
(60.0%) of 30 cases of Warthin tumor and in 3 (75.0%)
of 4 cases of cyst. Pleomorphic adenomas (20 of 22
cases; 90.9%), lipomas (2 of 4 cases; 50%), and malignant tumors (12 of 14 cases; 85.7%) had slightly
hypoechoic internal echoes. Each of 2 cases of lipoma
had a glandular parenchymal echo intensity level.
Multiple anechoic areas were observed in 28 (93.3%)
of 30 cases of Warthin tumor and in both cases of cystic
lymphoid hyperplasia in AIDS. Most pleomorphic
adenomas (20 of 22 cases; 90.9%) and cysts (3 of 4
cases; 75.0%) had homogeneous internal echoes. All 4
cases of lipoma had characteristic hyperechoic lines
regularly distributed in the lesions. Four of 5 cases
(80.0%) of lymphadenitis had a hilus in the lesions.
Malignant tumors had heterogeneous internal echoes in
10 (71.4%) of 14 cases, 6 cases of which were heterogeneous without characteristic features.
Attenuated posterior echoes were observed in only 2
cases of malignant tumors (14.3% of malignant tumors
and 2.3% of all lesions). Most pleomorphic adenomas
(16 of 22 cases; 72.7%) and cysts (3 of 4 cases; 75.0%)
had enhanced posterior echoes. On the other hand,
Warthin tumors (19 of 30 cases; 63.3%), lipomas (2 of
3 cases; 66.7%) and 50% of malignant tumors (7 of 14
cases) had unchanged posterior echoes.

p(Y) =

uY =

exp(uY)/[1 + exp(uPleo) + exp(uW

+exp(uMalig)]
Y + b Y (X ) + b Y (X ) + . . . + b Y (X ),
bXO
X1 1
X2 2
Xk k

where Xi, i = 1, 2, . . . k, are independent variables, xi,

i = 1, 2, . . . k, are categories of independent variables,
Y (Xi) are regression coefficients corresponding to Xi
b Xi
Y is constant.16 For example, if a
for group Y, and bXO
lesion showed a very clear boundary, oval shape, very
hypoechoic echo intensity level, homogeneous internal
echoes, and enhanced posterior echoes and if shape and
distribution of internal echoes remained as independent
variables in the final logistic regression model, then
p(Pleo) is calculated as follows:
p(Pleo) = exp(uPleo)/[1 + exp(uPleo) + exp(uWarth) +
exp(uMalig)]
uPleo =

Pleo
b Pleo
XO + b shape(oval) +
Pleo (homogeneous).
b distribution

The equation for p(Others) is

p(Others) =

1 [p(Pleo) + p(Warth) + p(Malig)].

To select an effective set of predictors for differential diagnoses from the 5 variables, stepwise analysis
was performed as a forward-stepping procedure with
a P value of less than .05 for variable inclusion and
greater than .06 by likelihood ratio test for exclusion
from the model.

RESULTS
Histopathologic diagnoses of the lesions
Table II contains the histopathologic results for
the lesions. There were 72 cases of benign lesions,
58 of which were benign tumors (67.4% of all lesions).
Most were pleomorphic adenomas (22 cases, 37.9% of
benign tumors, 25.6% of all lesions) or Warthin tumors
(30 cases, 51.7% of benign tumors, 34.9% of all
lesions). Figs 1 and 2 are representative cases of pleomorphic adenoma and Warthin tumor, respectively.
Four cases of lipoma and 2 cases of basal cell adenoma
were also seen as benign tumors. We had 14 cases of
various malignant tumors (16.3% of all lesions; Fig 3).
For the other benign lesions, 4 cases of cyst and 5 cases
of lymphadenitis were observed.
Sonomorphology of the tumorous lesions
Table III contains the sonomorphologic results for
the tumorous lesions. Pleomorphic adenoma showed
relatively clear boundaries in 17 (77.3%) of 22 cases,

Statistical analysis
In the final logistic regression model, 2 variables
were selected as independent variables: distribution of
internal echoes and shape (P < .0001 and P = .0007 for
likelihood ratio test, respectively). The BMDP PR
module provided coefficient, standard error (SE), coefficient/SE, odds ratio, and 95% confidence interval to
reference category for each independent variable. Table
IV contains coefficient and SE for pleomorphic
adenomas, Warthin tumors, and malignant tumors.
Though SE for some independent variables showed
very wide range because of the small number of cases,
the P value was less than .05 by likelihood ratio test.

230 Shimizu et al

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

August 1999

Table III. Sonomorphology of lesions

Lesions shapes
Histopathologic
diagnoses
(no. of cases)
Pleomorphic adenoma (22)
Warthin tumor (30)
Other benign lesions (20)
Lipoma (4)
Basal cell adenoma (2)
Cyst (4)
Lymphadenitis (5)
Miscellaneous (5)
Malignant tumors (14)
Totals (86)

Very
clear

Boundaries
Relatively
clear

Partially
unclear

Oval

5
21

17
8

2
19

3
1
3
4
3
2
42

1
1
1
1
2
5
36

7
8

1
1
3
5
1
2
34

Lobular

Polygonal

Unable
to
classify

20
8

3
1
1

3
7
43

4
6

1
1
3

Hetero 1, Heterogeneous with characteristic structures; Hetero 2, heterogeneous without characteristic structures.

Table V shows predictive proportions (percentages)

that were calculated with coefficients in Table IV. For
example, if we have a lesion that shows lobular shape
and multiple anechoic areas, then:
uPleo =

Pleo
b Pleo
XO + bshape(lobular) +
Pleo
b distribution
(multiple anechoic areas) =
-0.1612 + 1.870 + (-19.75),

uWarth =

-0.05095 + (-0.9848) + 2.932,

uMalig =

-0.5108 + 0 + (-25.67),

p(Pleo) = exp(uPleo)/[1 + exp(uPleo) +exp(uWarth)

+ exp(uMalig)] < 0.001,
p(Warth) =

0.869,

p(Malig) <

0.001,

and p(Others) = 1 [p(Pleo) + p(Warth) + p(Malig)]

= 0.131.
Therefore, this lesion could be predicted to be a
Warthin tumor, because it takes the maximum proportion. Table V shows that the predictive proportion for
pleomorphic adenoma was high only when the lesion
was lobular and homogeneous (Fig 1). On the other
hand, if we found multiple anechoic areas in a lesion
(Fig 2), the possibility of Warthin tumor was very high
(greater than 86.9%), regardless of the shape of the
lesion. Polygonal shape (Fig 3) predicted malignant
tumors with very high proportion, if the lesion showed
no multiple anechoic areas. Heterogeneous internal
echoes without characteristic structures (Fig 3) also
predicted malignant tumors with very high proportion,
regardless of the shape of the lesions.
Table VI contains sensitivity, specificity, accuracy, and
true positive rates for each dependent variable. Sixty-four
of 78 cases were diagnosed correctly (accuracy, 82.1%).

DISCUSSION
Ultrasonography has become the preferred diagnostic
method for tumorous lesions in the salivary glands. If we
could diagnose lesions precisely by means of ultrasonography, it would be of great value for treatment and operation planning. Inasmuch as there are many kinds of tumorous lesions, especially in parotid glands, and some of
them show characteristic sonomorphology, it should be
possible to make reliable diagnostic criteria for parotid
tumorous lesions. Although many authors1-5,7-15 have
reported ultrasonic differential diagnoses for the salivary
gland diseases, their criteria have not been used widely,
because the diagnoses were made empirically in the
authors own hospitals and are not uniform. Complexes
of characteristic findings of some representative diseases
were stated, and sensitivity, specificity, and accuracy in
each disease were calculated; however, because no statistical analyses were performed, the most important factors
for differential diagnoses have remained obscure. Therefore, we performed a statistical study to clarify which
sonographic features should be emphasized in the criteria
for differential diagnoses of parotid tumorous lesions.
We used 5 sonographic features (Table I): boundary,
shape, echo intensity level, distribution of internal
echoes, and acoustic enhancement.1,4 Other authors
have used only the first 4 of these features.7,8,11 However,
because attenuated posterior echoes in acoustic
enhancement have been stated to be one of the malignant signs,3,15 we used the former series of features.
Sonomorphologic features of pleomorphic adenoma
were reported in the literature as follows: clear
boundary,3,7,10 lobulated shape,1 slightly hypoechoic
echo intensity level,1,7,10 homogeneous internal
echoes,3,7,8,10,11 and enhanced posterior echoes.3,7,11
Because lobular shape was a main characteristic
feature of pleomorphic adenoma (Table V) and is also
listed among the histopathologic features of pleomor-

Shimizu et al 231

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

Volume 88, Number 2

Echo intensity level

Glandular
parenchymal-slightly
Very
hypoechoic
hypoechoic

Distribution of internal echoes

Multiple
Homo- anechoic Hetero
Hetero
geneous
areas
1
2

Acoustic enhancement

Enhanced Unchanged Attenuated

Unable
to
classify

20
12

2
18

20
2

28

16
11

4
19

4
2
1
3
2
12
56

3
2
3
2
30

1
3
1

4
31

30

4
1
1
4
3
4
19

6
6

3
2
2
5
40

2
2
1
3
3
7
41

2
2

phic adenoma,17 this feature should be emphasized in

differential diagnoses.
The sensitivity of diagnosing pleomorphic adenoma
has been reported in the literature as 71% to 86%.2,7-9,12
In this study, as shown in Table VI, the sensitivity for
pleomorphic adenoma was 80%. Including 2 cases for
which the acoustic enhancement could not be estimated, the sensitivity for pleomorphic adenoma in the
present study was 81.8%. This is acceptable when it is
compared with other reported results.
A wide range of sonomorphologic findings, from
cystlike to solidlike,10,11 has been reported for Warthin
tumors. In our study, we had 1 cystlike case and 1
solidlike case in 30 Warthin tumors; however, the other
28 cases showed multiple anechoic areas. According to
our results (Table V), the finding of multiple anechoic
areas was the main characteristic feature of Warthin
tumor. Such a finding has been reported to be one of
the characteristic sonomorphologic features of Warthin
tumor.3,7-10,12,14 Other reported sonomorphologic
features of Warthin tumor were clear boundary,3,7-10
oval shape,8 very hypoechoic internal echoes,1,9-11 and
weak enhancement of posterior echoes.11
Some authors have reported the sensitivity of diagnosing Warthin tumor (33%12; 70% to 86%2,7-9). Table
VI shows that the sensitivity for Warthin tumor in this
study was 93.1%. We had one case in which the shape
could not be estimated; it was therefore excluded from
the statistical study. Because this case showed multiple
anechoic areas, the sensitivity for Warthin tumor in this
study was actually 93.3%. This high sensitivity
suggests that we applied proper criteria.
The following sonographic features of malignant
tumors have been found to be diagnostically useful: illdefined (irregular) border,1,3,4,7-11,14 polygonal (irregular)
shape,4,8-10 hypoechoic echo intensity level,1,7-11 heterogeneous internal echoes,1,3,4,7-11,15 and attenuated poste-

rior echoes.3 Miwa et al15 mentioned that lesions with

internal shadow were malignant tumors with very high
specificity (98%). Therefore, we diagnosed 2 cases with
attenuated posterior echoes that had internal shadow as
malignant tumors by means of this finding alone. These
2 cases were excluded from the statistical analysis.
Heterogeneous internal echoes without characteristic
structures were one of the most important factors for
malignancy, inasmuch as they predicted malignancy
regardless of the shape of the lesions (Table V). On the
other hand, polygonal shape was also one of the main
characteristic features for diagnosing malignant tumors
(Table V). Histopathologically, the polygonal shape
turned out to be the portion of tumor cell infiltration.
Therefore, this finding should be emphasized in diagnosing malignant tumors. We had 2 cases of Warthin
tumor that also showed polygonal shape. Fine-needle
aspiration biopsy was performed in these cases, and the
capsule in each case showed partial involvement of
tumor cells histopathologically.
Among the other factors, Gritzmann14 emphasized
unclear boundary as a malignant sign. He used only
this criterion for differentiating malignant from benign
tumors. In the present study, only malignant tumors
showed unclear boundaries, the only exception being 1
degenerated Warthin tumor (Table III). Therefore,
unclear boundary seemed to be an important factor for
diagnosing malignant tumors. However, these 7 malignant tumors with unclear boundaries showed either
heterogeneous internal echoes without characteristic
structures or polygonal shape. Our results showed that
using heterogeneous internal echoes without characteristic structures and/or polygonal shape for diagnosis
predicted malignancy more precisely than did using
unclear boundaries (Table V).
In the statistical analysis, 7 (63.6%) of 11 malignant
tumors could be diagnosed correctly (Tables V and VI).

232 Shimizu et al

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

August 1999

Table IV. Results of stepwise analysis

Independent variables
Distribution of internal echoes
Homogeneous
Multiple anechoic areas
Hetero 1
Hetero 2
Shape
Oval
Lobular
Polygonal
Constant

Dependent variables
Warthin tumors
Coef
SE

Pleomorphic adenomas
Coef
SE
0
19.75
2.894
1.017
0
1.870
6.121
0.1612

0
4590
0.958
1.660 l05

0
2.932
20.78
1.265

0
0.943
385
0.901

0
0.9848
8.331
0.0510

Malignant tumors
Coef
SE

0
1.06
7420.
1.560 l05

0
25.67
1.887
26.12

0
1.02
117
0.851

0
0
19.84
0.5108

0
2790.
1.27
1.270 l05
0
0
191
0.730

Coef, Coefficient; SE, standard error.

Table V. Predictive proportions

Independent variables
Shape

Distribution of internal echoes

Lobular
Oval
Lobular
Polygonal
Oval
Lobular
Polygonal
Polygonal
Polygonal
Oval
Oval
Lobular
Totals

Homogeneous
Multiple anechoic areas
Multiple anechoic areas
Multiple anechoic areas
Hetero 2
Hetero 2
Homogeneous
Hetero 1
Hetero 2
Homogeneous
Hetero 1
Hetero 1

Predictive proportion (%)

Pleo Warth Malig Others
73.9
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
25.0
4.1
21.9

4.7
94.7
86.9
100.0
0.0
0.0
0.0
0.0
0.0
27.9
0.0
0.0

8.0
0.0
0.0
0.0
100.0
100.0
100.0
100.0
100.0
17.6
8.0
6.5

13.4
5.3
13.1
0.0
0.0
0.0
0.0
0.0
0.0
29.4
87.9
71.6

No. of cases in present study

Prediction
Pleo
Warth
Warth
Warth
Malig
Malig
Malig
Malig
Malig
Others
Others
Others

Pleo Warth Malig Others Total

16*
0
0
0
0
0
0
0
0
2
0
2
20

1
18*
7*
2*
0
0
0
0
0
1
0
0
29

3
0
0
0
1*
2*
1*
3*
0*
0
0
1
11

2
1
1
0
0
0
0
0
0
3*
6*
5*
18

22
19
8
2
1
2
1
3
0
6
6
8
78

Pleo, Pleomorphic adenoma; Warth, Warthin tumor; Malig, malignant tumors; Others, other benign lesions.
*Case predicted correctly.

We had 2 other cases that could be diagnosed directly by

means of attenuated posterior echoes. The last case, the
shape of which could not be estimated, showed heterogeneous internal echoes without characteristic structures. If we add these 3 cases, the sensitivity for malignant tumors was 71.4%, which was higher than the
sensitivity in other reports (67%,4 57%,9 and 50%12).
As to accuracy in differentiating benign from malignant lesions by means of ultrasonography, rates of 87%,1
90%,4 78%11 and 85%12 have been reported. We misdiagnosed 4 cases of pleomorphic adenoma as other benign
lesions, 2 cases of Warthin tumor as pleomorphic adenoma (1 case) and other benign lesion (1 case), and 4 cases
of other benign lesions as pleomorphic adenoma (2
cases) and Warthin tumor (2 cases; Table V). Although
some benign lesions were misdiagnosed, we did not diagnose them as malignant. On the other hand, 4 cases of
malignant tumors were misdiagnosed as pleomorphic adenoma (3 cases) and other benign lesion (1 case). There-

fore, 74 of 78 cases could be diagnosed correctly as either

benign or malignant (accuracy, 94.9%). This high accuracy suggests that our criteria could be very effective.
As to other benign lesions, some sonomorphologic
characteristics have been pointed out. For lipoma, rather
hyperechoic internal echoes with hyperechoic lines9,12,14
have been reported. As to cysts, very homogeneous,
hypoechoic lesions with very clear boundary and
enhanced posterior echoes4,8-11,14 have been reported. As
to lymphadenitis, the hilus, an eccentric echogenic structure in a hypoechoic lesion, is reported to be one of the
typical sonomorphologies.14 If these characteristics are
considered after diagnosis of an other benign lesion, a
lesion could be diagnosed more precisely.
In conclusion, the results of our statistical analysis
showed important factors to differentiate tumorous
lesions of the parotid gland. Lobular shape and homogeneous internal echoes predicted pleomorphic adenoma. If
a lesion showed multiple anechoic areas, it would be

Shimizu et al 233

ORAL SURGERY ORAL MEDICINE ORAL PATHOLOGY

Volume 88, Number 2

Table VI. Sensitivity, specificity, accuracy, and true positive rates for dependent variables
Dependent variables

Sensitivity

Specificity

Accuracy

True positive rate

Pleomorphic adenoma

80.0 (16/20)
0.557-0.965
93.1 (27/29)
0.758-1.019
63.6 (7/11)
0.306-0.966
77.8 (14/18)
0.519-0.959

93.1 (54/58)
0.824-0.991
95.9 (47/49)
0.848-1.012
100 (67/67)
0.932-1.023
90.6 (58/64)
0.800-0.971

89.7 (70/78)
0.802-0.959
94.9 (74/78)
0.867-0.994
94.9 (74/78)
0.867-0.994
92.3 (72/78)
0.834-0.977

72.7 (16/22)
0.495-0.907
93.1 (27/29)
0.758-1.019
100 (7/7)
0.561-1.150
70.0 (14/20)
0.474-0.926

Warthin tumor
Malignant tumors
Other benign lesions

*Percentage (no. of cases); 95% confidence intervals (lower and upper limits).

Warthin tumor with very high sensitivity. Malignant

tumors showed attenuated posterior echoes, heterogeneous internal echoes, or polygonal shape. The differential diagnosis is important for operation and therapy planning. Therefore, these sonomorphologic features should
be observed so that more exact diagnoses can be made.
We thank Mr A. Pommert (Institute of Mathematics and
Computer Science in Medicine, University Hospital Eppendorf,
Hamburg University, Hamburg, Germany) and Drs K. Araki
(Department of Oral & Maxillofacial Radiology, Faculty of
Dentistry, Showa Dental University, Tokyo, Japan) and K.
Yoshiura (Department of Oral & Maxillofacial Radiology,
Faculty of Dentistry, Kyushu University, Fukuoka, Japan) for
helpful discussion and pertinent comments on the manuscripts.
We thank Ms Tomrlin for photographs and all of our coworkers
in the Department of Oral Pathology and Ear-Nose-Throat
Clinic, University Hospital Eppendorf (Hamburg University,
Hamburg, Germany) for their cooperation.

6.
7.
8.
9.
10.
11.
12.
13.
14.

REFERENCES
1. Wittich GR, Scheible WF, Hajek PC. Ultrasonography of the
salivary glands. Radiol Clin North Am 1985;23:29-37.
2. Wagner W, Bttcher HD, Schadel A, Mllmann M.
Untersuchungen zur Sensitivitt and Spezifitt der Sonographie
und Sialographie bezglich der Diagnostik von Parotistumoren.
Ultraschall Med 1987;8:175-7.
3. Rinast E, Gmelin E, Hollands-Thorn B. Digital subtraction
sialography, conventional sialography, high-resolution ultrasonography and computed tomography in the diagnosis of salivary gland diseases. Eur J Radiol 1989;9:224-30.
4. Akin I, Esmer N, Gerceker M, Aytac S, Erden I. Sialographic
and ultrasonographic analyses of major salivary glands. Acta
Otolaryngol (Stockh) 1991;111:600-6.
5. Landwehr P, Hhmann D, Krahe T, Lackner K. Wertigkeit der
digitalen Subtraktionssialographie im Vergleich mit konventioneller Sialographie, Speicheldrtisensonographie und Oper-

15.

16.
17.

ationsbefund. Fortschritte auf dem Gebeite der Rntgenstrahlen

und der Neuen Bildgebenden Verfahren 1992;156:437-42.
Westhofen M, Reichel C, Nadjmi D. Die farbkodierte
Duplexsonographie der Halslymphknoten. Otorhinolaryngol
Nova 1994;4:285-91.
Haels J, Lenarz T. Ultrasscalldiagnostik benigner und maligner
Parotistutnoren. Laryng Rhinol Otol 1986;65:480-4.
Wagner W, Bttcher HD. Die Abklrung tumorser
Ohrspeicheldrtisen-erkrankungen
durch
Sonographie.
Ultraschall Med 1985;6:341-5.
Klein K, Trk R, Gritzmann N, Traxler M. Der Stellenwert der
Sonografie bei Speicheldrtisentumoren. HNO 1989;37:71-5.
Schroeder HG, Schwerk WB, Eichhorn T. Hochauflsende
Real-Time Sonographie bei Speicheldrtisenerkrankungen, II:
Speicheldrtisentumoren. HNO 1985;33:511-6.
Mann W, Wachter W. Ultraschalldiagnostik der Speicheldrtisen.
Laryng Rhinol Otol 1988;67:197-201.
Hausegger KW, Krasa H, Pelzmann W, Grasser RK, Frisch Ch,
Simon H. Sonographie der Speicheldrtisen. Ultraschall Med
1993;14:68-74.
Cvetinovic M, Jovic N, Mijatovic D. Evaluation of ultrasound in
the diagnosis of pathologic processes in the parotid gland. J Oral
Maxillofac Surg 1991;49:147-50.
Gritzmann N. Sonography of the salivary glands. AJR
1989;153:161-6.
Miwa K, Yuasa K, Yonetsu K, Kanda S, Higuchi K, Shinohara
M. Diagnostic accuracy of ultrasonography for salivary gland
tumors [in Japanese]. Journal of the Japanese Stomatological
Society 1995;44:240-4.
Dixon WJ, Brown MB, Engelman L, Jennrich RI. BMDP statistical software manual. Vols. 1 and 2. Berkeley, Calif: University
of California; 1993.
Thackray AC, Lucas RB. Pleomorphic adenoma. In: Tumors of
the major salivary gland. Washington DC: Armed Forces
Institute of Pathology; 1974. p. 16-39.

Reprint requests:
M. Shimizu, DDS, PhD
Department of Oral & Maxillofacial Radiology
Faculty of Dentistry, Kyushu University
Maidashi 3-1-1, Higashi-ku, Fukuoka, 812-8582
Japan