Full text of DOI:10.1089/105072501753211000

THYROID
Volume 11, Number 10, 2001
Mary Ann Liebert, Inc.
Techniques in Thyroidology
Computer-Assisted Cell Morphometry and Ploidy Analysis
in the Assessment of Thyroid Follicular Neoplasms
1
2
1
3
1
Andrea Frasoldati, Marcella Flora, Marialaura Pesenti, Angelo Caroggio, and Roberto Valcavi
Conventional cytology cannot discriminate between benign and malignant follicular neoplasms. Our study eval-
uated the diagnostic role of computer-assisted image analysis in the presurgical assessment of thyroid follicu-
lar neoplasms. Fifty-eight patients (14 males, 44 females, age range, 45–75 years) who underwent surgery for
cytologic diagnosis of thyroid follicular neoplasm were studied. All patients were first evaluated on clinical
grounds and assigned a high/low suspicion of malignancy on the basis of gender, age, and nodule size. Cell
image analysis was subsequently performed using a Cytometrica BYK Gulden microscope image processor on
Feulgen-stained thyroid cytologic smears. A different population of 50 benign and 50 malignant, histologically
evaluated nodules was studied in order to establish image analysis criteria suggestive of thyroid malignancy.
Ploidy histogram, proliferation index (PI), nuclear area coefficient of variation (NACV), and anisocariosis ratio
(
AR) were studied. Thyroid cancer was diagnosed in 16 of 58 follicular neoplasms. Only 7 of these lesions were
clinically suspicious (43.7%), whereas 14 of 16 (87.5%) malignant tumors were identified by image analysis.
Positive and negative predictive values of image analysis versus clinical evaluation were 46.6% versus 30.4%
and 92.8% versus 74.3%, respectively. The distribution of ploidy pattern was different in benign versus malig-
2
nant follicular neoplasms (x 8.25, p 5 0.016), malignant lesions showing an increased frequency of heteroclonal
aneuploid DNA content (37.5% vs. 7.1%). Increased PI (mean 6 standard deviation (SD) 5 11.3 6 5.7 vs. 7.1 6
6
.1; p , 0.01) and NACV (mean 6 SD 5 25.28 6 1.89 vs. 20.14 6 0.91; p , 0.01) levels were also observed in ma-
lignant follicular neoplasms. In conclusion, computer-assisted image analysis may profitably support clinical
evaluation in the assessment of thyroid follicular neoplasms.
Introduction
of malignancy (male gender, age $50, nodule size $4 cm,
single and solid nodules) should be surgically treated. How-
HYROID FOLLICULAR ADENOMA and hyperplastic adeno- ever, some authors suggest a “wait-and-see” strategy in low-
T
matoid nodule may share overlapping cytologic pattern risk patients (female gender, age ,50, cystic structure,
with well-differentiated thyroid follicular carcinoma and fol- multinodular goiter, nodule size ,4 cm) (8,11,12).
licular variant of thyroid papillary carcinoma (1–5). Fine-nee-
The role of ploidy analysis and nuclear morphometry in
dle aspiration biopsy (FNAB) is therefore characteristically the differential diagnosis between benign and malignant thy-
limited in making a specific diagnosis in follicular lesions roid follicular lesions has been investigated since the late
showing high cellularity and absence of colloid. These le- 1970s (13–17). The expanding application of computer-as-
sions are conventionally termed follicular neoplasms (6).
sisted image analysis has led to a reappraisal of cytomor-
The prevalence of thyroid cancer among follicular neo- phometrical studies (18–20). Furthermore, a new image anal-
plasms is 10%–20% (4,7,8). Many studies have recently in- ysis-based technique is now available for the evaluation of
vestigated clinical (8), cytologic (9), and ultrasonographic ploidy levels on cytologic samples (21). The goal of our study
(
10) features of follicular neoplasms, with the goal of im- was to evaluate the contribution of computer-assisted cell
proving the selection of patients for surgical referral. There image analysis on thyroid fine-needle aspirates in the presur-
is agreement that follicular neoplasms presenting high risk gical identification of thyroid malignant follicular lesions.
1
2
3
Unità Operativa di Endocrinologia, Unità Operativa di Anatomia Patologica e Unità Operativa di Otorinolaringoiatria, Centro Inter-
dipartimentale per le Malattie della Tiroide, “Azienda Ospedaliera di Reggio Emilia,” Arcispedale S. Maria Nuova-Reggio Emilia, Italy.
9
41

942
FRASOLDATI ET AL.
Materials and Methods
Study population
From June 1999 until June 2000, 1,218 ultrasound (US)-
guided fine-needle aspiration biopsies (FNABs) of thyroid
nodules were performed at our institution. A cytologic report
of follicular neoplasm, defined according to Lowaghen and
Sprenger (6), was obtained in 58 (4.7%) patients (14 males, 44
females; ages 34–75 years). These patients were assigned on
clinical basis to a high- or low-risk of malignancy. High-risk
patients presented at least one of the following features: male
gender, age $50 years and nodule size $4 cm (8,11,12). Thy-
roid smears of the 58 follicular neoplasms were evaluated by
computer assisted image analysis by a single operator (M.F.),
who was unaware of clinical data. All patients with follicu-
lar neoplasms were subsequently referred to neck surgery for
histologic examination. Image analysis criteria of thyroid ma-
lignancy were established on thyroid smears obtained from
FIG. 2. Image analisis of a thyroid fine-needle preparation
(
computer screen). The image of a thyroid cell nucleus (Obj
Type) is digitized by the videocamera and analyzed by cal-
culation of the nuclear area (Obj Area) and the integrated
a different population of 50 histologically evaluated benign optical density of the nucleus (Obj IOD).
nodules (13 males, 37 females; ages 26–71 years) and 50 his-
tologically studied differentiated thyroid malignancies (14
males, 36 females; ages 19–68 years), (control study).
the integrated optical density (IOD) parameter. IOD was mea-
sured on a 570-nm wavelength and calculated the amount of
2
Cell image analysis
absorbent material on each pixel (0.16 mm ) from a given nu-
clear area, computed on a high-resolution mode (4003). For
each lesion, the ratio between thyroid cells main population
IOD and lymphocytes IOD was calculated (DNA index).
Three different types of ploidy histogram were observed (Figs.
Thyroid smears were obtained from US-guided FNAB, us-
ing a 7.5-MHz sector probe equipped with a needle pointer
device. Specimens were obtained with a 22–27 gauge spinal
needle, inserted under US visual control. The needle, con-
nected with a 20 mL Cameco-mounted (Precision Dynamics
Co., Burbank, CA) syringe was repeatedly moved inside each
nodule, until the needle hub was filled with material. Com-
puter-assisted cell image analysis was performed on Feulgen-
stained thyroid cytologic smears (Fig. 1). Cell nuclei (150 per
slide) were analyzed using a Cytometrica BYK Gulden mi-
croscope (Konstanz, Germany) image processor (Fig. 2). Feul-
gen-stained nuclei of lymphocytes were used as controls.
3
A–3C): diploid, monoclonal (main population $70%; DNA
index 5 0.9–1.1); aneuploid, monoclonal (main cell popula-
tion .70%; DNA index ° 0.9–1.1); aneuploid, heteroclonal
(main cell population ,70%). The proliferation index (PI) was
calculated as the percentage of cells in the S and G2 phases.
Data analysis
Results of cell image analysis were elaborated by one-way
2
Analysis of nuclear morphometry included the following pa- analysis of variation (ANOVA) and x test. Positive predictive
rameters: nuclear area of thyroid cells (NA); nuclear area co- value (PPV) and negative predictive value (NPV) were defined
efficient of variation (NACV 5 standard deviation of nuclear according to Sox et al. (22). The calculation of posttest proba-
area/mean nuclear area 3 100); anisocariosis ratio (AR 5 thy- bility of thyroid malignancy was based on Bayes’ theorem (22).
roid cells mean nuclear area/lymphocytes mean nuclear area).
The nuclear DNA histograms resulted from the calculation of Results
Control study
The distribution of DNA ploidy pattern was different in
2
benign versus malignant thyroid nodules (x 8.40, p 5 0.015)
(Table 1). In malignant lesions, the heteroclonal aneuploid
DNA content was more frequent (10/50; 20%) as compared
to benign lesions (2/50; 4%). Furthermore, increased PI,
NACV, and AR values were detected in malignant nodules
(Table 2). Cutoff values, calculated as the mean plus 1 SD of
benign lesion values, were the following: PI $ 12.0%;
NACV $ 25.5%; AR $ 3.84%. Therefore, the finding of a het-
eroclonal aneuploid DNA content and/or PI, NACV, AR val-
ues above the cutoff values, suggested thyroid malignancy.
FIG. 1. Thyroid cytologic smears stained by Feulgen These criteria were used in image analysis evaluation of fol-
Method. Left: Benign follicular neoplasm (4003). Follicular licular neoplasms.
structures and regular nuclei with finely dispersed chromatin
may be observed. No nucleoli, grooves or cytoplasmic inclu-
Follicular neoplasms
sions are evident. Right: malignant follicular neoplasm
(
4003). Severe pleomorphism and chromatin clumping with-
At histologic examination 16 of 58 follicular neoplasms
out intranuclear inclusions and nuclear grooves are visible. were found to be malignant: papillary thyroid cancer, follic-

IMAGE ANALYSIS IN THYROID FOLLICULAR NEOPLASMS
943
A
B
C
FIG. 3. Ploidy patterns obtained from computer-assisted image analysis of thyroid smears. Diploid, monoclonal pattern
A): aneuploid, monoclonal pattern (B): aneuploid, heteroclonal pattern (C). Graphs on the left show cell nuclear area val-
(
ues (Area, y axis) plotted against the integrated optical density values (IOD; x axis). In diploid, monoclonal pattern nuclear
area and IOD values are clotted; in aneuploid, heteroclonal pattern they are widely scattered. Histograms on the right show
thyroid cell populations with variable DNA content.
ular variant, was the prevailing histologic type (n 5 8). follicular neoplasms, presented follicular groups arranged in
Among benign lesions, 28 follicular adenomas and 14 hy- monolayer sheets with occasional nuclear grooves and chro-
perplastic nodules were diagnosed. Clinical features of be- matin clearing. These features, consistent with the diagnosis
nign and malignant follicular neoplasms are shown in Table of follicular variant of papillary thyroid carcinoma (23) were
3
. After revision, 2 cytologic smears, previously reported as underestimated at a first examination.

944
FRASOLDATI ET AL.
TABLE 1. PLOIDY PATTERNS IN HISTOLOGICALLY
PROVEN BENIGN AND MALIGNANT
TABLE 3. THYROID FOLLICULAR NEOPLASMS: CLINICAL DATA
THYROID NODULES (CONTROL STUDY)
Follicular neoplasms
Control study
Benign
Malignant
Total
(
n 5 42)
(n 5 16)
(n 5 58)
Benign
n 5 50)
Malignant
(n 5 50)
(
Sex
31 F/11 M
42.1 6 12.6
13 F/3 M
47.7 6 14.4
44 F/14 M
43.8 6 13.5
Age
Monoclonal, diploid
Monoclonal, aneuploid
Heteroclonal, aneuploid
34 (68%)
14 (28%)
2 (4%)
22 (44%)
18 (36%)
10 (20%)
mean 6 SD
Size (mm)
22.9 6 10.2
22.4 6 10.4
22.7 6 10.2
mean 6 SD
x2 5 8.40; p 5 0.015
F, female; M, male; SD, standard deviation.
Clinical evaluation. Twenty-three (39.6%)patients with fol- ical practice, the appropriate management of follicular neo-
licular neoplasms were suspicious of malignancy according plasms is controversial. In light of the 10%–20% prevalence
to age ($50 years), gender (male), and nodule size ($40 mm). of thyroid cancer recorded in the largest series (4,7) surgical
Only 7 of these 23 patients indeed had malignant tumors, excision of all follicular neoplasms has been recommended
while 16 had benign lesions. This rated clinical evaluation (4). However, in order to avoid unnecessary surgery, a con-
sensitivity and specificity to 43.7% and 57.1%, respectively. servative approach has been proposed for those lesions
PPV and NPV were 30.4% and 74.3%, respectively.
showing low clinical risk of malignancy (8).
In our series, had surgery been restricted to patients
Image analysis. Thirty lesions (51.7%) were suspicious for judged at high risk of malignancy according to clinical cri-
malignancy according to image analysis criteria. Of these, 14 teria, diagnosis of thyroid cancer would have been missed
were malignant and 16 benign. This rated image analysis in 9 of 16 (56.2%) patients. One possible explanation for the
sensitivity as high as 87.5%, with a 57.1% specificity. PPV low sensitivity of clinical evaluation may be because of the
and NPV were 46.6% and 92.8%, respectively.
small size of nodules in this series. Thus, it seems that clin-
Image analysis features of malignant follicular neoplasms ical evaluation needs to be supported by further techniques
were comparable to those observed in the malignant lesions that may help identify less apparent malignant follicular
of the control study (Tables 4–5).
lesions.
The application of Bayes’s theorem estimated that thyroid
Our results indicate that computer-assisted image analy-
cancer probability in follicular lesions suspected to be ma- sis of follicular neoplasms by fine-needle aspirates allows to
lignant at image analysis was 0.332 (assuming a 15% preva- identify the majority (14/16) of malignant lesions prior to
lence of thyroid cancer among follicular neoplasms). In fol- neck surgery. Specificity of image analysis and clinical eval-
licular lesions lacking these criteria, cancer probability uation were comparable. Higher sensitivity of image analy-
decreased to 0.031.
sis was therefore not associated with an overwhelming in-
crease in unnecessary surgery.
Discussion
In our study, computer-assisted image analysis of thyroid
follicular neoplasm was based on two lines of data: ploidy
analysis and thyroid cell nuclei morphometry. Much evi-
dence indicates that ploidy analysis is useful for the prog-
nostic assessment of thyroid cancer once it is diagnosed (21).
The relevance of ploidy analysis in the identification of ma-
lignant lesions is still under debate. Aneuploidy is not ex-
clusive of thyroid cancer: aneuploid follicular adenomas
have been reported (15–17), which may represent lesions un-
Cytologic examination may not discriminate between ma-
lignant and benign follicular lesions. In these cases the re-
port is usually an indeterminate follicular neoplasm. In clin-
TABLE 2. PROLIFERATION INDEX AND NUCLEAR MORPHOMETRY
IN HISTOLOGICALLY PROVEN BENIGN AND MALIGNANT
THYROID NODULES (CONTROL STUDY)
Control study
TABLE 4. THYROID FOLLICULAR NEOPLASMS: PLOIDY PATTERNS
Benign
Malignant
(
n 5 50)
(n 5 50)
Follicular neoplasms
PI (%; mean 6 SD)
5.96 6 6.10
68.60 6 25.4
20.87 6 4.78
2.81 6 1.03
10.85 6 9.10a
2
Benign
(n 5 42)
Malignant
(n 5 16)
NA (mm ; mean 6 SD)
76.85 6 19.26
25.23 6 6.61a
NACV (%, mean 6 SD)
AR (AU, mean 6 SD)
3.36 6 0.87b
Monoclonal, diploid
17 (40.5%)
22 (52.4%)
3 (7.1%)
5 (31.2%)
5 (31.2%)
6 (37.5%)
a
b
Monoclonal, aneuploid
p , 0.01; p , 0.05.
PI, proliferation index; NA, nuclear area; NACV, nuclear area Heteroclonal, aneuploid
coefficient of variation; AR, anisocariosis ratio; SD, standard devia-
x2 5 8.25; p 5 0.016.
tion; AU, arbitrary units.

IMAGE ANALYSIS IN THYROID FOLLICULAR NEOPLASMS
945
TABLE 5. PROLIFERATION INDEX AND NUCLEAR MORPHOMETRY
IN BENIGN AND MALIGNANT THYROID FOLLICULAR NEOPLASMS
8. Tuttle RM, Lemar H, Burch HB 1998 Clinical Features asso-
ciated with an increased risk of thyroid malignancy in pa-
tients with follicular neoplasia by fine needle aspiration.
Thyroid 8:377–383.
Follicular neoplasms
9
. Gardner HAR, Ducatman BS, Wang HH 1993 Predictive
Benign
Malignant
value of fine-needle aspiration of the thyroid in the classifi-
cation of follicular lesions. Cancer 71:2598–27603.
0. Lin JD, Hsueh C, Chao TC, Weng HF, Huan BY 1997 Thy-
roid follicular neoplasms diagnosed by high resolution ul-
trasonography with fine needle aspiration cytology. Acta
Cytol 41:687–691.
1. Tyler DS, Winchester DJ, Caraway NP, Hickey RC, Evans
DB 1994 Indeterminate fine needle aspiration biopsy of the
thyroid: identification of subgroups at high risk for invasive
carcinoma. Surgery 116:1054–1060.
(
n 5 42)
(n 5 16)
1
1
PI (%; mean 6 SD)
7.18 6 6.14
76.17 6 24.52
20.14 6 5.92
3.17 6 1.01
12.13 6 5.51a
83.84 6 19.21
25.64 6 7.17a
3.46 6 0.77
2
NA (mm , mean 6 SD)
NACV (%, mean 6 SD)
AR (AU, mean 6 SD)
ap , 0.01.
PI, proliferation index; NA, nuclear area; NACV, nuclear area
coefficient of variation; AR, anisocariosis ratio; SD, standard devia-
tion; AU, arbitrary units.
12. Schlinkert RT, van Heerden JA, Goellner JR, Gharib H, Smith
SL, Rosales RF, Weaver AL 1997 Factors that predict malig-
nant thyroid lesions with fine-needle aspirations is “suspi-
cious for follicular neoplasm.” Mayo Clin Proc 72:913–916.
dergoing malignant transformation (13,14,21). In our series, 13. Lukacs GL, Balasz GY, Zs-Nagy I 1979 Cytofluorimetric
the finding of an aneuploid monoclonal pattern applied to
many benign follicular lesions; nevertheless, the heteroclonal
measurements of the DNA contents of tumor cells in human
thyroid gland. J Cancer Res Clin Oncol 95:265–271.
aneuploid DNA histogram was prevalent in thyroid malig- 14. Greenebaum E, Koss LGL, Elequin F, Silver CE 1985 The di-
nancies. In agreement with previous observations (21), ma-
lignant follicular neoplasms showed increased PI levels.
Many studies report that an increased nuclear diameter is a
feature of thyroid malignancy (28,24–26,27). However, this
finding has not been confirmed in our and other studies
agnostic value of flow cytometric DNA measurement in fol-
licular tumor of the thyroid gland. Cancer 56:2011–2018.
5. Joensuu H, Klemi P, Eerola E 1986 DNA aneuploidy in fol-
licular adenomas of the thyroid gland. Am J Pathol 1243:
1
1
3
73–376.
6. Hruban RH, Huvos AG, Traganos F, Reuter V, Lieberman
PH, Melamed MR 1990 Follicular neoplasms of the thyroid
in men older than 50 years of age: a DNA flow cytometric
study. Am J Clin Pathol 94:527–532.
7. Grant CS, Hay ID, Ryan JJ, Bergstrahl EJ, Rainwater LM,
Goellner JR 1990 Diagnostic and prognostic utility of flow
cytometric DNA measurements in follicular thyroid tumors.
World J Surg 14:283–289.
(
20,28). Instead, we observed a wider range of variation of
nuclear size (NACV) in malignant follicular lesions. This re-
sult is in good agreement with a recent report showing a
good (85%) NACV sensitivity in the detection of malignant
thyroid follicular lesions (20). Altogether, risk of malignancy
turned out to be about 10-fold higher in follicular lesions
with suspicious features at image analysis.
In conclusion, computer-assisted ploidy analysis and nu-
clear morphometry applied to fine-needle aspirates should
be considered a valuable technique that may help in the clin-
ical management of thyroid follicular neoplasms.
1
1
1
8. Kashyap V, Kaushik N, Bhambhani S, Das DK, Luthra UK
1
992 Supportive role of image analysis and DNA ploidy pat-
tern. Diagnostic Cytopathology 8:228–230.
9. Zacks JF, De Las Morenas A, Beazley RM, O’Brien MJ 1998
Fine needle aspiration cytology diagnosis of colloid nodule
versus follicular variant of papillary carcinoma of thyroid.
Diagn Cytopathol 18:87–90.
References
1
. Lowhagen T, Willems JS, Lundell G, Sunblad R, Granberg 20. Nagashima T, Suzuki M, Oshida M, Hashimoto H, Yagata
PO 1981 Aspiration biopsy cytology in diagnosis of thyroid
cancer. World J Surg 5:61–73.
. Silverman JF, West RL, Larkin EW, Park HK, Finley JL,
H, Shishikura T, Koda K, Nakajima N 1998 Morphometry
in the cytologic evaluation of thyroid follicular lesion. Can-
cer Cytopathol 84:115–118.
2
Swanson MS, Fore WW 1986 The role of fine needle aspira- 21. Salmon I, Gasperin P, Remmelink M, Rahier I, Rocmans P,
tion biopsy in the rapid diagnosis and management of thy-
roid neoplasm. Cancer 57:1164–1170.
. Hall TL, Layfield LJ, Philippe A, Rosenthal DL 1989 Sources
of diagnostic error in fine needle aspiration of the thyroid.
Pateels JL, Heimann R, Kiss R 1993 Ploidy level and prolif-
erative activity measurements in a series of 407 thyroid tu-
mors or other pathologic conditions. Hum Pathol 24:912–
920.
3
Cancer 63:718–725.
22. Sox HC, Blatt MA, Higgins MC, Marton KI (eds.) 1988 Med-
4. Gharib H 1994 Fine needle aspiration biopsy of thyroid nod-
ical Decision Making. Butterworth-Heinemann, Boston.
ules: advantages, limitations and effects. Mayo Clin Proc 69: 23. Logani S, Gupta PK, LiVolsi VA, Mandel S, Baloch ZW 2000
4
4–49.
. Oertel YC 1996 Fine needle aspiration and the diagnosis of
thyroid cancer. Endocrinol Metab Clin North Am 25:69–91.
Thyroid nodules with FNA cytology suspicious for follicu-
lar variant of papillary thyroid carcinoma: Follow-up and
management. Diagn Cytopathol 23:380–385.
5
6
. Lowhagen T, Sprenger E 1974 Cytologic presentation of thy- 24. Boon ME, Lowhagen T, Willems JS 1980 Planimetric studies
roid tumors in aspiration biopsy smear. A review of 60 cases.
on fine needle aspirates from follicular adenoma and follic-
Acta Cytol 18:192–198.
ular carcinoma of the thyroid. Acta Cytol 27:145–148.
7
. Hamburger JL, Kaplan MM 1996 Evaluation of thyroid nod- 25. Kini SR, Miller JM, Hamberger JI 1985 Cytopathology of fol-
ules by needle biopsy. In: Braverman LE, Utiger RD (eds) The
licular lesions of the thyroid. Diagn Cytopathol 1:123–132.
Thyroid. Lippincott-Raven, Philadelphia-New York, p 447. 26. Suen KC 1988 How does one separate cellular follicular le-

946
FRASOLDATI ET AL.
sions of the thyroid by fine needle aspiration biopsy? Diagn
Cytopathol 4:78–81.
7. Deshpande V, Kapila K, Sai KS, Verma K 1997 Follicular neo-
plasma of the thyroid. Decision tree approach using morpho-
logic and morphometric parameters. Acta Cytol 41:369–376.
8. Luck JB, Mumaw VR, Frable WJ 1982 Fine needle aspiration
biopsy of the thyroid: differential diagnosis by the video-
plan image system. Acta Cytol 26:793–796.
Address reprint requests to:
Roberto Valcavi, M.D.
Servizio di Endocrinologia
Arcispedale S. Maria Nuova
Viale Umberto I, 50 42100 Reggio Emilia
Italy
2
2
E-mail: valcavi.roberto@asmn.re.it