[CANCER RESEARCH 38, 4688-4697, December 1978]

0008-5472/78/0038-0000$02.00

Computerized Nuclear Morphometry as an Objective Method for

Characterizing Human Cancer Cell Populations1
BjOrn Stenkvist,2 Sighild Westman-Naeser, Jan Holmquist, Bo Nordin, Ewert Bengtsson, Jan Vegelius,

Olle Eriksson,and Cecil H. Fox3

Department of Clinical Cytology, University Hospital, S-750 14 Uppsala, Sweden (B. S., S. W-N.J; Departments of Computer Science (J. H., B. N., 0. E.j,
Physics (E. B.), and Statistics (J. V.1, Uppsala University, Uppsala, Sweden; Department of Pathology, Karolinska Institute, 104 01 Stockholm, Sweden (C.
H. F.); and Laboratory of Biochemistry, Division of Cancer Biology and Diagnosis, National Cancer Institute, Bethesda, Maryland 2@X@14
(C. H. F.)

ABSTRACT

galbocyanin. The results of this objective grading system

are compared with the results of subjective grading and
A new method for measuring differences in nuclear classification. In addition, we have succeeded in obtaining
detail in chrome alum galbocyanin-stained
nuclei of cells a quantitative expression of nuclear pleomorphism that
from human breast cancers was comparedwith conven allows the determination of the reliability of our sampling
tional subjectivegradingand classificationsystems.The methods and also allows the examination of the biological
new method,termed computerizednuclearmorphometry homogeneity of a breast cancer.

(CNM), gives a multivariatenumerical score that corre

bateswellwith nuclearatypiaandgivesa higherreproduc
ibility of classification

than do subjective

observations

with conventional histological preparations. When 100

individualnuclei from each of 137 breast cancers were
examinedbyCNM, therewas a broadCNM scorevariation
between patients but a good reproducibilityfor each
tumor. When different parts of the same tumor were
sampled, there was good reproducibilitybetween sam
pIes, indicating that some breast cancers at least are
geometrically
monocbonal.
When these cancers were
compared by the grading systems of WHO and Black,
correlationsof 0.43 and 0.48, respectively,were found.
There was a poor correlationbetween CNM and classifi
cations of tumor type, but in general there were high

MATERIALSAND METHODS
Patients
Cellular material was obtained from 162 of
tive breast cancers removed from patients
counties during a period of 5 months. The
unusually well characterized as a part of a
correlating

epidemiology

and morphology

181 consecu
in 4 Swedish
patients were
larger project

of breast cancer.

A detailed history, physical examination, and questionnaire

for epidemiobogical information were administered to each
patient, and blood samples were drawn both before and
after surgery for use in hormone and other assays. To date,
the patients' course has been followed for 2 years, with
collection of clinical course observations continuing.

values for CNM in medullarytumors and low values in

mucous tumors. Correlations between CNM and tumor
progressionand prognosisawait future studyof patients Specimen Preparation
participating

in the study.

INTRODUCTION
Human breast cancers are diagnosed on the basis of the
appearance of tissue or cells removed from the patient and
prepared for microscopic evaluation. In addition to the
diagnosis of cancerous tissue, most pathologists also at
tempt to classify the histological origin of the tumor and to
estimate the grade
of the tumor. While the significance
of subjective grade estimates for prognosis has been de
bated (8, 22, 24), there have been a number of such systems
proposed, i.e. , those of Bloom and Black (4, 5). A number
of classification systems have also been suggested, i.e.,
those of WHO, Ackerman, and Armed Forces Institute of
Pathology(1,21,25).
In an effort to improve reproducibility of grading of breast
cancers, we have used digitized image analysis of the nuclei
of breast cancer cells that were stained with chrome alum
I This

work

was

supported

by NIH-National

Cancer

Institute

Contract

NOl

CB-53968 through the Breast Cancer Task Force.

a To whom requests for reprints should be addressed.
3 On

leave

from

Laboratory

of

Biochemistry,

Division

of

Cancer

and Diagnosis, National Cancer Institute, Bethesda, Md.

Received December 27, 1977; accepted September 15, 1978.

4688

Each tissue specimen was taken directly to the laboratory

following its removal. The fresh tumor was biopsied first by
fine-needle aspiration with a 22-gauge needle and a 20-mI
syringe in a syringe holder. This method produces single
dispersed cells that are well suited to cytological study.
Following aspiration biopsy the tumor was divided, the size
of the tumor was measured, and portions were taken for
determination of estrogen receptors and histopathology.
Imprint specimens were obtained from the cut surface of
the tumor by gently pressing a microscope slide against the
surface. All cytological specimens were fixed immediately
in fresh Carnoy's solution and stained with chrome alum
gallocyanin (9, 10). Of the 179 tissue specimens, adequate
and complete cytological material was obtained from 162.
Of these 137 specimens have been analyzed by CNM.4
For determination of the biological homogeneity of the
tumors, fine-needle biopsies were taken from different parts
of the same tumor in 14 cases. For demonstration of
congruence between imprint and needle biopsy specimens,
fine-needle biopsies and imprint specimens were obtained
from different parts of the same tumor and CNM values
were compared.

Biology
4 The

abbreviation

used

is:

CNM,

computerized

nuclear

morphometry.

CANCER RESEARCHVOL. 38

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

Computerized

Selectionof TumorCell Nuclei

Cytological

preparations

were carefully reviewed , and

100 nuclei from each specimen

were chosen for recording.

Only typical cancerous epithelial cells with undamaged

nuclei were chosen. Each subsequent cell was selected for
recording by a standard process to avoid the operator's
making subjective decisions of cell selection. At the same
time that each cell image was digitized, a color photograph
(Ektachrome Film; Eastman Kodak, Stockholm, Sweden)
was made in white light so that the selected field of vision
could be checked against the computer image.

Image Recordingand Preprocessing

A diode array scanner coupled with a Leitz Orthoplan
microscope (E. Leitz, Wetzlar, Germany) was used to record
the cell images. The details of the scanning system have
been described elsewhere (2, 13) and will only be summa
rized here. The diode array scanning microscope was
developed by Aslund and coworkers (17, 18) at the Royal
Institute of Technology, Stockholm, Sweden and is not yet
commercially available. It consists of a linear array of 256
diodes that measure the light in the image. The image is
divided into narrow bands by a swinging prism device, and
each of these image bands is measured by the diodes. The
signal from each diode is digitized and transmitted to the
computer interface and thence to a central processing unit.
The image recording work was done with a PDP-8 minicom
puter. The subsequent image processing was done with a

Nuclear Morphometry

PDP 11-55 (Digital Equipment Corporation, Maynard,

Mass.). The light source for the microscope was fitted with
a 570-nm interference filter (30-nm bandwidth) for scan
ning. Rectangular fields were scanned with a final spatial
resolution of 0.2 @m
with a maximum image size of 256 x
256 image points

and 6 bits of gray-level

resolution.

The

hardware is diagrammed in Chart 1.

The digitization process for each image was followed
with a Tektronix 4010 (Tektronix Inc., Beavertown, Oreg.)
display. The nuclear images were isolated from their sur
roundings by a rectangle defined by a cross-hair cursor and
stored for further processing together with a gray level
corresponding to the nuclear border. These gray levels

were determined from the gray-bevelhistogram of the visual

display by the use of an interactive graphic computer
technique previously developed (3, 20). All values above the
nuclear border threshold were initially considered as be
longing to the nucleus.
The images were then cleaned by local image processing
operations (2, 13). After this preprocessing image plots
showing the border of each nucleus were visually reviewed,
and images not clearly defined were excluded from the
population. This occurred in less than 0.2% of the cases,
and in no instance were more than 10 nuclei from any one
tumor excluded.

Mathematical Methods to Describe Nuclei and Nuclei

Populations
The mathematical procedure for describing the numerical
RIN!COSPUTCS

pop @s/r

Chart 1. Hardware configuration. CPU, central proc

essing unit; DEC, Digital Equipment Corporation.

DECEMBER 1978
Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

4689

B. Stenkvist et a!.
features of a tumor cell population obtained from a breast
cancer was structured into (a) parameters describing the
cell nuclei and (b) parameters describing the entire cell
population based on the parameters extracted from the
single nuclei.
Numerical Features of Nuclei
These parameters were subdivided into 3 categories,
namely, into parameters describing density, shape, and
texture.
Density. The density parameters were derived from the
nuclear gray-level histograms, in which h is defined as the
number of image points in the region attaining gray-level i
and i= 0 ...63 so that:
AREA=@h

EXTINC=

Descriptors based on the Fourier series must be inde

pendent of size, rotation, and orientation of the contour to
be useful as cell shape descriptors. Position and size
independence are obtained by the normalizations. Since
rotation through angle 0 affects the coefficients by a multi
plicative factor e,the absolute values a,,f are independent
of rotation of the contour. Equal contours with different
orientation (mirror images) give the Fourier coefficients a,,
and a,,. Shape descriptors containing a,, and a@,,symmet
rically will give orientation independence. From the Fourier
coefficients we calculated:
MAJTOMIN

(I)

ffa_,fla,lI

BDYVAR (m, p) = (a,2 + &,2)

(A)

[Iog(64 log(i)}.h,

= a_,( + la,l

(J)

MAJTOMIN is an elongation measure and BDYVAR is a

(B)

measure

of the

boundary

variations

in a part

of the

fre

quency spectrum.
@

AVLEVEL

ih/AREA

(C)

Texture.

These

parameters

were

basically

extracted

from

differences in light transmission between neighboring im

@

VARLEVEL

(i

AVLEVEL2h/AREA

(D)

age

points

within

the

nuclear

contour.

Two

classes

of

parameters were calculated:

ENTROPY = @h, . 109(h)/AREA

(E)

The AREA parameter is the area over the region. EXTINC is

the amount of light absorbed by the region. AVLEVEL is the
average gray-level over the region and VARLEVEL is the
corresponding variance. ENTROPY is a measure of the
information content of the gray-level histogram.

Shape.The shapeparameterswerebasedon the workof

Granlund (11) from which we developed computer subrou
tines for extracting shape descriptors from Fourier expan
sion of closed contours.
A point moving around a closed contour at constant
speed generates a complex function of time u(t. This
function is periodic and can be expanded in a Fourier
series.
@

u(t) = ,,

a, =

2ir

a,,e''

f211u(t) .e'dt

(F)

from

differences

range

to

8 gray

levels.

average coarseness

These

of chromatin

u(t) = ,, @m
e@'

in gray

level

between

parameters

structure

measure

the

in the nucleus

from an 8 x 8 transition probability matrix. A set of 4

parameters were calculated for each nucleus. The matrix
had the following

appearance:

p(lIlp(112
p(211p(212

p(118
p(218i

p(811)P(812)

P(818)

in whichp(ilj@is definedas the probabilityof gray-levelj

occurring after gray-bevel i in the nucleus when moving
along the scan lines within the nucleus. The 4 parameters
were:
ASM = ,@,

derived

pairs of image points along the scan lines in the nucleus

taken in steps of 1, 2, etc., to 10 steps of points. The larger
the local variation in light transmission the more frequen
cies of differences for a given step length will be shifted
towards larger values for those differences. The first mo
ment of the frequencies after normalization according to
nuclear size are used as parameters and are called DIFF1,
DIFF2
DIFF1O depending on step length; (b) a class
derived from a transition probability matrix over the nucleus
previously used by Haralick et a!. (12) and Pressman (19).
To facilitate processing we reduced the nuclear gray-level

(G)

assuming that the period of u is 2@r.The coefficients a,, are

complex and were calculated by the fast Fourier transform
algorithm after the closed contour had been interpolated
into 512 sampled time points and moved to the origin of the
complex plane followed by a normalization according to
size.
A truncated Fourier series

(a) a class

p(i@f)2

(K)

(H)

was used to approximate the contour. The first 2 coeff Icients determine an ellipse that approximated the original
contour (in least square sense). The major axis MAJAXIS of
this ellipse is Ia,I + Ia@,@,and the minor axis MINAXIS is lla@l

CONTRAST= fl@,

DIFMOMENT=

(i

p(ilJ)]

(L)

j)2.p(ilj)

(M)

1a,ll.

4690

CANCER

RESEARCH

VOL. 38

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

Computerized Nuclear Morphometry

@@[p(iIj

COEFVAR=

_ m@

(N

64 m

Histopathobogical
Methods

where

p(iI/)
m

64

Description of the Tumor Cell Population

To describe the tumor based on the nuclear measure

ments, we approximated the sample of multidimensional
random vectors that resulted from the measurements of
individual nuclei by analytical probability distributions. The
parameters of the distributions were used as cell population
descriptors. In this way the tumor cell population was
represented by a mean vector and a covariance matrix. The
cell population descriptors can be presented in the follow
ing way. Let
be theJth descriptor,J = 1 . . . m, measured
from the ith cell, i = 1 . . . n, in the sample from a specific
tumor. The means of the descriptors

will then be

D@
= 11,@,d,@
D@can be used as a tumor descriptor.

(0)

The covariance

matrix
@

Cj.k

@,
(d,@D@(d,@0k)

(P)

can be used as another descriptor of the tumor.

OptimizationProcedure
Many methods can and will be used to evaluate the
importance of cancer cell population descriptors. For in
stance, the descriptors can be evaluated on the basis of
patient survival, metastatic spread at the time of diagnosis,
epidemiobogical characteristics, etc. In this basic study the
descriptors were used to model the degree of atypia based
on the cytological judgment.
In this study the nuclear atypia of a tumor cell population
level was described by a linear mathematical model:
@

CNM = nuclear

population

atypia =

wD.

nuclear

population

atypia

measure

can

be altered

(under the limitation of linearity) to correspond to various

patient parameters.
For adjustment of weighting, photographs of recorded
cell populations were ranked by increasing atypia accord
ing to cytological criteria. The weights w were determined
from the atypia ranking by using multiple linear regression
analysis performed stepwise with one complete regression
per step. At each step the one parameter that caused the
largest reduction in the error sum of squares was added. A
combination of m(AREA), m(AVLEVEL), m(ENTROPY),
m(BDYVAR), m(COEFVAR), and m(DIF1) was first included
DECEMBER1978

When the tumor was divided, a slice about 3 mm thick

was taken through the greatest diameter of the tumor, thus
ensuring that the central portion of the tumor was included
as well as the borders. This slice was fixed immediately in
Carnoy's solution. The remainder of the tumor, surrounding
tissues, and axillary tissues were fixed in 4% neutral form
aldehyde for further study. The Carnoy's-fixed portion was
sectioned and stained with the following: periodic acid
Schiff stain with and without prior diastase digestion (mucin
and glycogen); Voerhoff's orcein-iron hematoxylin (elastic
fibers); Azan-Heidenhain's stain (connective tissue); methyl
green-pyronin (mast and plasma cells), and Alcian green
stain (keratin and mucus). In addition, formalin-fixed as
well as Carnoy's-fixed tissue was sectioned and stained
with hematoxylin-eosin or van Gieson's stain and used for
morphological diagnosis. Similar staining studies were
made of lymphatic metastases when warranted. Histobogi
cabbyconfirmed lymphatic metastases were found in 53 of
the patients.

Gradingand Classification
Each tumor was graded by the WHO (5, 21) and Black (4)
systems. In addition, tumors were classified by the Acker
man (1), WHO (21), and the Armed Forces Institute of
Pathology (25) classifications. These gradings and classifi
cations were performed with the use of supplementary
stains and histochemical methods and represent a more
detailed examination than that usually given breast cancer
cases. To arrive at a grade or classification, 2 pathologists
examined the slides from each case and then reexamined
the same slides after a passage of 6 months. Thus each
case was graded or classified 4 different times. The variable
reproducibility of the grading and classification systems
required that a system be developed for arriving at an
approximated final score. If 3 or 4 of the 4 readings of the
tumor were in agreement, the tumor received the designa
tion of these judgments. In cases in which only 2 readings
of 4 were in agreement, the final designation for that
specimen was made by the pathologist who showed the

(Q) highestrateofreproducibilityofsubjectivegradingfor that

D is a population descriptor and w1is the weight assigned

to that particular parameter. By choosing suitable weights
Wi, the

with the corresponding sample correlation coefficient be

tween computer and visual assessment of nuclear popula
tion atypia.

particular system.
The tumors were also categorized to indicate the degree
of differentiation of the tumor tissue by dividing the tumors
into low, middle, and highly differentiated.
Statistical Methods
Statistical comparisons were made between CNM values
and classification and gradation variables. CNM is an inter
val scale variable. In determination of the linearity of the
relationship between CNM and some other interval variable,
the production moment correlation coefficient and its cor
responding 2-tailed p value were used as a test of linear
independence (15). When ordinal variables or interval van
ables that are not normally distributed were compared with
CNM values, Spearman's rank correlation coefficient was
4691

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

B. Stenkvist

et a!.

used and , if the independence hypothesis was to be tested,

its corresponding 2-tailed p value was also analyzed (14).
For comparisons with nominal variables such as the
classification systems, correlation coefficients for nominal
scales such as Tschuprow's coefficient (26, 27) on Cramer's
V (7) were used. In some instances,

correlation

coefficients

may have questionable reliability, and in those instances

the most frequent categories in the nominal variable were
taken out and the t test for independent samples (6) or the
Mann-Whitney U test (23) were used to determine whether
the CNM values for tumors in one category are significantly
different from those in another category.

CNM values with a correlation coefficient of 0.79 (Chart 3).

Chart 4 demonstrates a scatter plot of the result of
multiple-site sampling from the same tumor by repeated
fine-needle aspiration biopsy. The correlation coefficient
for these values is 0.86, obtained by plotting 2 different
populations against each other. Because of the good con
relation obtained between the CNM values of these appar
ently geometrically
monocbonab
cells, we used this set of
data to test some of the individual descriptors that make up
659

550

r@i
P
G

RESULTS

R
A

Fig. 1 shows the quality of the cytological preparations

stainedwith chromealum gallocyanin.A comparisonbe

tween a high-resolution light photomicrograph of a cell
nucleus and a computer image of the same nucleus is
shown in Fig. 2. The image stored in the computer is at a
higher geometrical and photometric resolution than that
shown by the plotter, and for that reason the image is
somewhat coarser than that recorded on film.
A binary map was made of the images of the cells
scanned from each tumor. One such map is shown in Fig.
3. A frequency distribution of CNM factors for these cells is
shown in Chart 2.
The role played by the cell sampling method was studied
by comparing a population of cells collected from the same
tumor both by taking imprints of the cut surface of the
tumor and by using fine-needle aspiration biopsy. When
plotted against each other, the 2 populations of cells gave

@9150
181

Chart3. Scatterplot illustratingthe correlation betweenCNMvaluesfor

cell populationstakenfrom the sametumor by fine-needleaspirationbiopsy
(DATGRADE)and by imprints (IMPGRADE) from the cut surface ofthe tumor.
The coefficient of variation for the reproducibility of CNM is 0.79.

/L
5.555(454-0.256(451

AREA

5.455(45k5.155(451

ENTROPY

5.355(485-@5.188(481Ai4@EV
8.555(485EXTDIC

5.555(452--

-8.188(451DIFF4
-@nA@fl,Lrc
Chart2. Histogramsof the distribution of different

5.785(4515.588(455

8.755(451-DIFFI

parameters of the nuclei in Fig. 3 showing the density,

shape, and texture parameters.
--8.185(451

5.155(451

5.785(481I@8.555(455
DIFF8

kcONTRAS
5.555(455

4692

DIAGlIOII

8.155(482

BDYV55II

5.355(453

5.855(455

@CEF@i

S. 1@4S2

CANCER RESEARCH VOL. 38

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

Computerized

Nuclear Morphometry

Table 1
Correlationswith CNM for 4 variables(including 2-tailed valuesof

731.150

the independence hypothesis) used as positive and negative

statistical controls
Prod
uctSpear
man'smentrankcome
mo

P
G
R
A

D
E

No.lationCorre
ofcoeffi
spend

correla
tionCorre
spend
coeffi

pairscientingpcientingpSubjective

1370.4570.0000.4400.000ment
judg
of mi
tosesDegree
of differ
1370.0270.7580.0410.632entiationGrade

180.850

according1370.4760.0000.4630.000to
BlackGrade
according1370.4330.0000.4220.000to
WHO

@52.

Chart 4. 5catter plot showing the correlation between CNM values for
cells taken from different parts of the same tumor. One population value on
the horizontal axis corresponds to the value on the vertical axis that was
obtained by measuring another population from a different part of the tumor.
Correlation coefficient, 0.86.

9227.100
R
M

DISCUSSION

F
F
I

1230.900

1578 .

Chart 5. Average DIFF4 values measured twice from each of 14 tumors,

where the first determination is plotted on the horizontal axis (MDIFF4) and
the second is plotted on the vertical axis (RMDIFF4) as a scatter diagram.

the CNM value. Chart 5 shows the comparison of one such

highly abstract measure based on the frequencies of differ
ence in gray levels, average DIFF4 values, with a correlation
coefficient of 0.83. Other such comparisons gave correla
tion coefficients that varied between 0.75 and 0.86.
Two additional measures were added for comparison
with CNM values, mitoses and subjective assessments of
tumor differentiation. One of these, mitoses, was intended
as a positive control of CNM and was expected to be
correlated with a high CNM value, while the other, differ
entiation degree, was expected to be independent of CNM.
The results of these comparisons are seen in Table 1.
Another comparison between a subjective grading system
and CNM was made with the Black grading system (Chart 6
A). The product

the Black grading and CNM was 0.47 and between the WHO
gradingand CNM was 0.42(Table1).
A comparison was also made between final pathological
diagnoses based on the different classification systems and
CNM (Chart 6, B to 0).

moment

correlation

coefficient

between

The texture of the nuclear image is an important compo

nent in the histopathological-cytopathological
determina
tion of cancer. For this purpose cells are fixed and stained
in such a way that a reproducible pattern is obtained and
this pattern is then judged subjectively by the eye of the
pathologist. The chrome alum gabbocyanin used here is
generally believed to form (specific) stoichiometnic com
plexes with both RNA and DNA. Thus CNM is most probably
a measure of nucleic acid content (i.e., cell pboidy, tran
scniption rate) and distribution (i.e. , euchromatin, hetero
chromatin, RNA distribution) within the nucleus. Based on
its specificity, stoichiometry and intensity, chrome alum
galbocyanin appears to be a better choice for use with CNM
than does a non-stoichiometnic, classical staining method
such as hematoxylin-eosin.
In a recent paper by Nicolini et a!. (16), it was reported
that the texture of Feulgen-stained Wl-38 nuclei changed
when cell growth was stimulated by addition of serum to
the medium, although the DNA content of the cells re
mained the same. These textural changes were apparently
rebated to the cell cycle: cells in Go or G2 arrest had a
different nuclear DNA morphology than did cells in other
portions of the cell cycle. They also found that chromatin
dispersion is followed by a modulation of the nuclear shape
to give an irregular, less circular shape to the nuclear
border. This report, coupled with the positive correlation
between CNM value and number of observable mitoses in a
tumor

(Table 1), is reason to believe

that another

factor

influencing CNM values is the growth activity of the cell

population. The possible correlation between growth activ
ity of a tumor cell population and CNM values should be
further explored because of its implications for choice and
planning

of chemotherapy.

That is, many chemotherapeutic

DECEMBER1978
Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

4693

B. Stenkvist

et a!.
Gradin
( B1ack
N: 31

6A

CLASSIFICATIONS

6C

AFTER ARLED 0ORCES

INSTITUTE OF PATHOLOGY

LOBULAR CARc:N0VA

N: 2

Ii!:V

IINVASIVE)
CGLLOID CARCIGOMA
MEDULLARY CARCIr,O.G
CARCINOMA WITH F:BRos:s

I1:B:3

COMEDOCARC INOVA

II:s:2

PAPILLARY CARCINOMA

N: 34

175

5
CLASSIFICATION

CNN value

CNM
VALUE

84@

844

6B

ACCORDING TO

ACKERMAN
HIGHLY METASTASIZING

I@

INTRADUCTAL CARCINOMA

111:2

6D

CLASSIFICATION ACCORDING TG INC

WITH STROMAL INVASION

ADENOCARCINOMA

WELL-DIFF. ADENOCARCINOMA

MEDALLARY CARCINOMA

MUCINOUS CARCINOMA

LOBULAR CARCINOMA

C:111:E

MUCOUS CARCINOMA

C:I11:o

111:1

1:3
PAPIlLARY CARCINOMA

C:111:B

MEDALLARY CARCINOMA

C:III:@

INVASIVE CARCINOCA

C:11

11:2

11:1

175
175

CNM
VALUE

C4M VALUE

844

844

Chart 6. Distribution of CNM values. A , grading system advocated by Black; b , Ackerman classification system ; c , tumors classified by the AFIP criteria; d,
tumors classified according to the WHO system. 01FF, differentiated.

agents are effective only in S phase and, if a major portion

of the cells are in a resting phase, these antitumor agents
will not be effective. Prediction by CNM of this effect (or
back of effect) would be a valuable clinical tool in cancer
treatment.
Low reproducibility of subjective grading and classifica
tion systems as found in this study has been reported
previously. Cutler et a!. (8) studied the reproducibility of
Black's grading system and achieved results similar to ours,
although a thorough statistical study of the results was not
conducted. Schiodt (22), in an analysis of the grading
system of Bloom and Richardson (5), reported a lower
reproducibility than did Bloom and Richardson.
A common assumption in a situation in which poor
results are obtained in subjective judgments is that those
observers who do not succeed are in some way less com
petent than the individual who devised the system. In this
case, the pathologists making the subjective judgments
represent at least an average standard of competence in
breast cancer diagnosis. They used a number of additional
histopathobogicab methods and certainly devoted more time
than is usual to making diagnostic decisions but failed to
achieve consistent reproducibility (24). We believe that
CNM, with its great reproducibility, can therefore serve as a
remarkable aid in helping the pathologist to a decision and
thus form a new part of the armamentanium of cancer
diagnosis.
4694

CNM is capable

of providing

both a continuous

interval

scale variable and a group of independent descriptors.

These allow statistical comparison between other charac
tenistics of the patients' disease such as hormonal status,
estrogen receptors, thyroid function, and the complete
battery of chemical findings with an objective measure of
the tumor. Such a study has now been completed; the
results will be reported elsewhere.
As with any system with possible patient diagnosis appli
cation, the question of throughput and cost effectiveness
should be addressed. To scan one cell for CNM requires
about 4 mm depending on the quality of the slide and the
adeptness of the operator. Thus the instrument is capable
of processing one patient per day in its current configura
tion. If justified by clinical applications, the throughput of
the system can be increased at least 50-fold for routine
scanning. The greatest amount of time is currently spent in
locating cells suitable for study. Most of the operations can
be performed by competent technical assistants provided
their judgments of cytopathology are examined by a pathol
ogist. Costs for such an instrument can probably not be
justified at present except as a research procedure, and
further estimates of its usefulness must await correlation of
long-term clinical observations with CNM values.
The Swedish population registry and the availability of
complete hospital records along with the exhaustive, pop
ulation-based cohort epidemiological study in this project
CANCER RESEARCHVOL. 38

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

Computerized
ensure that clinical course and prognosis is certain to be
obtained for all of the patients, as it has been for the last 2
years.
This study has demonstrated that individual cells ob
tamed from human tumors by fine-needle aspiration biopsy
are suitable for high-resolution
image analysis. CNM
through application of weighting optimization also allows
application of cytopathobogical data to the entire spectrum
of clinical data such as chemical, radiological, or epidemi
obogical information about patients.
Application of CNM to other solid tumors will be a major
part of the future for this apparatus. Not only breast tumors
but also other epithelial tumors (with only one or a few
major cell types) such as those of the prostate and thyroid,
as well as lymphomas, are suitable for study. We believe
that CNM, with its great reproducibility, can be a remarkably
useful aid in helping the pathologist to more exact grading
of the malignancy of a tumor.

REFERENCES

9.
10.
11.
12.
13.
14.
15.
16.

Nuclear Morphometry

Reproducibility of Histopathologic Classification. Cancer, 19: 75-82,

1966.
Einarson, L. Gallocyanin-Chromlackreaktion als Grundlage eines ZellS
quivalentbildes. Laeknabladid, 20: 142-153, 1934.
Einarson, L. On the Theory of Gallocyanin-chromalum Staining and Its
Application for Qualitative Estimation of Basophilia. A Selective Staining
of Exquisite Progressivity. Acta Pathol. Scand., 27: 82-92, 1951.
Granlund, G. H. Fourier Pre-processing for Handprint Character Recog
nition. IEEE Trans. Computers, C-21: 195-201 , 1972.
Haralick, A. M. , Shanmugam, K., and Dinstein, I. Textural Features for
Image Classification. IEEE Trans. Systems Man Cybernetics, SMC-3:
610-622,1973.
Holmquist, J. On Analytic Methods and Software Design for Computer
Processing of Digitized Microscopic Cell Images. Doctoral thesis, Upps
ala University, Uppsala, Sweden, 1977.
Kendall, M. G. Rank Correlation Methods. London: Charles Griffin, and
Co., Ltd., 1948.
Kendall, M. G., and Stuart, L. A. The Advanced Theory of Statistics, Vol.
2. London: Charles Griffin, and Co., Ltd., 1967.
Nicolini, C., Giaretti, W., De Saive, C., and Kendall, F. The G0-G1
Transition of WI 38 Cells. II. Geometric and Densitometric Texture

Analysis.Exptl. Cell Res.,106:119-125,1977.

17. Norberg, A. OSIRIS-Electronics for Computer Control of an Image
Scanner with a Vibrating Prism and a Diode-Array. Doctoral Thesis,
Physics IV. Royal Institute of Technology, Stockholm, 1976.
18. Nordin, J. OSIRI5The Opto-mechanical Part of a Scanner Using a
Diode-Array to Digitize Microscope Images. Doctoral Thesis, Physics IV,
Royal Institute of Technology, Stockholm, 1976.
19. Pressman, N. J. Markovian Analysis of Cervical Cell Images. J. Histo

chem.Cytochem.,24: 138-149,1976.
1. Ackerman, L. v. , and del Regato, J. A. Cancer, Diagnosis Treatment and
Prognosis, Ed. 4. St. Louis, Mo.: The C. I.
Mosby Co., 1970.
2. Bengtsson, E. On the Design of Systems for Computer Aided Analysis of
Microscopic Images. Doctoral Thesis, Uppsala University, Uppsala,
Sweden, 1977.
3. Bengtsson, E., Eriksson, 0., Holmquist, J., and Stenkvist, B. A Software
System to Record and Analyze Digitized Cell Images. Computer Progr.
Biomed.,7:233-246,1977.
4. Black, M. M., and Spear, F. D. Nuclear Structure in Cancer Tissues.
Surg. Gynecol. Obstet., 105: 97-102, 1957.
5. Bloom, H. J. G., and Richardson, W. W. Histologic Grading and Prog
nosis in Breast Cancer. A Study of 1409 Cases of Which 359 Have Been
Followed for 15 Years. Brit. J. Cancer, 11: 359-377, 1957.
6. Brownlee, K. A. Statistical Theory and Methodology in Science and
Engineering, Ed. 2. New York: John Wiley & Sons, Inc., 1967.
7. Cramer, H. Mathematical Methods of Statistics. Uppsala, Sweden:
Almqvist & Wiksell, 1945.
8. Cutler, S. J., Black, M. M., Friedell, G. H., vidone, A. A., and Golden
berg, I. S. Prognostic Factors in Cancer of the Female Breast. II.

20. Prewitt, J. M. S., and Mendelsohn, M. L. The Analysis of Cell Images.

Ann. N. Y. Acad. Sci., 128: 1035-1053, 1966.
21. Scarff, R. W., and Torloni, H. Histological Typing of Breast Tumours.

International Histological Classification of Tumours No. 2. Geneva:

WHO, 1968.
22. Schiodt, T. Breast Carcinoma. A Histologic and Prognostic Study of 650
Followed-up Cases. Copenhagen: Ejnar Munksgaard, 1966.
23. Siegel, 5. Nonparametric Statistics for the Behavioral Sciences. New

York: McGraw-Hill,BookCo., 1956.

24. Stenkvist, B., Westman-Naeser, 5., Vegelius, J., Holmquist, J., Nordin,
B., Bengtsson, E., and Eriksson, 0. Analysis of the Reproducibility of

SomeSubjectiveGradingSystemfor BreastCarcinoma.J. Clin. Pathol.,

in press, 1979.
25. Stewart, F. w. Tumors of the Breast Atlas of Tumor Pathology, Sect. Ix.

Washington,D.C.:ArmedForcesInstituteof Pathology,1950.
26. Tschuprow, A. A. Grundbegriffe und Grundprobleme der Korrelation
stheorie. Stuttgart: B. G. Teubner, Verlagsges. MbH, 1925.
27. Tschuprow, A. A. Principles of the Mathematical Theory of Correlation.
New York: William Modge, 1939.

DECEMBER1978
Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

4695

@@1
B. Stenkvist et a!.

a..
@1'@
S

IL

.@

@..

.:.@

@...
Fig. 1. Photomicrograph (initial magnification, x 500) of a typical cell population from a fine-needle aspiration biopsy. Arrows, typical cells that would be
selected for measuring.
Fig. 2. Photomicrograph (initial magnification, x 1000 at N. A. 1.30) of a typical cancer cell nucleus obtained from fine-needle aspiration biopsy and
stained with gallocyanin-chrome alum and the image of the same cell obtained by the computer. Resolution of the computer image is deceptively low
because of limitations of the plotter resolution. There is some geometrical distortion due probably to a display artifact.

4696

CANCER RESEARCH VOL. 38

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

Computerized

I_@._____

Fig. 3. Binary plots of nuclei from 40

cancer cells obtained from a fine-needle
biopsy of a human breast cancer.

S
...
.11

DECEMBER 1978

Nuclear Morphometry

I
I

4697

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.

Computerized Nuclear Morphometry as an Objective Method for

Characterizing Human Cancer Cell Populations
Bjrn Stenkvist, Sighild Westman-Naeser, Jan Holmquist, et al.
Cancer Res 1978;38:4688-4697.

Updated version

E-mail alerts
Reprints and
Subscriptions
Permissions

Access the most recent version of this article at:

http://cancerres.aacrjournals.org/content/38/12/4688

Sign up to receive free email-alerts related to this article or journal.

To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at pubs@aacr.org.
To request permission to re-use all or part of this article, contact the AACR Publications
Department at permissions@aacr.org.

Downloaded from cancerres.aacrjournals.org on July 27, 2015. 1978 American Association for Cancer Research.