ANALYTICAL AND QUANTITATIVE CYTOPATHOLOGY AND HISTOPATHOLOGY

1. 44
Analytical and Quantitative Cytopathology and Histopathology®
0884-6812/21/4302-0044/$18.00/0 © Science Printers and Publishers, Inc.
Analytical and Quantitative Cytopathology and Histopathology®
OBJECTIVE: Many high-grade serous ovarian carcino-
mas are thought to originate from fallopian tube epithe-
lium. Karyometry detects chromatin abnormalities at the
nuclear level using high-resolution computer imaging
analyses. This study hypothesizes that karyometry can
detect nuclear abnormalities of fallopian tube epithelium
in women at high risk as compared to low risk for ovarian
cancer.
STUDY DESIGN: Fallopian tube tissue from 8 women
carrying BRCA1 or BRCA2 alterations (high risk) and
7 women at normal risk were obtained from the tissue
bank at NorthShore University HealthSystem. Tissues
were fixed, paraffin embedded, sectioned, and stained,
followed by high-resolution imaging and karyometric
analysis.
RESULTS: The distribution of nuclear features and
nuclear signatures in tubes from women at high risk
showed a distinct deviation from that of normal risk
cases. The two most segregating features in the discrim-
inant function scores (pixel optical density heterogeneity
Karyometry Identifies a Distinguishing Fallopian
Tube Epithelium Phenotype in Subjects at
High Risk for Ovarian Cancer
Samantha J. Russell, M.D., Gustavo C. Rodriguez, M.D., Michael Yozwiak, M.S.,
Charmi Patel, M.D., Melody Maarouf, M.D., Hubert G. Bartels, M.S.I.E.,
Jennifer K. Barton, Ph.D., Ahyoung Amy Kim, M.S., Swetha Atluri, B.S.,
Peter H. Bartels, Ph.D., M.D.(hon),† Hao Helen Zhang, Ph.D., and
David S. Alberts, M.D.
From the Departments of Medicine, Pathology, Biomedical Engineering, Statistics and Data Science, Optical Sciences, Mathematics,
Pharmacology, Public Health, and Nutritional Sciences, the Cancer Center, and the College of Medicine, University of Arizona, Tucson,
Arizona; Department of Obstetrics and Gynecology, NorthShore University HealthSystem, Evanston, Illinois; and the Department of
Obstetrics and Gynecology, University of Chicago, Chicago, Illinois.
Samantha J. Russell is Internal Medicine Resident, Department of Medicine, University of Arizona.
Gustavo C. Rodriguez is Clinical Professor, Department of Obstetrics and Gynecology, NorthShore University HealthSystem, and Asso-
ciate Clinical Professor, Department of Obstetrics and Gynecology, University of Chicago.
Michael Yozwiak is Principal Research Specialist, University of Arizona Cancer Center.
Charmi Patel is Assistant Professor, Department of Pathology, University of Arizona.
Melody Maarouf is Internal Medicine Resident, Department of Medicine, University of Arizona.
Hubert Bartels is Senior Systems Programmer, University of Arizona Cancer Center.
Jennifer K. Barton is Professor, Department of Biomedical Engineering, University of Arizona.
Ahyoung Amy Kim is Ph.D. Candidate, Statistics Graduate Interdisciplinary Program, Statistics and Data Science, University of Arizona.
Sri Saii Atluri is Medical Student, University of Arizona College of Medicine.
Peter H. Bartels was Professor Emeritus, Departments of Optical Sciences and of Pathology, University of Arizona.
Hao Helen Zhang is Professor, Department of Mathematics, University of Arizona.
David S. Alberts is Professor, Departments of Medicine, Pharmacology, Public Health, and Nutritional Sciences, University of Arizona.
This research was supported in part by a grant from the National Cancer Institute (Arizona Cancer Center Support Grant CA023074),
Bethesda, Maryland. The authors also gratefully acknowledge support from Humberto and Czarina Lopez, Tucson, Arizona, The Jim
Click Family Foundation, Tucson, Arizona, and the J. Russell Skelton Family, Phoenix, Arizona.
Address correspondence to: Hao Helen Zhang, Ph.D., ENR2 S323, University of Arizona, P.O. Box 210089, Tucson, AZ 85721-0089
(hzhang@math.arizona.edu).
Financial Disclosure: Drs. P. H. Bartels and Alberts own U.S. patents related to premalignant lesions and karyometry.

2. Volume 43, Number 2/April 2021 45
Distinct Fallopian Phenotype in High-Risk Patients
and the number of very dark stained pixels) showed a
pronounced shift from normal in high-risk nuclei and
represent a statistically significant difference (p<0.05)
at the nuclear level.
CONCLUSION: Karyometry detected an abnormal mor-
phometric phenotype in nuclei of fallopian tube epitheli-
um from women at high risk for disease as compared to
normal controls. (Anal Quant Cytopathol Histpathol
2021;43:44–51)
Keywords: BRCA1 genes, BRCA2 genes, carcino-
ma, fallopian tube, karyometry, karyometric image
analysis, ovarian cancer, ovarian epithelial carcino-
ma.
Ovarian cancer is the most lethal gynecological
malignancy in women and is responsible for 5%
of all cancer-related deaths in females.1 When de-
tected in early stages, the 5-year survival rate is
as high as 90%.1 Unfortunately, only 15% of cases
are diagnosed at this stage, and ovarian cancer
typically progresses to later, more aggressive stages
before detection in the majority of cases, thereby
resulting in a substantial drop in 5-year survival to
less than 40%.1
Women with mutations in breast cancer sus-
ceptibility genes, BRCA1 and BRCA2, have a 40%
and 18% lifetime ovarian cancer risk, respective-
ly.2 For these women, twice yearly screening has
been advised in the hope that ovarian cancers
might be detected early. Current screening tech-
niques including pelvic examination, transvaginal
ultrasound (TVU), and CA125 blood marker mea-
surement, however, have very poor specificity and
sensitivity.3 Furthermore, annual screening by TVU
and CA125 in high-risk populations is grossly inef-
ficient at detecting the disease in its early stages.4
Because of the limitations in ovarian cancer screen-
ing, it is recommended that women at high risk
for the disease undergo radical preventative mea-
sures, such as a prophylactic bilateral salpingo-
oophorectomy (BSO), when they have either com-
pleted childbearing or by age 35–45.5
Bilateral salpingo-oophorectomy, or the removal
of both ovaries and fallopian tubes, has been asso-
ciated with an 80% reduction in not only ovarian
cancer risk, but also fallopian tube and peritoneal
cancer risk.5 Although removal of the ovaries and
fallopian tubes is an effective method of ovarian
cancer prevention, it is accompanied by adverse
sequelae, including loss of fertility, sudden onset
of menopause, and rapid loss of bone mineraliza-
tion.6 In order to decrease the need for such harm-
ful surgeries, more efficient screening techniques
are required to accurately identify individuals har-
boring lesions at risk for malignant transformation
and to use these techniques to detect the disease in
its early, more curable stages.
Karyometry is a form of digital microscopy
that allows for quantitative analysis of diagnostic
images. All digital images are represented by dis-
crete points or picture elements known as pixels.
Electronically examining images on a pixel-by-
pixel basis offers novel, objective information and
descriptive features that are not readily visible to
the human eye. The difference between visual and
computed features can be demonstrated by read­
ing printed text. Although one can recognize visu-
al information like letters, words, and language,
more complex information, such as the relative
frequency of occurrence of one letter compared to
another, is overlooked. The relative frequencies of
certain letters or vowels compared to others, how-
ever, characterize the text while providing objec-
tive and computed descriptive features. Similarly,
examining the two-dimensional spatial and statis­
tical distribution of pixels in a digitized image
produces computed descriptive features. Through
the power of variance-analytic procedures, nu-
merical representations of these features allow
for subtle, yet consistent, differences to be iden-
tified between images. The computer processing
of digitally represented histopathologic images
provides the unique opportunity to collect quan­
titative, numerical data from an otherwise qualita-
tive source.
Karyometry can ascertain chromatin abnormali-
ties at the nuclear level by utilizing high-resolution
computer imaging analyses. It is designed to ex-
tract karyometric features that are undetectable to
the human eye, thereby providing a level of sen-
sitivity that allows the fine detection of variability
between samples despite the natural heterogene-
ity of nuclei.7 Not only does karyometry have the
ability to accurately detect nuclear abnormalities,
it also has provided objective and statistically valid
measures of chemopreventive efficacy in vivo. For
example, a statistically significant change toward
normalization in the karyometric features in the
ovary and fallopian tube epithelium in women at
increased risk for ovarian cancer treated with the
progestin levonorgestrel supported the notion that
progestins may act as chemopreventive agents
against ovarian and fallopian tube cancer.8 Similar
A
RTICLES

3. 46 Analytical and Quantitative Cytopathology and Histopathology®
Russell et al
studies investigating chemopreventive efficacy of
vitamin A in sun-damaged skin have shown that
karyometry analysis can provide more objective
measures of chemopreventive efficacy.9,10
When pathologists microscopically examine tu-
mor biopsies, they generally assign a grade based
on how abnormal the cells appear to visual inspec-
tion. On the other hand, digital analysis of biopsy
images through karyometry can provide the abil-
ity to compute and document certain features that
are too small to be appreciated by visual observa-
tion. For example, a previous study using karyo-
metry on ovarian surface epithelium discovered
that histologically normal-appearing epithelium
in women at high risk for ovarian cancer, as well
as in the normal epithelium adjacent to ovarian
cancer, harbors abnormal submicroscopic changes
in the nuclear chromatin, thus suggesting poten-
tial for malignancy.11 Furthermore, studies utiliz-
ing karyometry have uncovered similar significant
relationships between slight changes in the nuclei
of normal, preinvasive, and invasive lesions in var-
ious other tissues including breast, rectal, urinary
tract, prostate and skin.11-16 The ability to detect
distinct changes in nuclear chromatin pattern
between normal, preinvasive, and invasive lesions
provides the opportunity to characterize samples
in a progression curve. In fact, karyometry has
been shown to accurately detect the progression
of precancerous lesions of actinic keratosis to ag-
gressive squamous cell carcinomas through the
comparison of karyometric features.17 Although
further research is required, karyometry possesses
the potential of being an instrumental screening
technique for cancers of many different origins.
It is a widely supported theory that the cell of
origin for many high-grade ovarian cancers may
originate from the fallopian tube epithelium rath-
er than ovarian epithelium.18 This makes the fal-
lopian tube an attractive target for ovarian cancer
screening and prevention. In the current study, we
sought to determine whether karyometry might
be useful in detecting nuclear abnormalities of the
fallopian tube epithelium in women carrying the
BRCA1 or BRCA2 mutation at high risk for ovar-
ian cancer as compared to women with normal risk
factors. The ability to evaluate nuclear abnormali-
ties of fallopian tube epithelium through karyom-
etry may provide a more accurate assessment of
ovarian cancer risk beyond a simple genetic test
for BRCA mutations. Therefore, karyometry could
result in more individualized assessment of dis-
ease risk and the avoidance of unnecessary and
harmful procedures. It may even lead to the detec-
tion of preneoplastic lesions, resulting in appropri-
ate preventative measures and early surgical inter-
vention and, in this way, become a useful element
of precision oncology.
Materials and Methods
Sample Collection
A portion of fallopian tubes included in this ret-
rospective study were collected from an archive
of de-identified samples and received exemptions
from the Human Subjects Review Committees at
the University of Arizona and NorthShore Univer-
sity HealthSystem in Chicago. The remaining sam-
ples were collected under IRB protocol, utilizing
the U.S. common rule and obtaining patient con-
sent when appropriate. All materials were trans-
ferred to the University of Arizona from the Pa­
thology Department at NorthShore University
HealthSystem (Evanston Hospital) on the basis of
a material transfer agreement between these two
institutions.
The “normal risk” fallopian tubes were collect-
ed from women who lacked BRCA1 or BRCA2
mutations or a strong family history of breast and
ovarian cancer and underwent a salpingectomy
for benign gynecologic indications. The “high risk”
fallopian tubes were collected from women who
were positive for BRCA1 or BRCA2 mutations
and who underwent surgery for ovarian cancer
risk reduction. All high-risk cases were examined
carefully by pathologists to rule out the presence
of any neoplasia, including the presence of atypia,
STIC lesions, and cancers. They were therefore
confirmed to be histologically normal, and any
abnormality detected by karyometry was therefore
occurring prior to any histologic transformation.
Tissue Preparation
A total of 15 fallopian tubes were collected, con-
sisting of 8 tubes from women at high risk due to
BRCA1 and BRCA2 gene mutations and 7 tubes
from women at normal risk. Tissue fixation and
staining was strictly controlled to prevent unnec-
essary sources of variability. Tissues underwent
standard fixation in 10% neutral buffered formalin,
followed by paraffin embedding. They were then
sectioned at 5-micron thickness and stained with a
reproducible procedure using synthetic hematox-
ylin and eosin. Human tonsil tissue was utilized
in order to maintain quality control during the

4. Volume 43, Number 2/April 2021 47
Distinct Fallopian Phenotype in High-Risk Patients
staining procedure.7 Images of the fallopian tube
epithelium were captured at high magnification
(100:1) with a high-quality video microscope, utiliz-
ing a high numerical aperture (1.40) apochromatic
oil immersion objective. Images were recorded at
4 to 6 pixels per micron and consisted of a random
accumulation of well-defined nuclei that did not
overlap.7
Karyometric Protocol
Images were analyzed using a semi-automated
segmentation program with manual correction,
thereby isolating individual nuclei for further in-
vestigation.19 Segmentation of nuclei was a cru-
cial step in quantitative image analysis because
any error in defining nuclear borders would have
modified the values utilized by karyometric fea-
tures. For segmentation, an image-processing al-
gorithm was applied for object recognition. This
resulted in chain codes that outlined each indi-
vidual nucleus, thereby specifying the pixels that
generated the nuclear border. Interactive correc-
tions were required throughout the segmentation
process since a single algorithm rarely segments
all nuclei correctly. This program isolated, on aver-
age, 158 nuclei per fallopian tube for a total of 1,257
nuclei in the high-risk group. Likewise, an average
of 172 nuclei per fallopian tube for a total of 1,205
nuclei were isolated in the normal-risk group. Fig-
ure 1 provides a sample representative image of
the fallopian tube epithelium after nuclear seg­
mentation.
Following segmentation, 93 karyometric features
were used to analyze nuclear chromatin pattern.
Global features are the simplest and utilized all
pixels localized in a single nucleus, including such
characteristics as “nuclear area,” “nucleocytoplas-
mic ratio,” “measures of roundness of a nucleus,”
and “total optical absorbance.” These features rep-
resent individual nuclei as a whole and can be
classified as zero-order relationships. First- and
second-order features investigate the differences
or similarities between adjacent pixels, thereby
analyzing the 2-dimensional relationships of con-
tiguous pixels to determine chromatin patterns.
For example, several features summarize chroma­
tin pattern by utilizing “pixel run length” or “mean
pixel absorbance.” Finally, the associations of sev-
eral nuclei throughout a sample are organized
into third-order relationships.20 Essentially, the fea-
tures classified in higher-ordered groups are as-
sociated with more complex linear combinations.
A more descriptive list of the 93 features has been
published previously.11
Statistical Analysis
The 93 karyometric features collectively created
the “nuclear signature,” and data were normalized
to provide comparable data sets. Furthermore,
“nuclear abnormality” was calculated by averag-
ing the standard deviations for all 93 features of
each nucleus.7 The nuclear signature expresses
the values and correlations among all measured
karyometric features, thereby providing a quanti-
tative approach that can discriminate malignancy-
associated changes in the nuclear chromatin from
benign. When using karyometric features to detect
nuclear abnormalities in a tissue biopsy, values
are normalized to a reference data set and ex­
pressed in units of standard deviation, or z-value.
The magnitude of z-value for each feature mea-
sures its deviation from a normal reference set.
Normalized values add stability and provide data
that can be compared to similar samples. The z-
values averaged over all 93 karyometric features
offer a measure of “nuclear abnormality.”
In many cases the nuclear abnormality value
is not particularly sensitive, given that not all 93
karyometric features change between samples. A
Kruskal-Wallis test was utilized to ascertain the
Figure 1 The representative image illustrates nuclei of fallopian
tube epithelium that were randomly selected for further analysis.
Images of the fallopian tube epithelium were captured at high
magnification (100:1) with a high-quality video microscope
utilizing a high numerical aperture (1.40) apochromatic oil
immersion objective. The white outlines represent chain codes
that defined the border of each nucleus during segmentation.

5. 48 Analytical and Quantitative Cytopathology and Histopathology®
Russell et al
significance of very small deviations in data sets
in order to detect the submicroscopic differences
that cannot be appreciated by visual analysis
alone. Since there are 93 statistical tests being
performed, it is necessary to apply a multiple-test
correction such as the Bonferroni correction and
the Benjamini-Hochberg procedure to control for
false positives.21,22 In this work we selected features
with statistically significant value differences, at
p values <0.0001, which allows for a Bonferroni
correction. This test identified several important
features with statistically significant differences
between the high-risk group and the normal-risk
group.
These karyometric features identified by the
Kruskal-Wallis test were then used to construct
the linear discriminant function for separating the
two groups.20 A discriminant function analysis is
widely used in problems of categorizing obser-
vations into different groups, often called classi-
fication problems in statistics. Using the selected
features, a discriminant function analysis builds
a classification rule—in this case, high risk and
normal risk—by developing a function that can
separate different groups. Descriptive statistics
can characterize a set of nuclei by the numeric
values of their features. Finally, Wilks’ lambda
was used to determine a difference between the
nuclear abnormality in high-risk and normal-risk
fallopian tube samples with p value <0.05.
Results
Figure 2 plots the nuclear signature of the sam­
ples, i.e., the standardized scores of 93 features, for
normal-risk fallopian tube epithelium (top panel)
and for high-risk fallopian tube epithelium (bot­
tom panel). It shows that the nuclear feature dis-
tribution of high-risk cases demonstrates a distinct
deviation from that of normal-risk cases. Wilks’
lambda was reduced to 0.63, thereby providing
evidence of a definite difference between high-risk
and normal-risk fallopian tube features. Discrimi-
nant function analysis highlighted two of the most
segregating karyometric features, including pixel
optical density heterogeneity and the number of
very-dark-stained pixels. The feature “pixel optical
density heterogeneity” measured the degree of
diversity in the pixel optical density, and the fea-
ture “number of very-dark-stained pixels” mea-
sured the total number of dark-stained pixels in
the nucleus. Figure 3 displays the distribution
of discriminant function scores for nuclei from
Figure 2
The graphs represent the
nuclear signatures for
both high-risk (bottom) and
normal-risk (top) groups. Each
bar represents a single
karyometric feature and
provides a quantitative value
for the 93 global features
selected in the karyometric
analysis. In other words, each
bar represents a z-value that
was calculated using the
absolute difference of the
corresponding reference
feature mean divided by the
corresponding standard
deviation for the reference
feature. There are marked
abnormalities in at-risk from
normal, as seen by the
variation in z-values.

6. Volume 43, Number 2/April 2021 49
Distinct Fallopian Phenotype in High-Risk Patients
high-risk cases (in black) and for nuclei from
normal-risk cases (in grey). It clearly shows that
the discriminant function for high-risk nuclei dis-
played a pronounced shift (p<0.05) away from
that of normal-risk cases, towards lower (more
negative) values. Additionally, utilizing these two
segregating karyometric features in a bivariate plot
with 95% ellipses for the case mean values dem-
onstrated a statistically significant difference (p<
0.05) between the two experimental groups at the
nuclear level (Figure 4). It was observed that nu­
clei from high-risk tissues, falling onto the upper
portion of Figure 4, had lower values of the pixel
optical density heterogeneity than those from nor-
mal risk tissues.
Discussion
This small, exploratory study was designed to
assess whether or not karyometric analysis could
detect changes in the nuclear chromatin pattern of
fallopian tube epithelium from women at high risk
for ovarian cancer. The results suggest that there
were, in fact, detectable and significant changes in
the chromatin signature of tubes in the high-risk
group as compared to the normal-risk controls.
As illustrated in Figure 2, there was a distinct dif-
ference in the nuclear signature of fallopian tube
epithelium of high-risk samples as compared to
normal controls. Changes in the nuclear signature
could represent the earliest transformation of histo-
logically normal-appearing tissue to premalignant
lesions. Previous studies confirmed that seemingly
normal tissue might harbor premalignant lesions
that can be detected through nuclear karyometry.12
For example, research conducted by Anderson et al
uncovered significant differences between nuclei in
normal-appearing breast tissue containing remote
cancer as compared to nuclei in breast tissue with
no cancer.12 These differences indicate a deviation
from normal, thereby suggesting the potential to
progress to malignancy.
Pixel optical density heterogeneity and the num-
ber of very-dark-stained pixels were computed as
the most segregated karyometric features. In fact,
these features provided statistically significant evi-
dence that nuclear chromatin patterns in high-risk
women differed from those in normal-risk controls
(Figure 4). When cells undergo molecular changes,
they develop altered chromatin structure. By utiliz­
ing karyometric nuclear analysis, it is possible to
characterize chromatin patterns by describing the
organization of pixel combinations.23
It is widely appreciated that BRCA1 or BRCA2
mutations predispose women to high-grade serous
ovarian cancer and that the fallopian tubes are
believed to be the origin of disease.18 Therefore,
abnormalities between the nuclear signatures of
fallopian tubes from women with these mutations
and normal controls were anticipated. Given the
preliminary nature of this study, one of the weak-
nesses is the small number of samples analyzed. In
the future, we hope to expand upon this research
and analyze, in a blinded fashion, a much larger
number of fallopian tubes in order to better under-
Figure 3
The distribution of normalized
discriminant function scores
for the nuclei of fallopian
tube epithelium representing
women at normal risk cases
(in grey) and women at high
risk cases (in black). The
discriminant analysis utilized
only karyometric features that
maximally reduced Wilks’
lambda (i.e., maximally
separated high-risk from
normal-risk epithelium). There
is a significant separation of the two curves (p<0.05). A shift to the lower score values as seen in the black bars indicates higher deviation
from normal.

7. 50 Analytical and Quantitative Cytopathology and Histopathology®
Russell et al
stand the difference between high-risk and normal-
risk nuclear features.
Finally, we are collaborating with biomedical
engineers at the University of Arizona in the de-
velopment of a novel miniature endoscopic de-
vice that can be threaded through the vagina and
into the fallopian tubes to collect fallopian tube
cells. This revolutionary device would eliminate
the need to surgically remove the tubes in order
to accurately assess cancer risk. With the assis-
tance of this device, we are hopeful that karyom-
etry can offer a novel and noninvasive mechanism
for ovarian cancer screening, thereby encouraging
noninvasive prevention strategies and early detec-
tion of the disease, ultimately contributing to pre­
cision oncology in gynecological malignancies.
Acknowledegments
This research was supported in part by a grant
from the National Cancer Institute (Arizona Can-
cer Center Support Grant CA023074), Bethesda,
Maryland. The authors also gratefully acknowl-
edge support from Humberto and Czarina Lopez,
Tucson, Arizona; The Jim Click Family Foundation,
Tucson, Arizona; and the J. Russell Skelton Fam-
ily, Phoenix, Arizona. In addition, we would like
to dedicate the present work to our brilliant co-
author, Peter Bartels, Ph.D., one of the founding
fathers of quantitative histopathology, who con-
tinued to be productive and amazingly creative
through his last days. He will forever be missed
and revered.
Contributions
Dr. Samantha Russell collated the karyometric
data and composed the manuscript. Dr. Gustavo
Rodriguez collected all the participant tissue sam-
ples, facilitated the production of the histopatho-
logical slides, and provided scientific input to the
project and manuscript. Dr. Hao Helen Zhang pro-
vided statistical advising, data interpretation, and
assistance throughout the editing process. Michael
Yozwiak provided oversight on the segmentation
of the microscopic studies and input into the inter-
pretation of the generated data. Dr. Charmi Patel
provided histopathological review of all samples
and input into selection of microscopic images
for karyometric analyses. Dr. Melody Maarouf
provided the segmentation of all histopathological
specimens. Hubert Bartels collected and main-
tained all analytical data. Dr. Jennifer Barton par-
ticipated in the discussion, interpretation of study
data, and editing of the manuscript as well as
potential application to the future collection of
tissue samples. Ahyoung Amy Kim provided sta-
tistical input throughout the editing process and
participated in the discussion. Sri Saii Atluri partic-
ipated in discussion of data and has been integral
in continuing segmentation of fallopian tubes for
future projects. Dr. Peter Bartels presided over the
final statistical analyses and the interpretation of
results. Dr. David Alberts developed experimental
design, supervised all laboratory functions, and
provided major input into the manuscript drafts.
References
1. American Cancer Society, “Cancer Facts and Figures 2016,”
2016
2. Chen S, Parmigiani G: Meta-analysis of BRCA1 and BRCA2
penetrance. J Clin Oncol 2007;25:1329-1333
3. Buys SS, Partridge E, Greene MH, Prorok PC, Reding D,
Riley TL, Hartge P, Fagerstrom RM, Ragard LR, Chia D,
Izmirlian G, Fouad M, Johnson CC, Gohagan JK; PLCO
Figure 4 The image above shows a bivariate plot of two
karyometric features that displayed the strongest weight in the
discriminant function, including pixel optical density
heterogeneity (ordinate) and the number of densely stained
pixels (abscissa). This graph demonstrates a statistical difference
(p<0.05) between the means and 95% confidence regions
between these two features. Triangles represent high risk, circles
represent normal risk, crosses represent means, and ellipses
represent confidence intervals.

8. Volume 43, Number 2/April 2021 51
Distinct Fallopian Phenotype in High-Risk Patients
Project Team: Ovarian cancer screening in the Prostate,
Lung, Colorectal and Ovarian (PLCO) cancer screening trial:
Findings from the initial screen of a randomized trial. Am J
Obstet Gynecol 2005;193:1630-1639
4. Woodward E, Sleightholme H, Considine A, Williamson S,
McHugo J, Cruger D: Annual surveillance by CA125 and
transvaginal ultrasound for ovarian cancer in both high-risk
and population risk women is ineffective. BJOG 2007;114:
1500-1509
5. Finch APM, Lubinski J, Møller P, Singer CF, Karlan B, Senter
L, Rosen B, Maehle L, Ghadirian P, Cybulski C, Huzarski
T, Eisen A, Foulkes WD, Kim-Sing C, Ainsworth P, Tung
N, Lynch HT, Neuhausen S, Metcalfe KA, Thompson I,
Murphy J, Sun P, Narod SA: Impact of oophorectomy on
cancer incidence and mortality in women with a BRCA1 or
BRCA2 mutation. J Clin Oncol 2014;32:1547-1553
6. Finch A, Metcalfe KA, Chiang JK, Elit L, McLaughlin J,
Springate C, Demsky R, Murphy J, Rosen B, Narod SA: The
impact of prophylactic salpingo-oophorectomy on meno-
pausal symptoms and sexual function in women who carry
a BRCA mutation. Gynecol Oncol 2011;121:163-168
7. Ranger-Moore J, Alberts DS, Montironi R, Garcia F, Davis J,
Frank D, Brewer M, Mariuzzi GM, Bartels HG, Bartels PH:
Karyometry in the early detection and chemoprevention of
intraepithelial lesions. Eur J Cancer 2005;41:1875-1888
8. Rodriguez GC, Kauderer J, Hunn J, Thaete LG, Watkin WG,
Russell S, Yozwiak M, Basil J, Hurteau J, Lele S, Modesitt
SC, Zivanovic O, Zhang HH, Bartels PH, Alberts DS: Phase
II trial of chemopreventive effects of levonorgestrel on ovar-
ian and fallopian tube epithelium in women at high risk
for ovarian cancer: An NRG Oncology Group/GOG Study.
Cancer Prev Res (Phila) 2019;12:401-412
9. Bartels PH, Ranger-Moore J, Stratton MS, Bozzo P, Einspahr
J, Liu Y, Thompson D, Alberts DS: Statistical analysis of
chemopreventive efficacy of vitamin A in sun-exposed, nor-
mal skin. Anal Quant Cytol Histol 2002;24:185-197
10. Alberts D, Ranger-Moore J, Einspahr J, Saboda K, Bozzo P,
Liu Y, Xu XC, Lotan R, Warneke J, Salasche S, Stratton S,
Levine N, Goldman R, Islas M, Duckett L, Thompson D,
Bartels P, Foote J: Safety and efficacy of dose-intensive oral
vitamin A in subjects with sun-damaged skin. Clin Cancer
Res 2004;10:1875-1880
11. Brewer MA, Ranger-Moore J, Baruche A, Alberts DS, Greene
M, Thompson D, Liu Y, Davis J, Bartels PH: Exploratory
study of ovarian intraepithelial neoplasia. Cancer Epidemiol
Biomarkers Prev 2005;14:299-305
12. Anderson N, Houghton J, Kirk SJ, Frank D, Ranger-Moore
J, Alberts DS, Thompson D, Bartels PH: Malignancy-
associated changes in lactiferous duct epithelium. Anal
Quant Cytol Histol 2003;25:63-72
13. Ranger-Moore J, Frank D, Lance P, Alberts D, Yozwiak M,
Bartels HG, Einspahr J, Bartels PH: Karyometry in rectal
mucosa of patients with previous colorectal adenomas. Anal
Quant Cytol Histol 2005;27:134-142
14. Scarpelli M, Montironi R, Tarquini LM, Hamilton PW, Lopez
Beltran A, Ranger-Moore J, et al: Karyometry detects sub-
visual differences in chromatin organisation state between
non-recurrent and recurrent papillary urothelial neoplasms
of low malignant potential. J Clin Pathol 2004;57:1201-1207
15. Bartels PH, Montironi R, Duval da Silva V, Hamilton PW,
Thompson D, Vaught L, Bartels HG: Tissue architecture
analysis in prostate cancer and its precursors: An innovative
approach to computerized histometry. Eur Urol 1999;35:484-
491
16. Glazer ES, Bartels PH, Liang J, Prasad AR, Yozwiak ML,
Krutzsch M, Clark C, Kha S, Bartels HG, Einspahr JG,
Alberts DS, Krouse RS: High proportion of nuclear pheno-
type identifies aggressive cutaneous squamous cell carcino-
ma. Anal Quant Cytopathol Histpathol 2015;37:302-309
17. Ranger-Moore J, Bozzo P, Alberts D, Einspahr J, Liu Y,
Thompson D, et al: Karyometry of nuclei from actinic kera-
tosis and squamous cell cancer of the skin. Anal Quant Cytol
Histol 2003;25:353-361
18. Piek JMJ, Verheijen RHM, Kenemans P, Massuger LF, Bulten
H, Van Diest PJ: BRCA1/2-related ovarian cancers are of
tubal origin: A hypothesis. Gynecol Oncol 2003;90:491
19. Korde VR, Bartels H, Barton J, Ranger-Moore J: Automatic
segmentation of cell nuclei in bladder and skin tissue for
karyometric analysis. Anal Quant Cytol Histol 2009;31:83-89
20. Glazer ES, Bartels PH, Prasad AR, Yozwiak ML, Bartels HG,
Einspahr JG, Alberts DS, Krouse RS: Nuclear morphometry
identifies a distinct aggressive cellular phenotype in cutane-
ous squamous cell carcinoma. Cancer Prev Res 2011;4:1770-
1777
21. Bonferroni C: Teoria statistica delle classi e calcolo delle
probabilita. Pubbl del R Ist Super di Sci Econ e Commericiali
di Firenze 1936;8:3-62
22. Benjamini Y, Hochberg Y: Controlling the False Discovery
Rate: A Practical and Powerful Approach to Multiple Test-
ing. J R Stat Soc Series B (Methodological) 1995;57:289-300
23. Weyn B, Jacob W, da Silva VD, Montironi R, Hamilton PW,
Thompson D, Bartels HG, Van Daele A, Dillon K, Bartels PH:
Data representation and reduction for chromatin texture in
nuclei from premalignant prostatic, esophageal, and colonic
lesions. Cytometry 2000;41:133-138