Original Article
Evaluation of an Automated Digital Imaging System,
Nextslide Digital Review Network, for Examination
of Peripheral Blood Smears
Hongbo Yu, MD, PhD; Chi Young Ok, MD; Adam Hesse, BS; Peter Nordell, MEM, MBA; Diane Connor, MT; Erica Sjostedt, MT;
Liberto Pechet, MD; L. Michael Snyder, MD
Context.—Several automated digital imaging systems
Nhave
been introduced in recent years to improve turnaround time and proficiency in examining peripheral blood
smears in hematology laboratories.
Objective.—To evaluate a new automated digital imaging system, Nextslide Digital Review Network (Nextslide),
for examination of peripheral blood smears.
Design.—We evaluated 479 peripheral blood smears,
of which 247 (51.6%) were included for comparison of
Nextslide and manual white blood cell differential
counts and morphology evaluation, 204 (42.6%) were
included for comparison of Nextslide and CellaVision
(DM96) differential counts, and 28 (5.8%) were neonatal samples examined for enumeration of nucleated red
blood cells.
he primary step in assessing hematologic function and
T
the presence of disease is an examination of the
cellular elements in the peripheral blood. Examination of
1
the blood frequently provides important information that
aids in the diagnosis of hematologic diseases and may
suggest further testing.2 The complete blood cell count
(CBC) is one of the most commonly ordered laboratory
tests. A CBC test may be ordered as a simple count of
blood elements and red blood cell (RBC) indices or as a
test that includes a white blood cell (WBC) differential
count. Most laboratories use automated analyzers for
CBCs. These analyzers can flag abnormalities in RBCs,
WBCs, and platelets and can, thus, trigger examination of
the peripheral blood smear. Automated blood cell analysis
and visual morphologic examination of blood cells each
offer significant advantages and disadvantages.3 It is
Accepted for publication August 22, 2011.
From the Department of Hospital Laboratories, University of
Massachusetts Memorial Medical Center, Worcester (Drs Yu, Ok,
Pechet, and Snyder and Mss Connor and Sjostedt); and Nextslide
Imaging LLC, Cleveland, Ohio (Messrs Hesse and Nordell).
Adam Hesse is a system architect and Peter Nordell is a managing
partner at Nextslide Imaging LLC. The other authors have no relevant
financial interest in the products or companies described in this article.
Presented at the annual meeting of the American Society for Clinical
Pathology; October 29, 2010; San Francisco, California.
Reprints: Hongbo Yu, MD, PhD, Department of Hospital Laboratories
and Department of Pathology, University of Massachusetts Memorial
Medical Center, One Biotech Park, 365 Plantation Street, Worcester,
MA 01605 (e-mail: [email protected]).
660
Arch Pathol Lab Med—Vol 136, June 2012
Results.—Results from both method comparisons showed
excellent correlation for all major white blood cell classes with
correlation coefficients ranging from 0.70 to 0.99. Evaluation of
white blood cell, red blood cell, and platelet morphology also
showed good correlation among methods. White blood cell
preclassification capability in the system was evaluated for rate
and accuracy. Leukopenic samples demonstrated markedly
decreased review time with Nextslide. Enumeration of nucleated red blood cells showed good correlation among methods.
Conclusions.—Our evaluation of Nextslide shows excellent correlation when compared with conventional manual
differentials and CellaVision (DM96) differentials for evaluation of peripheral blood smears.
(Arch Pathol Lab Med. 2012;136:660–667; doi: 10.5858/
arpa.2011-0285-OA)
unlikely that either approach will totally supplant the
other. Automated analyzers provide clinical laboratories
with the capability to efficiently handle a large volume of
samples and provide superior accuracy and precision
in quantitative blood cell measurements. However, the
complexity and remarkable variation of formed blood
elements offer formidable challenges for any automated
analyzer. A significant proportion of samples still require
manual examination of the blood-test slide for definitive
diagnosis of morphologic abnormalities.
Examination of peripheral blood slides by light microscopy remains one of the major labor-intensive procedures
in the hematology laboratory, requiring a highly trained
staff. In addition, manual differential counts remain
subject to significant statistical variance.4 Thus, automated
morphology-analysis systems have been developed. Ideally, these automated systems should be able to morphologically analyze a peripheral smear-test slide, with
reproducible results faster than, or at least as fast as,
medical technologists. The first automated morphological
analysis system introduced was the Cydac Scanning
Microscope System (Cydac, Uppsala, Sweden) in 1966.5
A few digital analysis systems in the past include the
LARC (leukocyte automatic recognition computer; Corning Medical, Raleigh, North Carolina), the Hematrak
(Geometric Data, Wayne, Pennsylvania), the Coulter Diff 3
and Diff 4 (Coulter S-Plus WBC histogram, Coulter
Electronics, Hialeah, Florida), and the ADC 500 (Abbott
Laboratories, Abbott Park, Illinois).6 However, those
Nextslide, an Automated Digital Imaging System—Yu et al
instruments were slow and offered a limited degree of
automation and, thus, failed to provide significant
improvements in workflow. Several automated digital
imaging systems have been introduced in recent years to
improve turnaround time and proficiency.7–9 CellaVision
(Lund, Sweden) has been the system most often used for
this purpose. CellaVision introduced 3 new systems, in
2001 the Diffmaster Octavia, in 2004 the DM96, and in
2010, the CellaVision DM1200 was introduced in the
United States through Sysmex America, Mundelein,
Illinois. These instruments scan slides at low power,
identifying WBCs, and then take digital images at high
magnification. The images are analyzed by an artificial
neural network based on a database of cells and
preclassified according to WBC class. The cells are
presented on a computer screen for confirmation or
reclassification.
We evaluated a new automated digital imaging system,
Nextslide Digital Review Network (Nextslide), for examining peripheral blood smears. This system was codeveloped by Nextslide Imaging (Cleveland, Ohio) and the
Department of Hospital Laboratories, University of
Massachusetts Memorial Medical Center (Worcester).
The goal of this study was to evaluate the clinical accuracy
and operational efficiency of this new image system in
comparison to the standard reference manual differential10
and to the CellaVision (DM96). Within this study, we
specifically evaluated 4 areas of performance important to
our operation: (1) classification of WBCs, (2) manually
produced peripheral blood slides, (3) leukopenic samples,
and (4) neonatal samples.
MATERIALS AND METHODS
Nextslide Digital Review Network
The Nextslide Digital Review Network consists of a highresolution slide scanner, a hosted software-application suite
running in a data center accessed through the Internet, and an
interface with the hospital laboratory information system. The
hosted application suite serves the same functions as traditional,
installed software, but the maintenance and support is fully
managed by Nextslide Imaging. The suite includes an imageprocessing application, a management application, and an online
review application. The image-processing application accepts
completed images from the slide scanner and processes them to
identify and classify WBCs. Users log into the online review
application and review the processed and raw images, as well as
the classification results, from their Web browser.
Scanning.—When a peripheral blood slide is placed on the
slide scanner, snapshot images of the test area and of the barcoded slide label are obtained and uploaded to the data center.
The laboratory information system interface is queried for the
upstream CBC results; that information is used to optimize the
scanning process and to provide part of the subsequent review
information. The low-resolution peripheral blood smear image is
then analyzed, and a scan location is selected for each peripheral
blood slide. The scan coordinates are transmitted back to the
scanner, and the high-resolution scanning begins. Slides are
loaded onto trays for scanning, and each tray holds 5 slides. The
throughput of the instrument is 26 slides per hour. A single slide
can be loaded into the instrument for rush cases or STAT review.
Image Analysis.—Once scanning is completed, the scanner
uploads the high-resolution image (3100) to the data center. The
image-processing applications analyze each image. The WBCs
are located, an image of each is cropped and copied, and each
WBC image is classified.
Slide Review.—An experienced technician reviews each case,
accessing the image remotely with a Web browser. The reviewer
Arch Pathol Lab Med—Vol 136, June 2012
examines 110 WBC images and some portion of the fullresolution slide image. The WBC images are presented to the
reviewer in a gallery grouped by cell class. The reviewer
confirms the classification and notes morphologic abnormalities.
The peripheral blood smear image is reviewed to evaluate the
RBC and platelet morphology and to estimate a platelet count in
a manner similar to using a manual microscope. Reviewers can
adjust brightness and zoom throughout the review as necessary;
zoom is used in a manner similar to ‘‘flipping’’ objectives on a
manual microscope, and because the scanned image is made at
3100 magnification, all image data used in the review is available
at full 3100. Additionally, CBC results from the blood analyzer
can be extracted from the laboratory information system and
presented in the user interface to assist in the review process.
When the technician has completed the review, the test results
are uploaded to the laboratory information system.
Hematology Review.—If necessary, the technician can flag the
case for hematology review, and the network will notify the
physician to access the case. Hematology review is performed
using the same application and image data as above but with a
different user profile, which optimizes the workflow for the
physician.
Blood Samples
Cases were retrieved from files archived from the hematology
laboratory at the Department of Hospital Laboratories, University of Massachusetts Memorial Medical Center (Worcester).
This study was approved by the University of Massachusetts
Memorial Medical Center Institutional Review Board.
We evaluated 479 peripheral blood smears in total. For
comparison between Nextslide and manual WBC differential
counts and morphology evaluation, we examined 247 peripheral
blood slides (51.6%), including 50 blood samples (10.4%) within
reference range and 197 blood samples (41.1%) with abnormalities flagged by the automated hematology analyzers. For
comparisons between Nextslide and CellaVision (DM96), 102
samples (21.3%) within reference range, and 102 samples (21.3%)
with abnormalities were selected. The blood samples with
abnormal results included various clinical conditions, such as
acute myeloid leukemia, acute lymphoblastic leukemia/lymphoma, myelodysplastic syndrome, myeloproliferative neoplasms,
chronic lymphocytic leukemia, other mature B-cell lymphomas,
plasma cell neoplasms, T-cell prolymphocytic leukemia, infections, chronic inflammatory conditions, leukopenia, and anemia
(Table 1). The WBC counts ranged from 200/mL to 197 400/mL
(reference range, 4500–11 000/mL). In addition, 28 neonatal
samples (5.8%) were studied to address the use of smaller
amounts of blood for examination and enumeration of nucleated
RBCs.
Methods
Peripheral blood smears were prepared either automatically or
manually. When prepared automatically, they were made and
stained either by Beckman Coulter LH785 (Beckman Coulter, Inc,
Miami, Florida) or a Sysmex SP-1000i (Sysmex America, Inc,
Mundelein, Illinois). The manually prepared peripheral blood
smears were then stained using Hematek (HemaTechnologies,
Lebanon, New Jersey) or Midas (Thomas Scientific, Swedesboro,
New Jersey) stainer with Wright-Giemsa stain.
The first segment of the study was done primarily to compare
manually prepared peripheral blood smears on Nextslide with a
manual microscope, with the focus on leukopenic samples. Our
previous experience with CellaVision found that it had limited
performance with manually prepared smears and leukopenic
samples. Poor performance on manually prepared smears has
required the addition of a slide maker at any location that
employs a digital microscope. Each peripheral blood smear was
then reviewed using both the reference method (manual) and the
test method (digital) in parallel. One of 2 experienced medical
technologists or a pathology resident performed all the reviews
using both methods. Reviews included a 100-cell differential
Nextslide, an Automated Digital Imaging System—Yu et al
661
Table 1.
Clinical Conditions Included in This Study
Clinical Conditions
Acute myeloid leukemia
Acute lymphoblastic leukemia/lymphoma
Myelodysplastic syndrome and myeloproliferative neoplasm
Chronic lymphocytic leukemia and other mature B-cell lymphoma
Plasma cell neoplasm
T-cell prolymphocytic leukemia
Infections and chronic inflammatory conditions
Leukopenia
Anemia
WBC count and a morphologic examination for WBC, RBC, and
platelet abnormalities. The cell counts for each peripheral blood
smear by each method were compared, and any cases where the
results of the 100-cell count were outside the Rumke limits4 were
rereviewed by both methods as 200-cell counts. Twenty-eight
neonatal manually prepared peripheral blood smears were also
included in the study. Enumeration of nucleated RBCs was
performed either by an experienced medical technologist or by a
pathology resident using a light microscope and Nextslide in
parallel for each case.
The second segment of this study was performed on all slides
produced by slide makers to compare Nextslide to CellaVision
(DM96). Two slides were made for each sample. Each slide was
reviewed by 2 different technologists, who used each method to
perform a 200-cell differential count. The results of each 400-cell
differential count were analyzed, and any samples that were not
within the 95% confidence interval were arbitrated by a senior
member of the laboratory staff.
Timing Study
The timing study was performed on 42 leukopenic samples
(WBC range, 200/mL to approximately 1800/mL). Review times
for the manual microscopic method and the Nextslide method
for each test slide were recorded.
Statistical Analysis
Statistical analysis was performed using an Excel spreadsheet (Microsoft, Redmond, Washington) for evaluation of
accuracy.
RESULTS
Accuracy of WBC Differential Counts
For this part of the study, 247 peripheral blood smears
were evaluated as described, the preclassification function
was not used. Nextslide acquires an image, which
includes, on average, 220 WBCs. By default, the user is
presented with the first 110 WBC images (Figure 1) but is
allowed to view and classify all available WBC images.
Accuracy was assessed by computing the regression slope
and the correlation coefficient (r2) between manual
differential counts and Nextslide results (Figure 2). Because of the limited number of slides of promyelocytes,
myelocytes, and metamyelocytes, these cells were grouped
together and reported as immature granulocytes. Strong
correlation was observed with neutrophils (Figure 2, A),
lymphocytes (Figure 2, B), atypical lymphocytes (Figure 2,
F), blasts (Figure 2, G), immature granulocytes (Figure 2,
H), and bands (Figure 2, I), with slopes being close to 1.0
(range, 0.93–1.06), and r2 greater than 0.85. Correlation
coefficients for monocytes (Figure 2, C) and eosinophils
(Figure 2, D) were slightly lower, with the values being
0.82 and 0.77, respectively. The low value for basophils
(Figure 2, E; r2 5 0.56) was not significant because of the
few cells counted on the peripheral blood smears.
662
Arch Pathol Lab Med—Vol 136, June 2012
No. of Cases
12
8
18
11
5
1
112
60
21
Preclassification of WBCs
Nextslide’s preclassification engine uses an artificial
neural network to classify images; therefore, it can be
deployed using any number of cell types. The current
version of Nextslide’s classification engine, however, does
not extract some of the subtle features necessary to
distinguish among some abnormal cell types. These
features are to be added to a future version of Nextslide’s
classification engine. Primarily, Nextslide’s preclassification feature is a productivity tool that eliminates the need
to classify most cells within reference range, which makes
up most of the classification effort. In addition, because the
reviewer receives a quick view of cells classified as normal,
the workflow for time spent on laboratory differential rates
should be reduced because reviewers can quickly evaluate
the patient’s WBC morphology results as within reference
range and move on to study RBC morphology.
The accuracy and classification rates of the preclassification function were analyzed as part of this study. This
function was not enabled for the study; it was developed
in parallel to the study, and its performance was evaluated
by using the manual classification results.
Functionally, a cell image is classified by comparing it to a
set of reference images and deciding within which cell
group the cell image best fits. In total, more than 25 000 cells
were manually classified by an operator in this study on
Nextslide; these manually classified images were used in
2 ways to evaluate the classification function. A reference
library was defined by collecting more than 1500 sample
images from the collection of cells classified in the study. In
addition, a second group of 2907 cell images were collected
and classified using the classification function. These
classification results were compared against the original
classifications evaluated by the technician.
Table 2 summarizes the classification rate and the
accuracy of the classification function. The classification
rate is determined by the total cells classified, divided by
the total cells attempted. Accuracy is defined by the total
number of cells classified correctly, divided by the total
number of cells classified. Both classification rate and
accuracy were evaluated on a per–cell-type basis. The rate
for each WBC class ranged from 75.3% to 96.9%, with the
highest rate in the neutrophil class and the lowest rate in
the basophil class. The accuracy for each WBC class
ranged from 94.6% to 99.7%.
Morphology Review of WBCs, RBCs, and Platelets
The reviewer can perform WBC morphology review
using the morphology codes within the same WBC
classification screen in Nextslide. In addition, Nextslide
allows the reviewer to examine RBC and platelet
morphology at a range of magnifications, with the highest
Nextslide, an Automated Digital Imaging System—Yu et al
Figure 1. Default white blood cell (WBC) review screen in the Nextslide Digital Review Network system. The unclassified WBC images are
presented on the left. Morphology codes for WBC can be entered or edited on the right.
magnification being 3100 (Figure 3). The WBC, RBC, and
platelet morphologies for the 247 peripheral blood smears
were evaluated by the examiner based on the morphology
codes listed in Table 3. Each peripheral blood smear was
scored as present or absent for each morphology code. The
percentage of positive agreement was calculated by
identifying all cases in which each morphology code
was identified in both the reference and test methods,
divided by all cases in which the morphology code was
identified in the reference method alone, that is, truepositive/(true-positive + false-negative). Each morphology
result was grouped by cell type (WBC, RBC, and platelet)
and the value of the overall percentage of positive
agreement was calculated for each cell type (Table 3). In
general, our study demonstrates good correlation between
the microscopic method and the Nextslide method.
In addition, an estimate of the platelet count can be
made. The Nextslide workflow process includes a screen
that allows the user to view any part of the scanned
peripheral blood smear image. A 199-mm by 126-mm
rectangle is annotated on the image, which is an area
similar to a single high-power (3100) field under the
microscope. This area can be moved to any location and
allows the user to estimate platelets in the same manner
used for manual microscopy.
Platelet satellitosis occurs occasionally in peripheral
blood samples collected in ethylenediaminetetraacetic
acid anticoagulant and thus leads to pseudothrombocytopenia. We examined 5 such cases, and Nextslide
successfully detected platelet clumps in all cases (data
not shown).
Leukopenic Samples
Leukopenic samples pose a challenge for any differential count, and they require longer review times than
samples with WBC counts that are within reference range
or are elevated. Nextslide offers a useful feature for
Arch Pathol Lab Med—Vol 136, June 2012
leukopenic samples, allowing the scanning area to be
adjusted based on the WBC count. For leukopenic
samples, the scanning size can be expanded to capture a
sufficient number of cells. In our study, we included
60 leukopenic samples with WBC count ranging from
200/mL to 1800/mL. The WBC images from a representative leukopenic sample and a blood sample within
reference range are shown in Figure 4. The quality of the
cell images for the leukopenic samples is comparable to
the samples within reference range. The average time to
scan and capture images for samples with normal WBC
count results is 2.3 minutes. For leukopenic samples, the
time varies based on the WBC counts. For samples with
WBC counts around 1500/mL, it usually takes approximately 4 minutes. The longest time was 15 minutes for a
peripheral blood smear with a WBC count of 200/mL. We
conducted 100-cell differential counts on 59 leukopenic
samples. Nextslide identified 21 WBC images on one
sample with a WBC count of 200/mL. We also compared
the review time of 42 leukopenic samples between a manual
WBC differential count and Nextslide. The mean review
time for the manual differential count was 10.9 (SD 4.28)
minutes. Without preclassification, Nextslide shows a mean
review time of 2.83 (SD 0.95) minutes. Therefore, our result
reveals that Nextslide can evaluate leukopenic samples and
does so with a markedly decreased review time.
Nucleated RBC Enumeration in Neonatal Samples
Neonatal samples are unique because of their smaller
sample volume and high frequency of nucleated RBCs.
In addition to the 247 samples studied above, peripheral
blood smears from 28 neonatal samples were also
examined. Nextslide clearly identified nucleated RBCs
(data not shown). Enumeration of nucleated RBCs was
performed for each peripheral blood smear and was
expressed as the number of nucleated RBCs per 100
WBCs. The correlation coefficient was 0.87 (Figure 5).
Nextslide, an Automated Digital Imaging System—Yu et al
663
Figure 2. Correlation of white blood cell (WBC) classification by the Nextslide Digital Review Network system and by manual differential
microscopic method. Correlation of neutrophils (A), lymphocytes (B), monocytes (C), eosinophils (D), basophils (E), atypical lymphocytes (F), blasts
(G), immature granulocytes (the sum of promyelocytes, myelocytes, and metamyelocytes) (H), and bands (I) between the 2 methods. Abbreviation:
R2, correlation coefficient.
Comparison of Nextslide to CellaVision (DM96)
The comparison of Nextslide to CellaVision demonstrated correlation coefficients similar to that found in the
comparison with the manual microscope (Figure 6). For
major cell classes, including neutrophils (Figure 6, A),
Table 2. Rate and Accuracy of Preclassified Cells
for Each White Blood Cell (WBC) Class by Nextslide
Digital Review Network System
Cell Class
Rate, %
Accuracy, %
Neutrophil
Lymphocyte
Monocyte
Eosinophil
Basophil
96.9
86.6
87.6
93.1
75.3
99.7
95.3
94.6
99.0
95.5
664
Arch Pathol Lab Med—Vol 136, June 2012
lymphocytes (Figure 6, B), monocytes (Figure 6, C),
eosinophils (Figure 6, D), blasts (Figure 6, G), and
immature granulocytes (Figure 6, H), correlation coefficients ranged from 0.92 to 0.97. Correlation coefficients for
basophils (Figure 6, E), atypical lymphocytes (Figure 6, F),
and bands (Figure 6, I) were lower at 0.71, 0.72, and 0.86,
respectively. The primary reason for the increased
statistical match was due to use of a 400-cell differential
count, rather than a 100-cell differential count. Approximately 10% of the slides failed to yield results on the
CellaVision, whereas all the slides were successfully
evaluated on the Nextslide.
COMMENT
Examination of peripheral blood smears by light
microscopy is one of the major labor-intensive diagnostic
Nextslide, an Automated Digital Imaging System—Yu et al
Figure 3. A representative Nextslide Digital
Review Network system review screen for red
blood cell and platelet morphology evaluation (original magnification 3100).
procedures in hematology laboratories. Moreover, leukopenic samples often generate additional burden on
medical technologists to accurately classify the WBCs
and achieve an appropriate turnaround time. We report
the evaluation of a new, automated digital imaging
system, Nextslide Digital Review Network, for examination of peripheral blood smears.
We evaluated 479 peripheral blood smears using
Nextslide, CellaVision, and microscopic manual differen-
tial. Our data show that Nextslide can generate WBC
differential results that are comparable to either method,
with correlation coefficients greater than 0.77 for major
WBC classes. The Nextslide preclassification rate and
accuracy for each WBC class was between 75.3% and
99.7%. This feature helped the reviewers to achieve
accurate results and to reduce review time.
From the evaluation of 28 manually prepared, neonatal
peripheral blood smears, our study showed that Nextslide
Table 3. Percentage of Positive Agreement Between the Nextslide Digital Review Network System and Microscopy
for White Blood Cell (WBC), Red Blood Cell (RBC), and Platelet Morphology Review
Cell Type
Morphology Code
Occurrences, No.
Concordance, %
Overall Concordance, %
WBC
Toxic granulations
Döhle bodies
Vacuolations
Smudge cells
Clefted lymphocytes
Hairy projections
Auer rods
Anisocytosis
Poikilocytosis
Microcytosis
Macrocytosis
Hypochromasia
Polychromasia
Echinocytes (burr cells)
Dacrocytes (teardrops)
Spherocytes
Target cells
Acanthocytes
Ovalocytes
Schistocytes
Stomatocytes
Howell-Jolly bodies
Basophilic stipplings
Clumped platelets
Platelet satellitosis
Large platelets
Giant platelets
28
5
38
32
10
2
2
54
31
7
2
3
35
19
16
6
2
1
28
5
2
1
5
3
5
8
8
78.57
80.00
84.21
65.63
60.00
100.00
100.00
74.07
83.87
71.43
50.00
100.00
80.00
89.47
81.25
66.67
50.00
100.00
71.43
80.00
0.00
100.00
40.00
66.67
100.00
100.00
75.00
76.10
RBC
Platelet
Overall
Arch Pathol Lab Med—Vol 136, June 2012
76.50
87.50
77.12
Nextslide, an Automated Digital Imaging System—Yu et al
665
Figure 5. Correlation of the enumeration of nucleated red blood cells
(NRBCs) in neonatal samples by the Nextslide Digital Review Network
system and by the manual microscopic method. Abbreviations: R2,
correlation coefficient; WBC, white blood cell.
Figure 4. Comparison of white blood cell (WBC) preclassification
review screens of a representative peripheral blood sample with normal
results (A) and a representative leukopenic sample (B). Representative
neutrophil images from a blood sample with normal results (C) and a
leukopenic sample (D) show similar quality.
could identify nucleated RBCs with a correlation coefficient of 0.87; thus, Nextslide can also be used for
numeration of nucleated RBCs.
For leukopenic samples, Nextslide offered distinct
advantages with an adjustable scanning area based on
WBC count, so the reviewer could enlarge the scanning
area and acquire additional WBC images. The image
quality of the cells was comparable to WBC differential
counts that are within reference range or are elevated. The
average review time with the Nextslide system was
2.03 minutes, markedly less than the average review time
of 10.9 minutes for manual differential counts, primarily
because the examiner does not have to hunt for the scarce
WBCs to reach a proper number of cells for review.
Nextslide’s Web-based review application provided
easy, remote accessibility, with Internet access and the
necessary software applications. Nextslide uses a hosted
architecture, which is not a traditional model for labora666
Arch Pathol Lab Med—Vol 136, June 2012
tory equipment. Typically, a piece of laboratory equipment is combined with the management, processing, and
user interface functions necessary to perform the test. In
this model, the scanner’s software is limited to basic
functions necessary to create an image (eg, motor,
illumination, and camera control), as well as the interface
to query the hosted application for instructions. The
hosted application is maintained and monitored by
Nextslide and is responsible for all functions beyond
creation of the image. Therefore, all storage, workflow,
and decision making is performed on the hosted application suite. The primary advantage of this architecture is
accessibility, the ability to access cases from anywhere
with a Web browser and an Internet connection. It can
liberate a reviewer from being physically at the site where
peripheral blood smear slides are prepared. For a central
hematology laboratory that oversees multiple local laboratories, this feature is useful. Shipping peripheral blood
smear slides to a central location for physician review is a
common practice for peripheral blood smears. With
Nextslide, the reviewing physician can evaluate the
review process remotely, without needing the actual
peripheral blood smear slide. In addition, the technician
can flag the cell images in question, and then the
reviewing physician can evaluate the same exact cells.
Nextslide does not require an automated slide maker
and stainer. Manually made slides can also be scanned
and evaluated. Thus, it can be used in a hematology
laboratory or even a physician’s office without an
automated slide maker.
A central data center in the Nextslide system automatically performs a daily, incremental backup of all data and
images. Thus, in cases of local computer malfunctions, the
users can still access the needed images and results on
another computer. The users can also access pertinent
images from the Nextslide data center to build personal
reference libraries. When a computer reconnects after
repair, the user can reload the established reference library
without having to rebuild the library. In addition,
laboratory managers or directors can use this function to
evaluate the competency of technicians and for proficiency testing.
In conclusion, our evaluation of Nextslide showed it
had excellent correlation when compared with conventional, manual differential counts for evaluation of
peripheral blood smears, an accurate preclassification
function, and reduced the review time required for
Nextslide, an Automated Digital Imaging System—Yu et al
Figure 6. Correlation of white blood cell (WBC) classification by the Nextslide Digital Review Network and the CellaVision (DM 96) systems.
Correlation of neutrophils (A), lymphocytes (B), monocytes (C), eosinophils (D), basophils (E), atypical lymphocytes (F), blasts (G), immature
granulocytes (the sum of promyelocytes, myelocytes, and metamyelocytes) (H), and bands (I) between the 2 methods. Abbreviation: R2,
correlation coefficient.
leukopenic samples. The system provided distinct operational advantages. We found the Nextslide Digital
Review Network to be a good option for the hematology
laboratory in need of an automated digital imaging system
because it improved workflow and efficiency and reduced
turnaround time.
References
1. Bell A. Morphology of human blood and marrow cells: hematopoiesis. In:
Harmening DM, ed. Clinical Hematology and Fundamentals of Hemostasis. 4th
ed. Philadelphia, PA: FA Davis Co; 2002:2.
2. Perkins SL. Examination of the blood and bone marrow. In: Lee GR, Foerster
J, Lukens J, et al, eds. Wintrobe’s Clinical Hematology. 10th ed. Philadelphia, PA:
Lea & Febiger; 1998:9–10, 19–22, 27.
3. Felgar RE, Ryan DH. Automated analysis of blood cells. In: Hoffman R,
Benz E, Shattil S, et al, eds. Hematology: Basic Principles and Practice. 4th ed.
Philadelphia, PA: Elsevier Inc; 2005:2673–2686.
Arch Pathol Lab Med—Vol 136, June 2012
4. Rumke CL. Impression of ratio-derived differential leukocyte counts. Blood
Cells. 1985;11(2):311–315.
5. Prewitt JM, Mendelsohn ML. The analysis of cell images. Ann N Y Acad Sci.
1966;128(3):1035–1053.
6. Tatsumi N, Pierre RV. Automated image processing: past, present, and
future of blood cell morphology identification. Clin Lab Med. 2002;22(1):
299–315.
7. Ceelie H, Dinkelaar RB, van Gelder W. Examination of peripheral blood
films using automated microscopy: evaluation of Diffmaster Octavia and
CellaVision DM96. J Clin Pathol. 2007;60(1):72–79.
8. Cornet E, Perol JP, Troussard X. Performance evaluation and relevance of
the CellaVision DM96 system in routine analysis and in patients with malignant
hematological diseases. Int J Lab Hematol. 2008;30(6):536–542.
9. Briggs C, Longair I, Slavik M, et al. Can automated blood film analysis
replace the manual differential?: an evaluation of the CellaVision DM96
automated image analysis system. Int J Lab Hem. 2009;31(1):48–60.
10. Koepke JA, Van Assendelft OW, Brindza LJ, et al. Reference Leukocyte
(WBC) Differential Count (Proportional) and Evaluation of Instrument Methods.
Vol 27, no 4. 2nd ed. Wayne, PA: Clinical and Laboratory Standards Institute;
2007. CLSI Reference Method Approved Standard document H20-A2.
Nextslide, an Automated Digital Imaging System—Yu et al
667