|

Received: 11 July 2017    Accepted: 11 September 2017

DOI: 10.1111/ijlh.12758

ORIGINAL ARTICLE

Digital microscopy as a screening tool for the diagnosis of

hereditary hemolytic anemia

R. Huisjes1  | W. W. van Solinge1 | M. D. Levin2 | R. van Wijk1 | J. A. Riedl3

1
Department of Clinical Chemistry and
Haematology, University Medical Center
Utrecht, Utrecht, The Netherlands
2
Department of Internal Medicine, Albert
Schweitzer Hospital, Dordrecht, The
Netherlands
3
Result Laboratory, Albert Schweitzer
Hospital, Dordrecht, The Netherlands
Correspondence
Dr. Jürgen A. Riedl, Result laboratory,
Albert Schweitzer Hospital, Dordrecht, The
Netherlands.
Email: j.riedl@resultlaboratorium.nl
Funding information
European Seventh Framework Program
Abstract
Introduction: Evaluation of red blood cell (RBC) morphology is an important first step
in the differential diagnosis of hereditary hemolytic anemia. It is, however, labor inten-
sive, expensive, and prone to subjectivity. To improve and standardize the analysis of
RBC morphology as a screening tool in the diagnosis of hereditary hemolytic anemia,
we studied its automated analysis by digital microscopy (DM).
Methods: Blood from 90 patients with hereditary hemolytic anemia and 32 normal
control subjects was analyzed by the CellaVision DM96 Digital Microscope.
Results: All hemolytic RBC abnormalities could be distinguished by the presence of at
least one aberrant red cell type. In particular, the percentage of microcytes was highly
sensitive and specific (AUCROC = 0.97) for RBC membrane disorders, and a cut-­off of
5.7% microcytes was calculated to be optimal to distinguish patients from healthy
controls. Subgroup analysis of patients with RBC membrane disorders revealed addi-
tional distinct differences according to the underlying gene defect. A number of cell
types were significantly elevated in sickle cell anemia patients, such as polychromatic
cells, macrocytes, and poikilocytes. The increase in helmet cells (AUCROC = 0.96) and
hypochromic cells (AUCROC = 0.91) was specific for β-­thalassemia, whereas patients
with pyruvate kinase deficiency showed a significant increased polychromatic cells,
macrocytes, and ovalocytes. Patients with hereditary xerocytosis showed significantly
higher numbers of polychromatic cells, macrocytes, and target cells.
Conclusion: DM holds a promise as a useful screening tool in the diagnosis of heredi-
tary hemolytic anemia by detecting and quantifying distinct morphological changes in
RBCs in patients with various forms of hereditary hemolytic anemia.

KEYWORDS
anemia, digital microscopy, erythrocyte, hereditary hemolytic anemia, morphology

1 |  INTRODUCTION
The differential diagnosis of hereditary hemolytic anemia is often
complicated and time-­consuming. It requires advanced laboratory
tests, such as hemoglobin characterization by high-­performance liq-
uid chromatography (HPLC), osmotic fragility tests, osmotic gradi-
ent ektacytometry, EMA-­binding test, and activity measurements of
1-7
multiple red cell enzymes. In addition, genetic tests are regularly
needed to confirm a suspected disorder, which increases financial
burdens for hospital budgets and healthcare insurance companies.
One of the first steps in the differential diagnosis of hemolytic anemia
is a careful microscopic examination of the morphological appearance
of the red blood cell (RBC). Manual morphology analysis, however,
is time-­consuming, expensive and always subjected to some level of
subjectivity. Manual morphology is also hampered by a lack of stan-
dardization,8 although attempts to standardize the nomenclature and
grading system of peripheral blood smears is an important step for-
ward.9 A relatively inexpensive and high-­throughput gatekeeper that

Int J Lab Hem. 2017;1–10. © 2017 John Wiley & Sons Ltd |  1
wileyonlinelibrary.com/journal/ijlh  
 |
2       HUISJES et al.
gives direction for further diagnostic testing would therefore strongly
increase the diagnostic efficiency in the field of hereditary hemolytic
anemia.
Digital microscopy (DM) is a relatively new technique that au-
tomatically classifies the cells using the accompanying software.
DM has first been validated for leukocyte differentiation, and good
correlations between automated and manual examinations of pe-
ripheral blood smears were observed for a number of different cell
types (eg, neutrophils, eosinophils, basophils, lymphocytes, and
monocytes).10,11 In addition, automated morphology analysis of leu-
kocytes has shown to have good interlaboratory reproducibility12
and is being used and implemented in routine hematology labora-
tories worldwide to detect (malignant) leukocyte abnormalities.
Automated blood smear analysis also has the possibility to analyze
the morphology of RBCs without human interference objectively.
Importantly, RBC morphology can be analyzed on a large number
of cells as between 2000 to 4000 cells can be rapidly analyzed and
13
classified per blood smear. The validation of abnormal RBCs clas-
sified by DM thus far has shown good accuracy for teardrop cells
and schistocytes when comparing pre-­ and postclassification results
and showed good within-­run and between-­run reproducibility for all
cell types.14-16 Preclassification represents the initial classification
by DM, and postclassification represents the manual changes after
the original preclassification by DM.
We describe here for the first time the use of preclassification by
DM as a prescreening tool in different forms of hereditary hemolytic
anemia. We have analyzed blood smears from 90 patients with vari-
ous forms of hereditary hemolytic anemia and 32 healthy controls. We
show that DM is capable of discriminating between healthy controls
and patients suffering from hereditary hemolytic anemia, and we re-
port on the optimal cut-­off values. Moreover, we have determined the
sensitivity and specificity of distinct morphological cell types, which
are significantly different (P ≤ .05) in patients with hereditary hemo-
lytic anemia in relation to healthy controls.
2 |  MATERIALS AND METHODS
2.1 | Patient selection
Blood in EDTA was collected from 90 patients and 32 normal control
subjects. Thirty-­eight patients were previously diagnosed with a RBC
membrane disorder, 10 patients with a hemoglobinopathy (ie, sickle
cell anemia), 12 patients with β-­thalassemia, 10 patients with pyru-
vate kinase deficiency (PKD), 9 patients with a defective antioxida-
tive stress defense (such as G6PD-­deficiency), and 11 patients with
hereditary xerocytosis (HX) due to PIEZO1 defects. The 32 control
subjects were healthy and nonanemic. Patients with RBC membrane
disorders were patients with either hereditary spherocytosis (HS) or
hereditary elliptocytosis (HE) and categorized according to the mo-
lecular defect (eg, SLC4A1, ANK1, and SPTA1). All sickle cell anemia
patients were genotyped with HbSS. One HbSS patient was in an
hemolytic crisis at the moment of blood collection, all other HbSS pa-
tients were stable. From the 12 patients with β-­thalassemia, 9 were
classified as β-­thalassemia minor and 3 as β-­thalassemia major. From
the patients with defective antioxidative stress defense, 8 patients
had G6PD-­deficiency, while 1 was diagnosed with glutamate cysteine
ligase deficiency. Two patients with defective antioxidative stress
defense were at the moment of blood collection recovering from a
recent hemolytic crisis, while others were stable and symptomless at
the time of blood collection.
2.2 | Diagnosis and ethical aspects
All patients were recruited for the “disturbed ion homeostasis in he-
reditary hemolytic anemia”-­study, which was approved by the medical
ethics review board (MERB) of the University Medical Center Utrecht
(UMCU), Utrecht, The Netherlands, under reference code 15/426M.
Diagnosis of all patients was previously established using gold stand-
ards techniques in the diagnosis of hemoglobinopathies, membrane
disorders, and enzymopathies. These techniques include hemoglobin
characterization by HPLC, osmotic fragility tests, osmotic gradient
ektacytometry by Lorrca, eosin-­5-­maleimide (EMA)-­binding test, and
RBC enzyme activity measurements. DNA analysis was carried out to
confirm the diagnosis. Patients receiving transfusion were excluded
from participation in this study. Healthy controls were recruited by
the Minidonordienst from the University Medical Center Utrecht,
Utrecht, The Netherlands, under medical ethics review board (MERB)
approval number 07/125.
2.3 | Routine diagnostic tests
EDTA blood from patients and healthy controls were collected, and
hemocytometry parameters were measured using the Abbott Sapphire
cell counter (Abbott Diagnostics, Santa Clara, CA, USA). Smears were
automatically prepared and stained with May-­Grunwald Giemsa using
the Abbott Cell-­Dyn SMS Slidemaker/Stainer. Subsequently, red
blood cell (RBC) morphology was automatically determined by digital
microscopy (DM) using the CellaVision DM96 (Lund, Sweden) and an-
alyzed using the accompanying software (version 5.0.1 build 11). The
DM is programmed to characterize and classify individual RBCs to the
following cell classes: polychromatic cells, hypochromic cells, micro-
cytes, macrocytes, poikilocytes, target cells, schistocytes, helmet cells,
spherocytes, elliptocytes, ovalocytes, teardrop cells, stomatocytes,
acanthocytes, echinocytes, sickle cells, and Howell-­Jolly bodies. All
distinct morphological cell types were expressed as a percentage of all
cells counted by DM. No postclassification (eg, manual changes after
original classification by DM) was carried out and all red cell types,
therefore, represent the initial classification (=preclassification) by the
CellaVision DM96 digital microscope.
2.4 | Statistical testing and defining cut-­off values
Statistical analysis was performed using IBM Statistics SPPS 6 and
GraphPad Prism 7. One-­way multicomparison ANOVA (post hoc
Tukey) was used to test for differences between the different patient
groups and healthy controls. Receiver operating characteristic (ROC)
HUISJES et al.       3 |
curves were made for the distinct morphological cell types which

Reticulocytes (%)
were significantly different (ANOVA P ≤ .05) between patient groups

18.9 (15.2)***
and healthy controls. Suitable morphological cell types were defined

11.2 (5.8)***

12.5 (7.1)***
7.3 (5.6)***

7.6 (3.5)**

6.0 (6.1)**
6.2 (5.5)**
1.2 (0.3)

7.2 (4.0)
3.1 (1.8)

6.6 (7.2)
as having ROC curves with areas under the curve (AUCROC) ≥ 0.8, and
we defined cell types with AUCROC ≥ 0.9 as excellent to discriminate
between healthy controls and patients. Optimal cut-­off values were
defined by Youden’s index, which is defined as the maximum value

Reticulocytes (*109/L)
of J = sensitivity + specificity—1. Following Youden’s index, where
the probabilities for false-­negatives and false-­positives are equally

285 (197)***

330 (160)***
424 (183)***

511 (342)***

505 (316)***
weighted by definition, we considered sensitivity and specificity to be

232 (206)**
238 (189)**
262 (178)

231 (190)
of equal importance.

156 (84)
58 (17)
3 | RESULTS

Significant differences in hemocytometry parameters compared to healthy controls are depicted with *(P ≤ .05), **(P ≤ .01) or ***(P ≤ .001)(one-­way ANOVA).
MCHC (g/L)

354 (15)***

356 (19)***
354 (13)**

359 (8)***
355 (13)*
334 (15)

349 (13)

326 (18)
333 (16)

327 (22)
356 (9)*
3.1 | Routine diagnostic tests
Baseline hematology parameters are depicted in Table 1. Hemoglobin
levels were significantly different for sickle cell anemia patients

MCH (pg/cell)
(P ≤ .001), β-­thalassemia patients (P ≤ .001), and pyruvate kinase-­

20.7 (3.6)***
35.1 (2.6)***

35.5 (2.6)***
T A B L E   1   Mean (SD) hemocytometry parameters in healthy controls and patients suffering from hereditary hemolytic anemia

29.4 (3.0)
31.3 (2.7)

32.6 (2.3)
30.2 (2.1)
31.2 (2.9)
31.4 (3.0)
33.5 (7.9)

29.6 (3.3)
deficient (PKD) patients (P ≤ .001) in relation to healthy controls.
Reticulocytes were significantly elevated in patients with red blood
cell (RBC) membrane disorders (P ≤ .001), PKD (P ≤ .001), and heredi-
tary xerocytosis (HX) (P ≤ .001). The mean corpuscular volume (MCV)

105.6 (10.0)***
63.3 (10.2)***
was significantly elevated in patients with HX (P ≤ .05) and PKD

90.3 (20.1)

98.6 (6.6)*
88.9 (3.6)
88.4 (6.5)

91.3 (4.9)
86.6 (6.5)
88.2 (7.2)
88.2 (6.7)

90.4 (8.9)
MCV (fL)

(P ≤ .001) if compared to healthy controls but was significantly de-

creased in β-­thalassemia patients (P ≤ .001). The increased MCV in HX
and PKD patients is likely due to the degree of reticulocytosis in these
patients, whereas the decreased MCV in patients with β-­thalassemia

0.28 (0.05)***
0.32 (0.05)***
0.33 (0.04)***
0.37 (0.06)**

0.34 (0.07)**

0.36 (0.05)*
0.42 (0.04)

0.40 (0.04)

0.38 (0.06)

0.39 (0.07)

0.39 (0.05)
is intrinsically related to its pathophysiology.
HCT (L/L)

3.2 | Automated morphology analysis by

digital microscopy
RBC (*1012/L)

3.1 (0.6)***
4.1 (0.7)*
4.7 (0.5)
4.2 (0.7)

4.4 (0.5)
4.0 (0.6)
4.3 (0.8)

3.9 (1.5)
5.2 (1.2)

4.6 (1.6)

4.0 (0.6)
On average, 2615 cells (SD = 603) per sample were analyzed and
classified by DM. The distinct morphological cell types microcytes
(P ≤ .001), poikilocytes (P ≤ .01), schistocytes (P ≤ .01), spherocytes
(P ≤ .01), and acanthocytes (P ≤ .01) were statistically different
9.0 (2.8)***
10.5 (1.9)***
10.9 (1.9)***
14.0 (1.2)
13.1 (2.0)

14.2 (1.2)
12.1 (2.3)
13.2 (1.9)
12.9 (2.0)

12.4 (3.1)

14.1 (1.6)
Hb (g/dL)

in patients with RBC membrane disorders if compared to healthy

controls. In sickle cell anemia patients, polychromatic cells (P ≤ .05),
macrocytes (P ≤ .001), poikilocytes (P ≤ .001), target cells (P ≤ .001),
schistocytes (P ≤ .05), ovalocytes (P ≤ .001), teardrop cells (P ≤ .05),
Female (N)

stomatocytes (P ≤ .01), and sickle cells (P ≤ .001) are significantly

different in relation to healthy controls. In β-­thalassemia, hypochro-
NA
20

3
4
6
7
4
8
5

mic cells (P ≤ .001), target cells (P ≤ .001), helmet cells (P ≤ .001),

and teardrop cells (P ≤ .05) are significantly different in relation
32
38

12
13
10
12
10

11
6
7

9
N

to healthy controls. In PKD, polychromatic cells (P ≤ .001), mac-

rocytes (P ≤ .001), and ovalocytes (P ≤ .01) are significantly differ-
Defective antioxidative

Hereditary xerocytosis

ent between healthy controls and PKD patients. Patients with HX

deficiency (PKD)
Sickle cell anemia

showed significant differences in the percentages polychromatic

Healthy controls

Pyruvate kinase
RBC membrane

stress defense
Not genotyped

β-­thalassemia

cells (P ≤ .001), macrocytes (P ≤ .001), and target cells (P ≤ .001) in

disorders
SLC4A1

relation to healthy controls. Distinct morphological cell types ana-

SPTA1
ANK1

(HX)

lyzed in patients with a defective antioxidative stress defense (such

as G6PD-­deficiency) were not significantly different in relation to
 |
4       HUISJES et al.

healthy controls, except for polychromatic cells (P ≤ .05). However,
the latter is likely due to the increase in reticulocytes in the two
patients in this group, who were recovering from a recent hemolytic
crisis (reticulocyte counts of 16.3% and 20.7%). An overview of the
distinct morphological cell types and distributions of healthy con-
trols and patients are depicted in Figure 1 and Table 2 (Howell-­Jolly
bodies are depicted in Fig. S1).
Subgroup analysis of the patients with RBC membrane disorders
revealed significant differences in distinct morphological cell types
(Figure 2). First of all, in patients with ANK1 mutations, microcytes were
significantly higher than in patients with SLC4A1 (P ≤ .05) and SPTA1
(P ≤ .001) defects. Interestingly, patients with mutations in STPA1 were
significant different in relation to healthy controls for the cell types
poikilocytes (P ≤ .05), schistocytes (P ≤ .001), and ovalocytes (P ≤ .01).

F I G U R E   1   Plots with all distinct morphological cell types measured and classified using digital microscopy (DM). Patients and healthy
controls are plotted, and patients are grouped in RBC membrane disorder, sickle cell anemia, β-­thalassemia, pyruvate kinase deficiency (PKD),
defective antioxidative stress defense, and hereditary xerocytosis (HX). * indicate statistically different groups when compared with healthy
controls (one-­way ANOVA, P ≤ .05) [Colour figure can be viewed at wileyonlinelibrary.com]
 HUISJES et al.

T A B L E   2   Percentage (SD) of different distinct morphological cell types analyzed and classified by digital microscopy (DM) for the healthy controls and the patient groups (RBC membrane
disorder, sickle cell anemia, β-­thalassemia, PKD, pyruvate kinase deficiency, defective antioxidative stress defense, and HX, hereditary xerocytosis)

RBC membrane Sickle cell Pyruvate kinase Defective antioxidative stress Hereditary
% (SD) Healthy controls disorder anemia β-­thalassemia deficiency (PKD) defense xerocytosis (HX)

Polychromatic cells 0.3% (0.3) 1.7% (1.4) 3.5% (1.6)* 2.5% (3.1) 6.6% (4.6)*** 3.7% (6.6)* 4.4% (1.7)***
Hypochromic cells 0.0% (0.0) 0.1% (0.2) 0.5% (0.7) 2.5% (3.7)*** 0.1% (0.2) 0.1% (0.1) 0.0% (0.0)
Microcytes 1.9% (1.5) 23.9% (15.9)*** 5.5% (6.3) 10.7% (8.2) 1.1% (0.6) 2.3% (1.9) 0.7% (0.4)
Macrocytes 3.6% (4.2) 1.7% (4.0) 27.4% (22.3)*** 12.7% (17.5) 25.1% (21.9)*** 12.7% (21.9) 26.0% (16.2)***
Poikilocytes 8.5% (4.0) 19.6% (17.6)** 29.6% (11.3)*** 15.7% (8.8) 15.5% (8.9) 10.3% (4.0) 16.1% (9.0)
Target cells 0.6% (0.7) 0.2% (0.3) 12.5% (6.9)*** 5.7% (4.2)*** 2.9% (2.9) 2.2% (3.2) 8.7% (7.9)***
Schistocytes 0.2% (0.2) 1.9% (3.0)** 2.4% (2.8)* 1.5% (2.0) 0.3% (0.2) 0.3% (0.3) 0.3% (0.2)
Helmet cells 0.1% (0.1) 0.1% (0.1) 0.3% (0.2) 0.5% (0.4)*** 0.2% (0.1) 0.2% (0.2) 0.3% (0.3)
Spherocytes 0.7% (0.9) 2.7% (3.7)** 0.4% (0.5) 1.6% (2.4) 0.2% (0.3) 0.3% (0.5) 0.4% (0.4)
Elliptocytes 1.4% (1.4) 5.7% (14.1) 1.9% (1.8) 1.1% (1.2) 1.3% (1.6) 0.5% (0.5) 0.8% (1.2)
Ovalocytes 0.0% (0.2) 0.2% (0.4) 0.8% (0.7)** 0.0% (0.0) 0.6% (1.0)** 0.1% (0.2) 0.2% (0.3)
Teardrop cells 0.6% (0.6) 0.7% (0.4) 1.3% (1.1)* 1.3% (0.6)* 0.9% (1.1) 0.5% (0.5) 0.4% (0.3)
Stomatocytes 2.8% (1.8) 2.3% (2.0) 5.9% (3.2)** 1.5% (1.3) 3.2% (2.4) 5.0% (3.3) 3.8% (1.7)
Acanthocytes 0.6% (0.4) 2.5% (3.2)** 1.1% (1.1) 2.3% (2.9) 1.7% (2.0) 0.8% (1.3) 0.7% (0.8)
Echinocytes 1.3% (1.2) 3.1% (4.1) 0.8% (0.9) 0.1% (0.1) 4.1% (4.6) 0.4% (0.3) 0.4% (0.6)
Sickle cells 0.0% (0.0) 0.2% (0.4) 2.3% (3.6)** 0.0% (0.1) 0.0% (0.0) 0.0% (0.0) 0.0% (0.0)
Howell-­Jolly bodies 0.1% (0.1) 0.2% (0.2) 0.2% (0.3) 0.2% (0.6) 0.3% (0.2) 0.1% (0.2) 0.1% (0.1)

Significant different cell types in patients compared to healthy controls are depicted with *(P ≤ .05), **(P ≤ .01) or ***(P ≤ .001) (one-­way ANOVA).
|
      5
 |
6       HUISJES et al.

F I G U R E   2   Percentage distinct morphological cell types in RBC membrane disorders classified by underlying molecular defects (SLC4A1,
ANK1, SPTA1, or genotype unknown). Observed significant differences in morphology between subgroups are noted with horizontal bars
and significant differences compared with healthy controls are indicated with * (one-­way ANOVA, P ≤ .05) [Colour figure can be viewed at
wileyonlinelibrary.com]

Patients with mutations in SLC4A1 or ANK1 showed no significant differ-

ences in these cell types. Furthermore, elliptocytes were only observed in
patients with SPTA1 defects, and not in patients with SLC4A1 or ANK1 de-
fects (Figure 2). The absence of elliptocytes is in line with the fact that het-
erozygosity for ANK1 and SLC4A1 mutations is associated with autosomal
dominant forms of hereditary spherocytosis whereas heterozygosity for
SPTA1 mutations causes autosomal dominant hereditary elliptocytosis.
3.3 | Defining optimal cut-­off values for cell types
Cut-­off values for significant different (one-­way ANOVA, P ≤ .05) cell
types to distinguish between patients and healthy controls were de-
termined when AUCROC were higher than 0.8 (Table 3). ROC curves
with excellent sensitivity and specificity (defined as AUCROC ≥ 0.9) are
depicted in Figure 3.
HUISJES et al. |
      7

T A B L E   3   Area under the curve (AUC)

Optimal cut-­off Specificity
from ROC curves (AUCROC) with optimal
AUCROC value (%) Sensitivity (%) (%)
cut-­off values (%) to discriminate healthy
controls from the different patient groups RBC membrane disorders
(RBC membrane disorder, sickle cell Microcytes 0.97 5.7 89.5 100
anemia, β-­thalassemia, PKD, pyruvate
Schistocytes 0.87 0.4 76.3 84.4
kinase deficiency, defective antioxidative
stress defense, and HX, hereditary Sickle cell anemia
xerocytosis) Polychromatic cells 0.99 1.1 100 96.9
Poikilocytes 0.99 15.9 90 96.9
Target cells 0.97 6.7 80 100
Macrocytes 0.97 7.6 100 87.5
Schistocytes 0.93 0.4 90 84.4
Ovalocytes 0.91 0.2 80 100
Sickle cells 0.90 0.2 80 100
Stomatocytes 0.86 4.5 70 87.5
β-­thalassemia
Helmet cells 0.96 0.3 83.3 96.9
Hypochromic cells 0.91 0.1 83.3 96.9
Target cells 0.90 1.7 83.3 93.8
Teardrop cells 0.80 0.6 83.3 62.5
Pyruvate kinase deficiency (PKD)
Polychromatic cells 0.98 2.1 90 100
Macrocytes 0.91 3.9 100 71.9
Defective antioxidative stress defense
Polychromatic cells 0.88 0.6 66.7 93.8
Hereditary xerocytosis (HX)
Polychromatic cells 1.00 2.1 100 100
Macrocytes 0.97 10.0 90.9 96.9
Target cells 0.90 4.4 81.8 100

Cut-­off values are only calculated when distinct morphological cell types were significantly different in
patients when compared with healthy controls (one-­way ANOVA, P ≤ .05) and, AUCROC were above
0.8.
Cut-­off values are calculated using Youden’s Index. Sensitivity and specificity were calculated using the
optimal cut-­off values.

4 | DISCUSSION
We here describe for the first the time the usage of digital micros-
copy (DM) in the morphology analysis of red blood cells from patients
with hereditary hemolytic anemia. We only used preclassification data
by the CellaVision Digital Microscope DM96 and did not carry out
postclassification. Using a unique set of patients, we found that DM
analysis can detect quantitative differences in distinct morphological
cell types between patients and healthy controls with high sensitivity
and specificity.
In particular, microcytes were found to be a very sensitive and
specific cell type to discriminate patients with RBC membrane dis-
orders from healthy controls (AUCROC = 0.97) (Figure 3). Using the
calculated cut-­off value of 5.7% for microcytes (Table 3), 34 of 38 pa-
tients (=89.5%) were correctly categorized as patients with RBC mem-
brane disorders. At the same time, from the 32 healthy controls, none
were falsely categorized using this cut-­off for microcytes. Nine of 12
patients (75%) with β-­thalassemia also showed more than 5.7% micro-
cytes, but these patients can be easily distinguished from patients with
RBC membrane disorders using routine hemocytometry parameters
(such as MCV in combination with total erythrocytes).17 Moreover,
the presence of target cells and helmet cells measured by DM can ex-
clude RBC membrane disorders and support a preliminary diagnosis
of β-­thalassemia, as target cells and helmet cells are only present in
patients with β-­thalassemia (Figure 1 and Table 2).
The number of spherocytes was significantly different in patients
with RBC membrane disorders when compared to healthy controls
(Figure 1 and Table 2). However, AUCROC for spherocytes in RBC
membrane disorders was ≤0.8 indicating that the cell type spherocyte
is less suitable to discriminate patients with RBC membrane disorders
from healthy controls. The cell type spherocytes lack sensitivity, as not
all peripheral blood smears from patients with RBC membrane dis-
orders have detectable amounts of spherocytes as measured by DM
(Figure 1). The lack of sensitivity of spherocytes as measured by DM
 |
8       HUISJES et al.

F I G U R E   3   ROC curves with excellent sensitivity and specificity (AUCROC ≥ 0.9). Depicted distinct morphological cell types can be used to
discriminate between the patient groups and healthy controls. Microcytes are an excellent distinct morphological cell type to distinguish RBC
membrane disorders from healthy controls (AUCROC = 0.97). Polychromatic cells (AUCROC = 0.99), poikilocytes (AUCROC = 0.99), target cells
(AUCROC = 0.97), macrocytes (AUCROC = 0.97), schistocytes (AUCROC = 0.93), ovalocytes (AUCROC = 0.91), and sickle cells (AUCROC = 0.90) are
suitable to distinguish patients with sickle cell anemia from healthy controls. Helmet cells (AUCROC = 0.96), hypochromic cells (AUCROC = 0.91),
and target cells (AUCROC = 0.90) can be used to distinguish patients with β-­thalassemia from healthy controls. Sensitive and specific distinct
morphological cell types to distinguish pyruvate kinase deficiency (PKD) patients from healthy controls are polychromatic cells (AUCROC = 0.98)
and macrocytes (AUCROC = 0.91). Polychromatic cells (AUCROC = 1.00), macrocytes (AUCROC = 0.97), and target cells (AUCROC = 0.90) are very
sensitive and specific in hereditary xerocytosis (HX) [Colour figure can be viewed at wileyonlinelibrary.com]

can be explained by either the fact that spherocytes are absent in pe-
ripheral blood smears or that the classification criteria for spherocytes
by DM requires adjustment. Manual examination of blood smears
from patients with RBC membrane disorders suggests that the num-
ber of spherocytes in these patients is underestimated by DM (Table
S1). Based on the lack of sensitivity of the cell type spherocytes in RBC
membrane disorders, we, therefore, recommend using the cell type
microcytes as a parameter to screen for RBC membrane disorders.
Elliptocytes were only present in patients with defects in STPA1
and were not observed in patients with a RBC membrane disorder due
to mutations in either ANK1 or SLC4A1 (Figure 2). This is in accordance
with our current understanding of the pathophysiology of hereditary
elliptocytosis (HE), which shows that most HE causing mutations are
found in genes coding for α-­spectrin (SPTA1), β-­spectrin (SPTB), pro-
tein 4.1 (ERB41), and glycophorin C (GPC).18 In terms of efficiency and
costs, we postulate that if microcytes exceed the limit of 5.7% and
elliptocytes are above background in peripheral blood smears genetic
testing could initially focus on genes SPTA1 and EPB41. A substan-
tial number of distinct morphological cell types were significantly
elevated in patients with sickle cell anemia (ie, polychromatic cells,
macrocytes, poikilocytes, target cells, schistocytes, ovalocytes, tear-
drop cells, stomatocytes, and sickle cells, Figure 1 and Table 2). Target
cells and macrocytes were previously also reported by others to be
elevated in sickle cell patients.19 Together with polychromatic cells,
poikilocytes, schistocytes, ovalocytes, and sickle cells, these cell types
were very sensitive and specific cell to discriminate patient with sickle
cell anemia from healthy controls (Table 3 and Figure 3). Although
sickle cells were significantly elevated in sickle cell anemia patients,
the number of sickle cells was relatively low in stable HbSS patients
(mean 1.3% (SD = 1.7%, range 0%-­5.1%)), whereas sickle cells were
elevated (11.2%) in the HbSS patient who was in an hemolytic crisis at
the ­moment of blood collection.
Our results show that the percentage polychromatic cells as de-
termined by DM is significantly different in patients with sickle cell
anemia, pyruvate kinase deficiency (PKD), hereditary xerocytosis
(HX), and patients with defective antioxidative stress defense in rela-
tion to healthy controls (Figure 1). Polychromatic cells are larger than
normal cells and show a pinkish blue-­gray color in a peripheral blood
smear. Their presence is known to correlate with reticulocyte num-
ber.9 We confirmed this by correlating the percentage of reticulocytes
to the percentage polychromatic cells as classified by DM (Fig. S2).
Interestingly, a similar increase in the number of polychromatic cells
was not observed in the group of patients with RBC membrane dis-
orders, despite the fact that these patients also showed a significant
degree of reticulocytosis (Table 1). This could suggest that the poly-
chromatic RBCs in RBC membrane disorders might be morphologically
HUISJES et al.
F I G U R E   4   Summary of proposed workflow for the usage of digital
microscopy as a screening tool in the diagnostic process of hereditary
hemolytic anemia after excluding immune-­mediated hemolysis. The
relatively low costs per sample make digital microscopy suitable
to increase the efficiency of advanced diagnostic tests (eg, Hb
characterization by HPLC, EMA-­binding test, osmotic gradient
ektacytometry, osmotic fragility test, and red blood cell enzyme
activity measurements). After performing these advanced diagnostic
tests, the diseases such as sickle cell anemia, α-­thalassemia,
β-­thalassemia, hereditary spherocytosis (HS), hereditary elliptocytosis
(HE), hereditary pyropoikilocytosis (HPP), hereditary xerocytosis,
pyruvate kinase (PK) deficiency, and glucose-­6-­phosphate
dehydrogenase (G6PD) deficiency can be confirmed
different in relation to polychromatic cells in PKD or HX, causing them
to escape detection by DM. Another possibility may be, that matu-
ration of RBCs and, consequently, degradation of ribosomal RNA is
more efficient in RBC membrane disorders in relation to PKD20 and
perhaps also to HX. This could lead to less efficient staining of poly-
chromatic RBCs in RBC membrane disorders using the May-­Grunwald
Giemsa staining. Furthermore, hemocytometry analysis and DM are
obviously two different techniques both using different criteria to des-
ignate RBCs as reticulocytes and hyperchromic cells, possibly leading
to minor differences in outcome. In addition, the cell type echinocytes,
defined as RBCs with many small protrusions, are often observed in
PKD but are considered to be a nonspecific marker for PKD.21,22 In
our study, only 3 of 10 PKD patients (=30%) had increased percent-
ages of echinocytes when compared to healthy controls (Figure 1 and
Table 2).
No significant differences were found in hemocytometry param-
eters (Table 1) and distinct morphological cell types in patients with
defective antioxidative stress defense (such as G6PD-­deficiency) in
|
      9
relation to healthy controls, except for polychromatic cells (Figure 1
and Table 2). This, however, is likely correlated with the increase in re-
ticulocytes in the two patients who were recovering from a recent he-
molytic crisis. Other distinct morphological cell types were unaffected
in patients with defective antioxidative stress defense, indicating that
RBC morphology is essentially normal in this group of patients.23
The specificity of the DM could be affected by conditions, such as
autoimmune hemolytic anemia, myelodysplastic syndrome (MDS), iron
deficiency anemia, and megaloblastic anemia. These acquired condi-
tions may also manifest with macrocytosis, microcytosis, spherocyto-
sis, and polychromasia. DM should therefore be used in addition to
well-­established and well-­available diagnostic tests such as the Direct
Antiglobulin Test (DAT), hemocytometry measurements (eg, reticu-
locytes, MCV), and biochemical analysis (eg, haptoglobin, LDH, iron,
ferritin, and vitamin B12).
In conclusion, we used for the first time DM as a screening tool in
the analysis of RBC morphology in hereditary hemolytic anemia. We
found a number of distinct morphological cell types that are highly
sensitive and specific in distinguishing patients from healthy controls
(Figure 3 and Table 3). We therefore believe that DM holds a promise
as a new diagnostic tool in the diagnosis of hemolytic anemia and pro-
pose its incorporation into various stages of the diagnostic workflow
of hemolytic anemia (Figure 4). At an early stage in the differential di-
agnosis of patients with Coombs negative (or Direct Antiglobulin Test
(DAT) negative) hemolytic anemia, the DM will give direction for fur-
ther diagnostic testing, whereas its use at later stages can provide use-
ful directions for further diagnostic testing in the differential diagnosis
of hereditary hemolytic anemia.
AC KNOW L ED G EM ENTS
The research leading to these results has received funding from the
European Seventh Framework Program under grant agreement num-
ber 602121 (CoMMiTMenT).
AU T HO R CO NT R I B U T I O N
RH, WvS, MDL, RvW, and JR designed the research. RH performed
the research. RH analyzed the data. RH, WvS, MDL, RvW, and JR
wrote the paper.
O RC I D
R. Huisjes  http://orcid.org/0000-0002-1069-3272
REFERENCES
1. Szuberski J, Oliveira JL, Hoyer JD. A comprehensive analysis of hemo-
globin variants by high-­performance liquid chromatography (HPLC).
Int J Lab Hematol. 2012;34(6):594‐604.
2. King M-J, Telfer P, MacKinnon H, et al. Using the eosin-­5-­maleimide
binding test in the differential diagnosis of hereditary spherocy-
tosis and hereditary pyropoikilocytosis. Cytometry B Clin Cytom.
2008;74(4):244‐250.
 |
10       HUISJES et al.
3. Clark MR, Mohandas N, Shohet SB. Osmotic gradient ektacytometry:
comprehensive characterization of red cell volume and surface main-
tenance. Blood. 1983;61(5):899‐910.
4. Parpart AK, Lorenz PB, Parpart ER, Gregg JR, Chase AM. The
osmotic resistance (fragility) of human red cells. J Clin Invest.
1947;26(4):636‐640.
5. Giordano PC. Strategies for basic laboratory diagnostics of the hemo-
globinopathies in multi-­ethnic societies: interpretation of results and
pitfalls. Int J Lab Hematol. 2013;35(5):465‐479.
6. Beutler E, Blume KG, Kaplan JC, Löhr GW, Ramot B, Valentine WN.
International Committee for Standardization in Haematology: rec-
ommended methods for red-­cell enzyme analysis. Br J Haematol.
1977;35(2):331‐340.
7. Lazarova E, Gulbis B, van Oirschot B, van Wijk R. Next-­generation os-
motic gradient ektacytometry for the diagnosis of hereditary sphero-
cytosis: interlaboratory method validation and experience. Clin Chem
Lab Med. 2017;55(3):394‐402.
8. Constantino BT. Reporting and grading of abnormal red blood cell
morphology. Int J Lab Hematol. 2015;37(1):1‐7.
9. Palmer L, Briggs C, McFadden S, et al. ICSH recommendations for the
standardization of nomenclature and grading of peripheral blood cell
morphological features. Int J Lab Hematol. 2015;37(3):287‐303.
10. 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.
11. Kratz A, Bengtsson H-I, Casey JE, et  al. Performance evaluation of
the CellaVision DM96 system: WBC differentials by automated digital
image analysis supported by an artificial neural network. Am J Clin
Pathol. 2005;124(5):770‐781.
12. Riedl JA, Stouten K, Ceelie H, Boonstra J, Levin M-D, van Gelder
W. Interlaboratory Reproducibility of Blood Morphology Using the
Digital Microscope. J Lab Autom. 2015;20(6):670‐675.
13. Egelé A, Stouten K, van der Heul-Nieuwenhuijsen L, et al. Classification
of several morphological red blood cell abnormalities by DM96 digital
imaging. Int J Lab Hematol. 2016;38(5):e98‐e101.
14. Egelé A, van Gelder W, Riedl J. Automated detection and classification
of teardrop cells by a novel RBC module using digital imaging/micros-
copy. Int J Lab Hematol. 2015;37(6):e153‐e156.
15. Hervent A-S, Godefroid M, Cauwelier B, Billiet J, Emmerechts J.
Evaluation of schistocyte analysis by a novel automated digital cell
morphology application. Int J Lab Hematol. 2015;37(5):588‐596.
16. Criel M, Godefroid M, Deckers B, Devos H, Cauwelier B,
Emmerechts J. Evaluation of the Red Blood Cell Advanced Software
Application on the CellaVision DM96. Int J Lab Hematol. 2016;38(4):
366‐374.
17. Hoffmann JJML, Urrechaga E, Aguirre U. Discriminant indices for
distinguishing thalassemia and iron deficiency in patients with mi-
crocytic anemia: a meta-­analysis. Clin Chem Lab Med. 2015;53(12):
1883‐1894.
18. Gallagher PG. Hereditary Elliptocytosis: Spectrin and Protein 4.1R.
Semin Hematol. 2004;41(2):142‐164.
19. Westerman MP, Bacus JW. Red blood cell morphology in sickle cell
anemia as determined by image processing analysis: the relationship
to painful crises. Am J Clin Pathol. 1983;79(6):667‐672.
20. Aizawa S, Kohdera U, Hiramoto M, et al. Ineffective erythropoiesis in
the spleen of a patient with pyruvate kinase deficiency. Am J Hematol.
2003;74(1):68‐72.
21. Zanella A, Fermo E, Bianchi P, Valentini G. Red cell pyruvate ki-
nase deficiency: molecular and clinical aspects.ac. Br J Haematol.
2005;130(1):11‐25.
22. Koralkova P, van Solinge WW, van Wijk R. Rare hereditary red blood
cell enzymopathies associated with hemolytic anemia -­ pathophys-
iology, clinical aspects, and laboratory diagnosis. Int J Lab Hematol.
2014;36(3):388‐397.
23. Danon D, Sheba C, Ramot B. The Morphology of Glucose 6 Phosphate
Dehydrogenase Deficient Erythrocytes: Electron-­microscopic Studies.
Blood. 1961;17(2):223‐234.
S U P P O RT I NG I NFO R M AT I O N
Additional Supporting Information may be found online in the
­supporting information tab for this article. 
How to cite this article: Huisjes R, van Solinge WW, Levin MD,
van Wijk R, Riedl JA. Digital microscopy as a screening tool for
the diagnosis of hereditary hemolytic anemia. Int J Lab Hem.
2017;00:1–10. https://doi.org/10.1111/ijlh.12758