S. Neppalli et al.
Clinical Imaging 80 (2021) 123–130
MRI.3
maximum of 125 ml, at 3 ml/s followed by 30 ml of normal saline
by hand injection.
During CEM a patient is given an IV injection of iodinated contrast,
and two minutes later, standard two view mammographic projections of
each breast are taken using dual-energy X-ray exposures, one just above
and the other below the K-edge of iodine. A dual-energy subtracted
(DES) image is produced by applying a weighted subtraction algorithm
to the low-energy (LE) and high-energy (HE) images, canceling the
normal tissue appearance and increasing iodine contrast.4 It is important
to note that there are significant vendor-related differences in X-ray
production and filtration, the types of detectors and the algorithms used
to produce DES images. For example, these differences are reflected in
substantially different contrast-to-noise ratio values measured between
the systems in equivalent technical evaluations.5,6
(2) Image acquisition: unaffected (left/right breast not containing
lesion of interest) craniocaudal (CC), affected CC, unaffected
mediolateral oblique (MLO) and affected MLO, with first image
acquired 2 min post injection. All imaging was completed within
8 min.
(3) All studies were performed by the same group of nine MITs with
mammographic experience ranging from 3 to 20 years.
2.3. Image analysis
The LE and DES images for all four views of each study were retro-
spectively reviewed by a subspecialist consultant breast radiologist and
one of two senior MITs, each with over five years of CEM experience,
using an Agfa PACS (Version 6.5.3.1509 2013) system and dual 5-mega-
pixel Planar Dome ES monitors. DES images were analyzed for each
artifact after excluding a cause for the artifact on the LE image.
Sixteen artifacts (described below) were observed as being present or
absent on each DES image.
Artifacts on the DES image can be produced by one of several patient-
related or machine-related factors from combining images acquired with
different X-ray energy spectra, possibly impeding diagnostic interpre-
tation.7 Radiologists and technologists should be familiar with these
artifacts to help avoid image misinterpretation.
The two most widely used types of CEM equipment are produced by
Hologic and General Electric (GE) with our clinic being one of only a few
with extensive experience in using equipment from both vendors. This
article will review a range of artifacts commonly seen on DES images
and compare their incidence and severity between vendor systems.
(1) Ripple: faint alternating black and white lines (Fig. 1).4,9–11
(2) Halo: an arc-shaped whiter region paralleling the breast edge
(Fig. 1).7,9–11
2. Materials and methods
(3) Skin line enhancement: highlighting of the skin line contour on
the DES image (Fig. 2).7,9,10
Informed consent was not required for this Quality Improvement
Activity, GEKO ID 20627. The radiology information system was inter-
rogated to identify CEM studies performed between September 2013
and March 2017. De-identified patients were listed in a Microsoft Excel
(Version 2010) spreadsheet. All artifacts were noted as present or ab-
sent. Those with a variable appearance were subjectively graded for
severity using criteria listed in the appendices, developed by a sub-
specialist breast radiologist in conjunction with two medical imaging
technologists (MITs) with expertise in mammography.
(4) Air gap: a dark artifact caused by incomplete contact due to air
trapped in the skin folds.9,11,12 We recorded air gap artifacts in
the DES image when the artifact appearance was either enhanced
compared to the LE image, or when no air gap artifact was visible
on the LE image (Supp. Fig. 1).
(5) Contrast contamination: non-mass enhancement seen only on the
DES images (Supp. Fig. 2).9,10,12,13
(6) Elephant rind: a term coined by our group due to its peripheral
location and likeness to the pattern and thickness of an elephant’s
skin (Fig. 3).
2.1. Patients and equipment
(7) Cloudy fat: the appearance of contrast enhancement of fatty tis-
sue regions (Fig. 4).
The GE Senographe Essential (General Electric Company, Boston,
MA) was the only CEM-equipped mammography machine in our clinic
between 2013 and 2016 (software versions 55.20 2013-2015, 56.12
2015-2016, and 56.21.3 2016-present), followed by installation of a
Hologic Selenia Dimensions (Hologic Inc., Marlborough, MA) unit
(software version 1.8.3 2016-end of QA activity). Some patients had
(8) Suppression of highly attenuating objects: It remains to be seen
whether suppression is artifactual or intended, our observations
have been recorded without statistical analysis, with some CEM
images having variable suppression (Fig. 1).
(9) Negative rim surrounding a lesion (Supp. Fig. 3).
(10) Black line on the pectoral muscle (Supp. Fig. 4).
(11) White line on the pectoral muscle (Supp. Fig. 4).
(12) Pectoral highlighting (Supp. Fig. 4).
CEM performed as part of clinical trials: CESM
ACTRN12613000684729 (GE equipment only), CESM
V
T
Study
Study
ACTRN12616000533493 (Hologic equipment only) and CESM D Study
ACTRN12616001696482 (GE Equipment only).8 From 2016 onwards,
patients not involved in a clinical trial had their study performed on
either machine.
(13) Bright pectorals (Supp. Fig. 4).
(14) Chest wall highlighting (Supp. Fig. 5).
(15) Black line on chest wall (Supp. Fig. 5).
(16) Corrugations: parallel, transverse grooves seen in the superior
part of breast on the MLO projection (Fig. 5).
A total of 200 consecutive CEM studies were reviewed, 100 had been
performed using the GE Essential and the other 100 were obtained with
the Hologic Dimensions. This ensured a reasonable sample size without
preferencing one vendor over the other. Once 100 consecutive cases
were reached from one vendor, the remaining cases were excluded from
that vendor until 100 cases from the other vendor was reached.
The commonest clinical indications for CEM included local cancer
staging, problem solving, and moderate risk screening for those unable
to have MRI.
Artifacts with a variable severity were graded from 0 to 3 in 0.5 in-
crements by consensus between the reviewing radiologist and MIT,
using criteria listed in the appendices. The rubric was developed based
on the minimum/mild/moderate/marked descriptors used for scoring
background parenchymal enhancement (BPE) on breast MRI, with de-
scriptors written by the radiologist in conjunction with the MITs.
2.4. Statistical analysis
2.2. CEM technique
The relative risk was calculated to assess whether there was a sig-
nificant difference in the incidence of each artifact between both ven-
dors. When calculating the relative risk of artifacts that do not occur in
one vendor, a fixed correction of 0.5 was added to both frequencies to
avoid a computational error.14,15 A Wilcoxon rank-sum test was used to
A standardized CEM technique was used for all 200 studies:
(1) IV contrast: Omnipaque 350 contrast media (GE Healthcare,
Chicago, Il) 350 mg I/ml, 1.5 ml per kg of body weight to a
124