Key words phantom-less - routine CT - contrast agents - calculation formula - bone mineral density
Introduction
Osteoporosis is characterized by a lowered bone mass and trabecular thinning, which
leads to an increased risk of fracture, higher mortality, and increased healthcare
costs. In addition, patients with osteoporosis suffer from decreased independence
and quality of life [1 ]
[2 ]
[3 ]
[4 ].
Osteoporosis is diagnosed by the assessment of bone mineral density (BMD). Commonly
used BMD measurements are dual-energy X-ray absorptiometry (DXA) and quantitative
computed tomography (QCT). A major problem with DXA is that in elderly populations
lumbar spine spondylosis causes false elevation of BMD when measured in this anatomical
site. [2 ]
[5 ]
[6 ]
[7 ]
[8 ].
Recent studies have shown that routinely performed multidetector computed tomography
(MDCT) scans can also be used for BMD measurements [9 ]
[10 ]. As MDCT is one of the most important radiological examination methods, especially
in tumor patients, and oncology patients also frequently suffer from osteoporosis
triggered by chemotherapy or hormonal therapy, BMD measurements obtained on routine
MDCT scans would be a promising method for the diagnosis of osteoporosis [11 ]
[12 ]. A very recent innovation in this field was the development of phantom-less BMD
measurement systems. The major advantage of this phantom-less BMD measurement system
is that the patient can be used as his/her own reference, so that no bone equivalent
phantom is necessary, and, consequently, BMD can be measured retrospectively on MDCT
scans initially performed for another reason. This saves the patient from additional
radiation exposure. Furthermore, beam-hardening and scatter effects, which might be
induced by an external phantom, do not play a role in phantom-less BMD measurement
methods [13 ]
[14 ].
However, it must be stated that there are potential problems with phantom-less BMD
methods. For example, heterogeneous or varying density values of muscle and fat, which
are used as reference standards, due to differences in hydration status can influence
the measurements [14 ]. In accordance with that, previous studies have already shown that intravenous contrast
media administration also leads to higher BMD values measured on routine MDCT scans
[15 ]
[16 ]
[17 ]
[18 ]. As oncologic staging investigations are mainly performed with the use of an intravenous
contrast agent, this could be a major drawback for the diagnosis of osteoporosis on
routine MDCT scans.
Therefore, the aim of our study was to evaluate the differences in phantom-less BMD
measurements on routine MDCT scans in the unenhanced, arterial, and venous contrast
phases, using the Philips BMD measurement tool (Philips Healthcare, Best, NL). Furthermore,
an algorithm for calculating a reliable BMD value from these contrast-enhanced MDCT
scans should be developed.
Materials and Methods
Patients
In this prospective study, we included 112 female, postmenopausal patients from the
age of 40 to 77 years (mean age: 57.31 years; SD 9.61), who underwent a routinely
performed contrast-enhanced MDCT scan for other indications in the period from November
2013 to June 2014. The inclusion criteria were a contrast-enhanced MDCT scan, consisting
of at least an unenhanced, an arterial and a venous phase, and a scan region including
vertebrae T12 to L4. A flowchart of excluded patients is depicted in [Fig. 1 ]. Indications for the MDCT scans were, for example, nausea, abdominal pain, portal
venous thrombosis and intestinal obstruction as well as follow-up examinations for
ovarian, colon, gastric and lung cancer. In total, 61 patients were oncologic patients,
59 of which were undergoing chemotherapy. 6 patients were smokers. Only vertebrae
T12 to L4 were included in the BMD analyses. Patients with metastases or hematologic
or metabolic bone disorders besides osteoporosis were excluded. Furthermore, 16 vertebrae
with benign osteolytic or osteoblastic lesions, for example, hemangiomas, 17 fractured
vertebrae, and 4 vertebrae with pronounced degenerative changes were excluded. In
one patient we had to exclude one vertebra because of a vertebroplasty. In total,
at least two vertebrae in each patient had to be evaluable.
Fig. 1 Flowchart of inclusion and exclusion of study participants.Abb. 1 Flussdiagramm der Ein- und Ausschlusskriterien.
All patients gave written, informed consent to scientific evaluation of their data.
The local ethics committee approved this prospective study.
Image acquisition
The MDCT scans were performed on a 256-row CT unit (Philips iCT 256, Philips Healthcare,
Best, NL). The scanning protocol was adapted to the clinical indications. The images
were acquired with a tube voltage of 120 kV, an average tube current of 200 mAs, and
a collimation of 128 × 0.625 mm. Examinations were performed using contiguous acquisition
(no overlap). Axial slices were reconstructed using a soft-tissue kernel and a slice
thickness of 5 mm. Zips or metal clips were avoided in the field of view (FOV). For
the contrast-enhanced series, we chose a standardized amount of contrast agent. Each
patient received 100 ml of intravenous contrast media (Omnipaque 300 mg/ml, GE Healthcare,
Little Chalfont, UK). For the injection, we used a Medrad injector with a flow rate
of 3.0 ml/second. The intravenous contrast media injection started with a delay of
35 seconds for the arterial phase and 70 seconds for the venous phase. Only MDCT scans
including unenhanced, arterial, and venous phases were included.
Image analyses
The BMD analyses were performed on a workstation, on which the required phantom-less
Philips bone mineral density application was installed previously.
Initially, the correct slice and height of the region of the vertebral body, which
should be measured at a safe distance from the cortical bone and tilted to the axis
of the vertebra, was adjusted in the axial, coronal, and sagittal planes. An oval
region of interest (ROI) was placed in the vertebral body on the axial plane, without
including cortical bone and basivertebral veins. Subsequently, a second ROI was placed
in the paravertebral muscle and a third ROI in the subcutaneous fat tissue ([Fig. 2 ], [3 ]). If the paravertebral muscle showed fatty atrophy of more than 50 %, the second
ROI was placed in the psoas muscle.
Fig. 2 Bone mineral density application measurement with three ROIs—one in the vertebra
(yellow), one in the paraspinal muscle (red), and one in the subcutaneous fat tissue
(blue).Abb. 2 Phantomlose Knochendichtemessungen mit drei ROIs: im Wirbelkörper (gelb), in der
paravertebralen Muskulatur (rot) und im subkutanen Fettgewebe (blau).
Fig. 3 Bone mineral density application measurement result, BMD of L1: 60 mg/cm³.Abb. 3 Ergebnis der phantomlosen Knochendichtemessung, Knochendichte des LWK 1: 60 mg/cm³.
The BMD was calculated according to an algorithm that is implemented in the phantom-less
BMD measurement tool [14 ]. All vertebrae were analyzed in each phase, including the unenhanced, arterial,
and venous phases.
The calculated BMD value for each evaluated vertebra and each phase as well as the
mean BMD values of all evaluated vertebrae for each individual patient in all phases
were documented.
The bone mineral density application also provided a graph, in which the patient’s
average BMD value was shown in relation to a European reference group ([Fig. 4 ]).
Fig. 4 Average BMD value in relation to a European reference group.Abb. 4 Durchschnittlicher Knochendichtewert in Relation zu einer Europäischen Referenzgruppe.
After a training session with a board-certified radiologist, a medical student in
the last academic year performed all BMD evaluations. The BMD application’s reproducibility
was evaluated using 40 randomly selected patients who were also evaluated by a resident
in the third year of training. In order to calculate the intrarater agreement, the
medical student evaluated 40 patients twice, blinded to patient-identifying data and
previously measured BMD values.
Statistical analyses
All statistical analyses were performed by a statistician, using IBM SPSS 22.0.
BMD was described using mean and standard deviation. In order to compare BMD obtained
on unenhanced scans and in the arterial and venous phases, repeated measures ANOVA
and post hoc Bonferroni corrected paired t-tests were used. By using the linear regression
analyses, two conversion formulas for the calculation of BMD values based on the contrast-enhanced
phases could be developed. The intra- and interobserver agreement was rated using
the intraclass correlation coefficient (ICC). A p-value of p < 0.05 was considered
to indicate significant results.
Results
Calculating the mean BMD of at least two vertebrae per patient, the mean BMD value
of all patients in the unenhanced phase was 79.76 mg/cm³ (SD 31.20). In the arterial
phase, the mean BMD value of the whole study population was 85.09 mg/cm³ (SD 31.61),
and, in the venous phase, the mean BMD value was calculated at 86.18 mg/cm³ (SD 31.30).
Repeated measures ANOVA and post hoc corrected paired t-tests showed that BMD values
measured in the unenhanced phase were significantly lower than the values acquired
in the venous and arterial phases (p < 0.001). However, there was no significant difference
found between BMD values calculated in the arterial phase and BMD values measured
in the venous phase (p = 0.228). Patients undergoing chemotherapy vs. patients without
chemotherapy did not demonstrate any significant difference in regard to BMD values
(p = 0.123). The same applies for smokers and non-smokers (p = 0.200).
[Fig. 5a, b ] visualize a positive correlation when comparing BMD values calculated in the unenhanced
MDCT scans versus BMD values measured in the arterial phase (a), and unenhanced measurements
versus BMD values in the venous phase (b), without showing outliers. The difference
between arterial and unenhanced BMD values, relative to the difference between venous
and unenhanced BMD values, is depicted in [Fig. 6 ].
Fig. 5 Scatter diagrams showing a positive correlation between unenhanced vs. arterial a and unenhanced vs. venous BMD values b .Abb. 5 Die Streudiagramme zeigen eine positive Korrelation zwischen nativ vs. arteriell
a und nativ vs. venös b gemessenen Knochendichtewerten.
Fig. 6 Box-and-Whisker plot indicating the difference between arterial-unenhanced and venous-unenhanced
BMD values.Abb. 6 Der Box-Whisker-Plot veranschaulicht die Differenz zwischen arteriell vs. nativ und
venös vs. nativ gemessenen Knochendichtewerten.
Finally, two conversion formulas, enabling calculation of the unenhanced, relatively
true BMD value from values measured in the arterial or venous phase, were defined
using linear regression:
The intrarater agreement of BMD measurements was calculated with an intraclass correlation
coefficient (ICC) of 0.984 and the interrater reliability was calculated with an ICC
of 0.991.
Discussion
Our study showed that phantom-less BMD measurements on contrast-enhanced MDCT scans
are possible, even though intravenous contrast agent elevates BMD values, which can
result in falsely high results. Taking this into account, it is possible to calculate
a converted BMD value using the formulas defined in this study.
In comparison to the suggested thresholds for osteoporosis (< 80 mg/cm³) and osteopenia
(> 80 to 120 mg/cm³) issued by the American College of Radiology, a remarkable observation
in our study was the generally low BMD values of our patients, which might be a population-related
finding, as former studies have shown lower BMD values in this ethnic population compared
to other ethnic populations [19 ]
[20 ]. Since we included 59 patients receiving chemotherapy and 6 smokers, which may have
an impact on BMD measurements, an additional statistical analysis was conducted: chemotherapy
or smoking did not significantly influence BMD values. In addition, a software-related
origin is possible, as Mueller et al. also found slightly lower BMD values using the
Philips BMD option compared to phantom-based QCT. However, in their study, the values
measured by the BMD software were generally only 0.9 mg/cm³ lower than the BMD values
calculated by phantom-based QCT, which is a negligibly low difference. Furthermore,
they demonstrated a slightly lower precision compared with phantom-based QCT, but,
nevertheless, a very good accuracy of the Philips bone mineral density application,
with some advantages compared to QCT using a phantom [14 ].
With regard to phantom-based QCT, a major disadvantage of the method is the need for
a phantom. Using the Philips bone mineral density option, no phantom is needed and
BMD measurements can be performed retrospectively in any patient who underwent a CT
scan for any reason, without the need for another investigation that might cause additional
radiation exposure. Furthermore, the phantom-less BMD measurement is a time- and cost-saving
method [14 ]
[21 ].
Previous studies have already investigated the possibility of BMD measurements on
routinely performed MDCT scans and the influence of intravenous contrast agent on
the measured BMD values [10 ]
[15 ]
[16 ]
[18 ]
[22 ]
[23 ]. These studies used different methods and showed somewhat divergent outcomes: Pompe
et al. measured attenuation values of L1 in different contrast agent phases and found
a significant difference between all phases [15 ]. In contrast, Pickhardt et al. compared Hounsfield Unit (HU) values of L1 on pre-contrast
CT scans with measurements on contrast-enhanced CT scans and did not find a significant
difference [10 ]
[22 ]. Some studies, such as the one by Bauer et al., used QCT for BMD evaluation. These
investigators described a 2 % increase in BMD values measured in the hip after intravenous
contrast agent administration versus a 31 % increase in BMD values measured in the
spine [18 ]. These results correspond very well with the results defined by Link et al., who
noted an increase of 30 % in BMD values measured in the spine after intravenous contrast
agent administration [16 ]. Baum et al. compared routine MDCT with a phantom to dedicated phantom-based QCT,
and also found an average increase in BMD values of 37.9 %, measured in the spine,
compared to QCT values [23 ]. A potential problem when using a phantom-less BMD measurement method, where the
patient serves as his/her own reference, might be the variable contrast enhancement
of bone, as well as muscle and fat tissue, which leads to increased HU values of all
measurements, and thus, may falsify the calculation algorithm.
Our study has some limitations. We did not correlate our results with the presence
of vertebral fractures, as outlined by Baum et al. [23 ]. Furthermore, the ROIs were placed manually, which gives rise to the risk of a lower
precision and a higher inter- and intraobserver variability. We minimized that risk
by providing both observers with an intensive training session before starting the
study, thus helping to achieve a very low inter- and intraobserver variability. In
contrast to other studies like those of Pompe et al. or Pickhardt et al., our technique
requires a specific software tool, which entails additional costs [10 ]
[15 ]. Another limitation of this study is that no additional DXA or QCT examinations
were available as a reference or for comparison.
Conclusion
In conclusion, routinely performed contrast-enhanced abdominal MDCT scans can be used
for BMD measurement using our method, but the administration of contrast agent should
be taken into account. The two formulas defined in this study enable the measurement
of BMD values on contrast-enhanced MDCT scans because the actual BMD value can be
calculated afterward. The Philips bone mineral density measurement tool used in our
study showed very good reliability and seems to be a promising phantom-less method
for retrospective BMD measurements on routine MDCT scans.
Clinical relevance of the study
Using the phantom-less Philips bone mineral density measurement tool tested in this
study, BMD measurements can be done retrospectively on any MDCT scan performed for
another reason.
Intravenous contrast media application increases BMD values measured in the arterial
as well as venous phases.
Applying the formulas defined in this study, a reliable BMD value can be calculated
from BMD values measured in the arterial or venous phase.
