SHORT COMMUNICATION
Utility of an open-source DICOM viewer software (OsiriX)to assess pulmonary fibrosis in systemic sclerosis: preliminaryresults
Alarico Ariani • Marina Carotti • Marwin Gutierrez •
Elisabetta Bichisecchi • Walter Grassi •
Gian Marco Giuseppetti • Fausto Salaffi
Received: 2 August 2012 / Accepted: 31 July 2013
� Springer-Verlag Berlin Heidelberg 2013
Abstract To investigate the utility of an open-source
Digital Imaging and Communication in Medicine viewer
software—OsiriX—to assess pulmonary fibrosis (PF) in
patients with systemic sclerosis (SSc). Chest high-resolu-
tion computed tomography (HRCT) examinations obtained
from 10 patients with diagnosis of SSc were analysed by
two radiologists adopting a standard semiquantitative
scoring for PF. Pulmonary involvement was evaluated in
three sections (superior, middle and inferior). For the
assessment of the extension of PF, the adopted semiquan-
titative HRCT score ranged from 0 to 3 (0 = absence of
PF; 1 = 1–20 % of lung section involvement;
2 = 21–40 % of lung section involvement; 3 = 41–100 %
of lung section involvement). Further, a quantitative
assessment (i.e. parameters of distribution of lung attenu-
ation such as kurtosis and mean lung attenuation) of PF
was independently performed on the same sections by a
rheumatologist, independently and blinded to radiologists’
scoring, using OsiriX. The results obtained were compared
with those of HRCT semiquantitative analysis. Intra-reader
reliability of HRCT findings and feasibility of OsiriX
quantitative segmentation was recorded. A significant
association between the median values of kurtosis by both
the quantitative OsiriX assessment and the HRCT semi-
quantitative analysis was found (p \ 0.0001). Moreover,
kurtosis correlated significantly with the mean lung
attenuation (Spearman’s rho = 0.885; p = 0.0001). An
excellent intra-reader reliability of HRCT findings among
both readers was obtained. A significant difference
between the mean time spent on the OsiriX quantitative
analysis (mean 1.85 ± SD 1.3 min) and the mean time
spent by the radiologist for the HRCT semiquantitative
assessment (mean 8.5 ± SD 4.5 min, p \ 0.00001) was
noted. The study provides the new working hypothesis that
OsiriX may be a useful and feasible tool to achieve a
quantitative evaluation of PF in SSc patients.
Keywords Pulmonary fibrosis � Systemic sclerosis �DICOM � HRCT � kurtosis � OsiriX
Introduction
Systemic sclerosis (SSc) is a systemic autoimmune disease
characterized by a micro- and macro-vascular damage [1,
2]. Pulmonary fibrosis (PF) is a frequent manifestation in
SSc. Its severity may vary considerably depending on the
underlying disease, and frequently it can be the cause of
death of these patients [3–5].
Currently, chest high-resolution computed tomography
(HRCT) is considered the most accurate not invasive
imaging method for PF assessment. Both severity and
extent of PF are usually estimated by the adoption of
semiquantitative scoring methods [6, 7]. However, the
correct evaluation of the semiquantitative scores could
represent a problem for the inexperienced physicians since
there is a wide interobserver variability even among expert
radiologists [8].
Recently, special softwares providing an automatic lung
parenchyma identification and a quantitative assessment of
PF have been developed [9, 10]. In particular, they allow to
A. Ariani � M. Gutierrez (&) � W. Grassi � F. Salaffi
Clinica Reumatologica, Dipartimento di Scienze Cliniche e
Molecolari, Universita Politecnica delle Marche, Ospedale ‘‘A.
Murri’’, Via dei Colli, 52 Jesi, 60035 Ancona, Italy
e-mail: [email protected]
M. Carotti � E. Bichisecchi � G. M. Giuseppetti
S. O. D. Radiologia Clinica, Dipartimento di Scienze
Radiologiche, Ospedali Riuniti, Ancona, Italy
123
Rheumatol Int
DOI 10.1007/s00296-013-2845-6
obtain a more detailed information in regard to the severity
of PF since they permit an accurate evaluation of lung
attenuation distribution parameters such as kurtosis, sym-
metry and mean lung attenuation (MLA) [11]. Unfortu-
nately, the lack of standardization and the relative high cost
of the licences limit their use in clinical practice.
The growing physicians’ demand about visualization
and management of medical digital images [i.e. the Digital
Imaging and Communication in Medicine (DICOM) files]
has urged to the development of open-source (and free of
charge) softwares which allow to view DICOMs and to
perform very complex imaging analysis (‘‘post-process-
ing’’). OsiriX is one of the most popular DICOM viewers
created for Mac OSX platform [12]. Some OsiriX tools,
such as the possibility to perform multimodality and mul-
tidimensional post-processing analysis and reworking on
3D data obtained from chest HRCT, open up a new
research area to attain PF assessment. The main aim of the
present study was to investigate the utility of OsiriX for the
rheumatologist to assess pulmonary fibrosis in SSc. The
study was also planned to compare the PF assessment
carried out by OsiriX with respect to a standard semi-
quantitative method based on lung findings performed by a
radiologist.
Patients and methods
Patients
Ten patients with diagnosis of SSc (1 male and 9 females)
were included in the present study. The diagnosis was
made according to the American College of Rheumatology
(ACR) classification criteria [13]. Mean ± SD age was
56 ± 7 years (range 45–61), and the mean ± SD disease
duration was 10 ± 2 years (range 7–11).
Inclusion criteria included diagnosis of SSc with docu-
mented PF, [ 18 years, chest HRCT performed no longer
than 6 months prior to the beginning of this study. Patients
who refer a history of pulmonary neoplasia or other causes
of interstitial fluid such as hearth failure, diastolic dys-
function, asthma or pulmonary oedema were excluded of
the study.
All patients were attending the outpatient and inpatient
clinics of the Rheumatology Department of the Universita
Politecnica delle Marche (Ancona, Italy).
Study design
Chest HRCT and OsiriX analysis were carried out at the
Radiology and Rheumatology Departments of the Univer-
sita Politecnica delle Marche, Ancona, Italy, respectively.
All HRCT examinations were reviewed by a radiologist
expert on the HRCT interstitial lung disease (MC), blinded
to the clinical data. A second experienced thoracic radiol-
ogist (EB), blinded with respect to the first one and HRCT
findings, scored the PF in order to assess the inter-reader
agreement. Prior to the study, the investigators reached a
consensus on the PF HRCT interpretation.
Afterwards, on the same pulmonary sections, a quanti-
tative assessment of PF was performed independently and
blinded to radiologists’ scoring, by a rheumatologist (AA),
using OsiriX. The results obtained were compared with
those of HRCT semiquantitative analysis.
In order to determine the feasibility, the time spent for
both semiquantitative HRCT analysis and quantitative
segmentation by OsiriX was recorded.
The study was conducted according to the Declaration
of Helsinki and local regulations. The institutional review
board approved the study and informed consent was
obtained from all patients.
Chest HRCT assessment
All HRCT examinations were performed by standard pro-
tocol using a CT 64 GE light Speed VCT power scanner
with a rotation tube scanning time of 0.65 s. Scans were
obtained at full inspiration from the apex to the lung base
with the patients in the supine position, at 120 kV and
300 mAs, with a slice thickness of 1.25 mm and slice
spacing of 7 mm. HRCT assessment do not include the use
of contrast media agents.
HRCT pulmonary involvement was evaluated in three
pulmonary sections for each patient (obtaining 30 pul-
monary segmentations) according to Sverzellati et al. [9]
[8]. The sections considered were superior (origin of the
large vessels), middle (carina of trachea) and inferior (right
inferior pulmonary vein).
The following elementary lesions were evaluated in
each section: ground glass opacity, interface irregularities
between peripheral pleura and lung parenchyma, septal
lines (due to interlobular septa and subpleural thickening),
honeycombing and thin-walled subpleural cysts.
For the assessment of the extent of PF, a semiquantita-
tive HRCT, score ranged from 0 to 3 (0 = absence of PF;
1 = 1–20 % of lung section involvement; 2 = 21–40 % of
lung section involvement; 3 = 41–100 % of lung section
involvement), was adopted.
OsiriX assessment
HRCT images were reconstructed and analysed by OsiriX,
a DICOM viewer software (OsiriX version 3.9; Apple
Computer) on a Mac Mini (2.4 GHz Intel Core 2 Duo
Rheumatol Int
123
Desktop Computer, 4 GB random-access memory; Apple
Computer, Cupertino, CA, USA) running Mac Operating
System X 10.5.
After inserting the CD-ROM or DVD containing HRCT
scan data in the drive, the DICOM data were automatically
extracted from the disc by OsiriX. The DICOM data were
stored in the OsiriX using the ‘‘Copy linked files to Data-
base folder’’ under ‘‘file’’ in the OsiriX dropdown menu.
The same three pulmonary sections evaluated previously
by the radiologist were assessed. For each section, a
semiautomatic lung parenchymal segmentation was per-
formed in order to obtain the corresponding histograms of
attenuation; then, descriptive parameters of the distribution
such as MLA, kurtosis and symmetry were calculated.
Figure 1 shows the representative sequences of the OsiriX
segmentation process.
Statistical analysis
Statistical analysis was performed using MedCalc (version
12.0 for Windows XP, Belgium).
Segmentation data did not follow a Gaussian distribution,
and therefore, we did not use parametric tests. Thus, the
kurtosis, symmetry and MLA values were represented as
median with the relative interquartile. Spearman’s rank order
test was used to evaluate the correlation between kurtosis and
MLA. Additionally, we determined the relationship between
kurtosis and MLA, and the different scores of HRTC, using
Kruskall–Wallis and Wilcoxon tests to assess the level of
significance of the different severity categories. The intra-
reader HRCT agreement between the two radiologists has
been calculated by weighted kappa statistic. A kappa value of
0–0.20 was considered poor, 0.21–0.40 fair, 0.41–0.60
moderate, 0.61–0.80 good and 0.81–1.00 excellent. Feasi-
bility of OsiriX was estimated by comparing the time
spent in quantitative analysis with respect to HRCT semi-
quantitative analysis by the independent samples t test. A
p value\0.05 was considered statistically significant.
Results
Figure 2 shows representative examples of semiquantita-
tive analysis with the relative distribution diagrams
(including kurtosis, symmetry and MLA) obtained using
OsiriX.
Fig. 1 Segmentation algorithm. The basic steps for the segmentation
of lung parenchyma of the section shown in a are as follows: selection
of the command to generate the Region of Interest (b), insertion of
parameters for the detection of lung parenchyma (c), applying the
command of ‘‘Brush ROI’’ to obtain a complete segmentation (d)
Rheumatol Int
123
On categorizing patients into different grades of sever-
ity, with respect to HRCT score, we found a significant
association between the median values of kurtosis of each
lung section assessed by OsiriX and HRCT semiquantita-
tive score (p \ 0.0001). In particular, kurtosis was signif-
icantly different in patients with low HRCT score (0 or 1)
compared with patients with higher scores (2 and 3)
(p \ 0.01) (Fig. 3a).
The kurtosis median value, measured by computer
analysis, was significantly higher in the apical lung sec-
tions, compared with the lower ones (p \ 0.05) (Fig. 3b).
Similarly, low score patients (0 or 1) and high score
patients (2 and 3) have significantly different
(p \ 0.01) MLA median values, measured over the three
lung sections (Fig. 3c).
A significant correlation between kurtosis and MLA was
found (rho = 0.885 p = 0.0001) (Fig. 3d).
The global kappa value for the intra-reader reliability of
HRCT findings reached by both radiologists was 0.92.
With respect to the feasibility, a significant difference
between the time spent by the rheumatologist for the
OsiriX quantitative analysis, including three pulmonary
sections, and the mean time spent by the radiologist for the
HRCT semiquantitative assessment was found (mean
1.85 ± SD 1.3 min vs mean 8.5 ± SD 4.5 min,
p \ 0.00001, respectively).
Discussion
To the best of our knowledge, this is the first study pro-
viding evidence in favour of the utility of this ‘‘novel’’
open-source DICOM viewer software (OsiriX) as an
adjunct method to assess PF in patients with SSc.
PF plays an unfavourable prognostic factor of life in
SSc patients. Chest HRCT allows an accurate assessment
of both severity and extension of PF; however, it requires
both adequate knowledge of semiquantitative methods and
wide experience to score PF. Moreover, the reported con-
troversial data about its reproducibility makes its system-
atic application difficult in clinical trials [14, 15]. In order
to overcome this barrier, sophisticated automatic algo-
rithms to identify accurately the PF areas (segmentation)
have been developed. These segmentations permit to obtain
histograms of distributions of the lung attenuation which
facilitate the identification and quantification of PF through
the analysis of the descriptive parameters (mode, mean
value, kurtosis and skewness) [16, 17]. Additionally, these
descriptive parameters demonstrated to be higher repro-
ducible than the PF semiquantitative assessment, usually
performed by the radiologist [18]. Although these soft-
wares appear to be a really promising tool for the assess-
ment of PF, their application in clinical practice still
remains limited, due to the lack of standardization and the
relatively high cost of licence [19].
In our study we used an open-source DICOM viewer
software, OsiriX, to quantify PF in SSc patients. This
opens up an interesting window of research focused on
its use as a useful additional tool to study pulmonary
changes in patients with SSc. OsiriX offers peculiar
characteristics for the pulmonary assessment: it is a free
of charge software and no intensive computer training
is mandatory to become able to perform lung segmen-
tation and to obtain the corresponding distribution his-
togram (essential for kurtosis, MLA and skewness
calculation).
From an analysis of our results, the following consid-
erations can be formulated. First, the quantitative OsiriX
analysis showed a high agreement with the assessment of
PF compared with the semiquantitative HRCT analysis
performed by experienced radiologists in lung fibrosis. The
sections with the lowest kurtosis and highest MLA had a
semiquantitative score indicative of an extensive pulmon-
ary involvement.
Fig. 2 Examples of segmentation of lung and its distribution
histogram. On the left the lower section of the lungs of three patients
with different semiquantitative scoring on a Likert scale (LS)
indicated in the upper right. The green area corresponds to the
region that the OsiriX algorithm recognizes as lung parenchyma. To
the right of each section, there is the corresponding distribution
histogram of lung attenuation values and the consequent descriptive
parameters: kurtosis (K), symmetry (S), mean lung attenuation
(MLA) expressed in Hounsfield unit. Lung attenuation values assume
approximately a bimodal distribution. Note that the ‘‘highest peak’’
(mode’s main distribution) tends to be progressively closer to the HU
value 0 (to the right along the horizontal axis) and the base tends to
expand making the ‘‘minor peak’’ hardly intelligible
Rheumatol Int
123
Second, OsiriX demonstrated to be reliable in the
assessment of PF since, as shown by histological studies,
it confirms that PF tends to increase progressively with an
apical–basal gradient in SSc patients [7]. In fact, we
observed that the basal lung areas showed a lower degree
of kurtosis than the apical ones. Third, the values of MLA
obtained are in line with those reported previously in other
studies [8, 11, 16]. Fourth, the mean time spent to perform
an OsiriX quantitative analysis for each patient was much
less than the semiquantitative assessment of HRCT per-
formed by the radiologist’s lecture.
Taking these observations into account, the results of
our study seem to be encouraging since they induce to
consider OsiriX a useful tool for the assessment of PF.
Despite these aspects, we are aware that there are limita-
tions in our study. First, a correct analysis by OsiriX
depends directly on a correct HRCT data acquisition.
Second, the low number of enrolled patients does not
permit an accurate evaluation in terms of sensitivity and
specificity which could more strongly support these data.
Third, OsiriX segmentation algorithm does not consider
completely the PF at the basal peripheral regions (usually
in late stage disease patients). This flaw probably could be
explained by the fact that the PF is quite difficult to dis-
tinguish (in terms of density) from surrounding tissues
[20]. However, the operator can manually complete the
segmentation, indicating the neglected areas. Fourth,
OsiriX makes a quantitative assessment of pulmonary
fibrosis but does not allow the automatic identification of
elementary lesions. Moreover, we have evaluated only
three pulmonary sections, not the entire lung. Note, how-
ever, that some authors [21] have demonstrated the corre-
lation between the extent of pulmonary fibrosis in three
sections and the PF width in the entire lung. Finally,
although OsiriX is a free of charge software, it needs to be
installed on a Mac OsX platform.
Despite the above limitations, we believe that OsiriX
can be used as an adjunct method in the assessment of PF.
Besides, it can play a relevant role for screening purposes
aiming towards the early identification of SSc patients that
require a chest HRCT.
In conclusion, our pilot study provides the new working
hypothesis that OsiriX DICOM viewer software may be a
useful post-processing tool to achieve a quantitative eval-
uation of PF in SSc patients. Nevertheless, additional
investigations on larger series of cohorts about the whole
lung, studying sensitivity and specificity, may be useful to
more strongly support these data.
Conflict of interest The authors declare that they have no conflict
of interest.
Glossary
Open source Computer software available in
source code form: the source
code and certain other rights
normally reserved for copyright
holders are provided under a free
Fig. 3 a MLA stratified
according to the
semiquantitative evaluation.
b Kurtosis stratified according
to pulmonary section. c Kurtosis
stratified according to the
semiquantitative evaluation.
d Correlation between kurtosis
and MLA. MLA mean lung
attenuation
Rheumatol Int
123
software licence that permits
users to study, change, improve
and at times also to distribute the
software
DICOM Digital Imaging and COmmu-
nications in Medicine is a standard
for handling, storing, printing and
transmitting information in medical
imaging. It includes a reproduction
of the original uncompressed image
and a series of additional data
(called metadata) that are useful
for medical interpretation (such as
patient demographics, spatial
resolution, distance between the
planes of acquisition, radiation
absorbed)
Attenuation Value expressed in units Hounsfield
(HU) indicative of the X-rays
absorption degree that pass
through a single spatial unit (pixel)
of the lung examined with the CT
Distribution histogram Histogram showing the
frequency with a given value of
lung attenuation occurs in a given
set of CT image pixels
MLA Mean lung attenuation: arithmetic
mean of all lung attenuation values
of a single section
Kurtosis Distribution parameter which
gives an account of the
similarity between the histogram
distribution ‘‘shape’’ with the
Gaussian curve pattern. High
values indicate a departure from
normal distribution
Skewness Distribution parameter which gives
an account of how the histogram
distribution is symmetrical about a
central value. High values indicate
an asymmetry of the distribution
References
1. Steen VD, Conte C, Owens GR, Medsger TA (1994) Severe
restrictive lung disease in systemic sclerosis. Arthritis Rheum
37:1283–1289
2. Karassa FB, Ioannidis JPA (2008) Mortality in systemic sclerosis.
Clin Exp Rheumatol 26:S85–S93
3. Goh NSL, Desai SR, Veeraraghavan S et al (2008) Interstitial lung
disease in systemic sclerosis: a simple staging system. Am J Respir
Crit Care Med 177:1248–1254. doi:10.1164/rccm.200706-877OC
4. Wells AU (2008) High-resolution computed tomography and
scleroderma lung disease. Rheumatology (Oxford) 47(Suppl
5):v59–v61. doi:10.1093/rheumatology/ken271
5. Goldin JG, Lynch DA, Strollo DC et al (2008) High-resolution
CT scan findings in patients with symptomatic scleroderma-
related interstitial lung disease. Chest 134:358–367. doi:10.1378/
chest.07-2444
6. Warrick JH, Bhalla M, Schabel SI, Silver RM (1991) High res-
olution computed tomography in early scleroderma lung disease.
J Rheumatol 18:1520–1528
7. Cozzi F, Chiesura Corona M, Rizzi M et al (2001) Lung fibrosis
quantified by HRCT in scleroderma patients with different dis-
ease forms and ANA specificities. Reumatismo 53:55–62
8. Zavaletta VA, Bartholmai BJ, Robb RA (2007) High resolution
multidetector CT-aided tissue analysis and quantification of lung
fibrosis. Acad Radiol 14:772–787. doi:10.1016/j.acra.2007.03.
009
9. Sverzellati N, Zompatori M, De Luca G et al (2005) Evaluation
of quantitative CT indexes in idiopathic interstitial pneumonitis
using a low-dose technique. Eur J Radiol 56:370–375. doi:10.
1016/j.ejrad.2005.05.012
10. Sumikawa H, Johkoh T, Yamamoto S et al (2006) Quantitative
analysis for computed tomography findings of various diffuse
lung diseases using volume histogram analysis. J Comput Assist
Tomogr 30:244–249
11. Best AC, Lynch AM, Bozic CM et al (2003) Quantitative CT
indexes in idiopathic pulmonary fibrosis: relationship with
physiologic impairment. Radiology 228:407–414. doi:10.1148/
radiol.2282020274
12. Rosset A, Spadola L, Ratib O (2004) OsiriX: an open-source
software for navigating in multidimensional DICOM images.
J Digit Imaging 17:205–216. doi:10.1007/s10278-004-1014-6
13. LeRoy EC, Black CM, Fleischmajer R et al (1988) Scleroderma
(systemic sclerosis): classification, subsets and pathogenesis.
J Rheumatol 15:202–205
14. Strange C, Seibold JR (2008) Scleroderma lung disease: ‘‘if you
don’t know where you are going, any road will take you there’’.
Am J Respir Crit Care Med 177:1178–1179. doi:10.1164/rccm.
200802-304ED
15. Collins CD, Wells AU, Hansell DM et al (1994) Observer vari-
ation in pattern type and extent of disease in fibrosing alveolitis
on thin section computed tomography and chest radiography.
Clin Radiol 49:236–240
16. Lynch DA (2007) Quantitative CT of fibrotic interstitial lung
disease. Chest 131:643–644. doi:10.1378/chest.06-2955
17. Sumikawa H, Johkoh T, Yamamoto S et al (2009) Computed
tomography values calculation and volume histogram analysis for
various computed tomographic patterns of diffuse lung diseases.
J Comput Assist Tomogr 33:731–738. doi:10.1097/RCT.
0b013e31818da65c
18. Camiciottoli G, Orlandi I, Bartolucci M et al (2007) Lung CT
densitometry in systemic sclerosis: correlation with lung func-
tion, exercise testing, and quality of life. Chest 131:672–681.
doi:10.1378/chest.06-1401
19. Kim HG, Tashkin DP, Clements PJ et al (2010) A computer-
aided diagnosis system for quantitative scoring of extent of lung
fibrosis in scleroderma patients. Clin Exp Rheumatol 28:S26–S35
20. Hoffman EA, Reinhardt JM, Sonka M et al (2003) Character-
ization of the interstitial lung diseases via density-based and
texture-based analysis of computed tomography images of lung
structure and function. Acad Radiol 10:1104–1118
21. Kazerooni EA, Martinez FJ, Flint A et al (1997) Thin-section CT
obtained at 10-mm increments versus limited three-level thin-
section CT for idiopathic pulmonary fibrosis: correlation with
pathologic scoring. AJR Am J Roentgenol 169:977–983
Rheumatol Int
123