Keywords
deep endometriosis - color score ratio - strain elastography - strain ratio - endometriotic
lesion
Introduction
Endometriosis is a common chronic inflammatory disease characterized by ectopic endometrial
tissue outside the uterine cavity [1]
[2]
[3]. Its prevalence is roughly 5–10% in the general female population but reaches 71–87%
in women with pelvic pain [2]
[3]. Endometriosis can be characterized by superficial implants on the abdominal serous
membrane or distant organs (e.g., pleura or pericardium) [3]
[4]
[5]. In the pelvis, endometriosis can lead to deeper lesions affecting the bladder,
rectum, sigmoid tract, rectovaginal septum, parametrium and/or uterosacral ligaments
[5]. These lesions are commonly aggregated under deep infiltrating endometriosis (DIE)
[5]
[6], which often requires extensive surgery leading to severe morbidity [7]. Several studies have now shown that transvaginal sonography (TVS) performed by
an expert sonographer can be regarded as being an accurate method in defining DIE
nodules and their extension, as magnetic resonance imaging (MRI) [5]
[8]
[9]; although, accurate diagnosis of nodules of the uterosacral ligaments and parametrium
remains problematic [5]. Strain elastography (SE) measures tissue stiffness/elasticity and is commonly used
to characterize lesions of the breast and other organs [10]
[11]. DIE lesions are stiffer than normal surrounding tissue, but few studies have applied
SE to their characterization [12]
[13]. Hence, the aim of this study was to assess whether SE discriminates DIE from surrounding
non-endometriotic tissue. The secondary objective was to evaluate if the capabilities
of elastography in distinguishing DIE lesions differed by location.
Methods
Design and participates
This observational study was conducted between October 2020 and December 2021 in a
third-level academic hospital gynaecology outpatient facility. The study did not involve
any intervention beyond standard clinical practice. Publication of the results was
approved by the Ethics Committee of IRCCS Ospedale Policlinico San Martino, Genoa
(CER Liguria n. 19/2022). Each participant signed informed written consent for the
anonymous use of their data in clinical research.
The sample comprised all consecutive patients of 18 to 45 years of age complaining
of endometriosis symptoms referred to the specialist outpatient facility during the
study period with a previous or current diagnosis of DIE. Demographic and clinical
data were collected for each. Presence of pain at menstruation, intermenstrual pain,
and pain at intercourse was recorded [14]
[15]
[16]. The intensity of each type of pain was estimated individually on a 100-mm visual
analogue scale (VAS) [3]. Each woman underwent standard bimanual vaginal examination to assess for the presence
of stiff nodules, tenderness and mobility of pelvic organs. Transvaginal sonography
was then performed by an expert qualified practitioner (A.X.). The presence of gynaecological
diseases such as uterine myomas and adenomyosis was assessed using the MUSA criteria
[17]. Ovarian endometriosis and DIE of the posterior and anterior compartments was diagnosed
based on the IDEA consensus opinion and in accordance with recent guidelines [9]
[18]. Women with a clinical and sonographic diagnosis of DIE were further evaluated by
SE. Some of these women had previously received a surgical diagnosis of endometriosis,
and some underwent surgery afterwards if clinically indicated [19]. When surgery was performed, endometriosis was confirmed by histology.
Sonographic Measurements
A sole operator performed all ultrasound assessments using a Voluson E6 General Electric
(GE Medical System, Zipf, Austria) instrument. The ultrasound examinations were conducted
with a transvaginal wideband 5–9 Mhz transducer and dedicated elastography software
(GE Medical System, Zipf, Austria). Volume (cm3) of endometriotic nodules identified
in B-mode was evaluated by the ellipsoid formula (3 main diameters × 0.5223). Tissue
stiffness/elasticity was evaluated by SE, which measures tissue deformation or displacement
provoked by an applied pressure [10]
[12]
[13]. The strain value can be depicted using colored shading superimposed on the B-mode
image. The strain of different regions of interest (ROI) can be concomitantly evaluated,
and the strain ratio, a measure of the discrepancy in the elasticity of different
tissues, can be used to improve SE accuracy [10]
[20]
[21]. For image acquisition in B-mode, the vaginal probe was positioned in the region
of the endometriotic nodule. Afterwards, the sonographer performed a series of approximately
5 compression-decompression cycles in the elastography modality, revealing the colored
shading superimposed on the B-mode image. The compression and decompression process
were achieved using sub-centimetric motions perpendicular to the axis of the endometriotic
lesion. A feedback control bar in the ultrasound real-time elastography program was
used to check and maintain optimal compression force ([Fig. 1]). The dynamic elastography acquisition process was recorded on video, to be analyzed
afterward offline ([Video 1]). The offline analyst was blinded to patient data. The endometriotic nodule was
well characterized under SE as a homogenous blue area distinct from surrounding tissue.
Three ROIs (circular areas of 7.06 mm2), placed at an equal distance from the probe,
were: the endometriotic nodule, the surrounding non-endometriotic tissue on the left
of the nodule, and the surrounding non-endometriotic tissue on the right ([Fig. 1]). The strains and the colour scores were computed at optimal compression. For each
of the three ROIs, the SE software provided the numerical value of the strain as the
percentage of tissue deformation. The CS (purple/blue as low elasticity, yellow/green
as intermediate, and red as high) was also coded as follows: from 0 = blue/purple
to 3.0 = red [20]
[21]. The mean value of three measurements was used. Raw values for strain and CS were
used to calculate the respective ratios. Strain or CS ratio of the endometriotic nodule
was defined as the ratio between the mean of the two non-endometriotic tissue ROIs
as the nominator, divided by the value obtained from the endometriotic nodule ROI
as the denominator. The resulting value was compared to the strain or CS ratio of
the non-endometriotic tissue; this was calculated as the ratio between the mean of
the two non-endometriotic tissue ROIs as the nominator, and the minimum value between
the two ROIs from surrounding non-endometriotic tissue as the denominator. This yielded
ratios with a scale starting from 1, and progressively increasing values that correspond
to tissue of greater consistency.
Fig. 1 Endometriotic nodule (white arrow) under B-mode elastography, showing colored shading
superimposed on the image (Panel A), and example of tissue strain elastography (Panel B). The yellow circular region of interest (ROI) is located on the endometriotic nodule,
and the other two ROIs (blue and purple) on the left and right surrounding tissue.
Dynamic elastography acquisition.Video 1
Sample size assessment
The sample size was calculated according to preliminary data collected from six patients
to find any difference in strain ratio for different DIE locations paired with normal
tissue using a nonparametric test. The target sample size was therefore 4 pairs, sufficient
to detect differences in the median strain ratio between endometriosis and normal
tissue in every DIE assessed, with power 80% and a 0.05 significance level on two-sided
testing.
Data analysis
Data were analyzed using the statistical package R [22] (version 3.6.3; R Core Team (2020). Kolmogorov–Smirnov was used to test the normal
distribution of data. A Wilcoxon test or t-test was applied to the continuous variables,
as appropriate (the endometriotic lesion and the paired normal tissue were tested
using a paired test). Dichotomic variables were tested using the chi-squared or Fisher’s
exact test. Continuous data are presented as the median and interquartile range (IQR)
or mean and standard deviation. Categorical variables are expressed as frequencies,
absolute values, and percentages. Intra-operator variability was assessed via the
intraclass correlation coefficient (ICC) and its 95% confidence interval (CI.95).
The performance of strain and CS ratios in determining the presence of an endometriotic
nodule was evaluated by generating the receiver operating characteristic (ROC) curves.
ROC curves are presented with their area under the curve (AUC) and relative CI.95.
The DeLong test was used to compare AUCs of different ROC curves. For all analyses,
a two-tailed p-value <0.05 was considered significant.
Results
Population
The study examined 46 DIE nodules from 32 women. Mean age at diagnosis was 37.31±8.43
years. [Table 1] shows the characteristics of participants. 18.7% of them had had a surgical diagnosis
of endometriosis within the previous 24 months. A subsequent histological evaluation
of the nodule detected at sonography was obtained in 32.1% of the nodules.
Table 1 Characteristics of the population.
|
Acronyms: COC = Combined oral contraceptives; VAS = Visual Analogue Scale.
|
|
Womenʼs background characteristics and therapy
|
|
|
Women (n.)
|
32
|
|
Age (years)
|
37.31 (±8.43)
|
|
Nulliparity
|
31.25% (10/32)
|
|
Actual surgery
|
34.38% (11/32)
|
|
Previous surgery
|
18.75% (6/32)
|
|
Medical therapy
|
34.38% (11/32)
|
|
|
28.12% (9/32)
|
|
|
6.25% (2/32)
|
|
Womenʼs symptoms
|
|
|
Dysmenorrhea
|
68.75% (22/32)
|
|
|
5 (0–8)
|
|
Ovulation pain
|
37.5% (12/32)
|
|
|
0 (0–5)
|
|
Chronic pelvic pain
|
62.5% (20/32)
|
|
|
3.5 (0–6.25)
|
|
Dyspareunia
|
46.88% (15/32)
|
|
|
0 (0–7.25)
|
|
Dyschezia
|
25% (8/32)
|
|
|
0 (0–1)
|
|
Back pain
|
43.75% (14/32)
|
|
|
0 (0–5.25)
|
|
Inguinal pain
|
9.38% (3/32)
|
|
Endometriotic nodules characteristics
|
|
|
Number
|
46
|
|
Volume (cm³)
|
0.62 (0.35–1.01)
|
|
Nodules per women
|
1 (1–2)
|
|
Locations
|
|
|
|
45.65% (21/46)
|
|
|
15.22% (7/46)
|
|
|
30.43% (14/46)
|
|
|
8.7% (4/46)
|
Transvaginal elastography
In a preliminary assessment of 6 nodules in 6 different subjects, the intra-operator
variability for strain ratio was 0.841 (CI.95 0.516–0.966), and for CS ratio it was
0.925 (CI.95 0.816–0.980).
Elastography of DIE nodules and surrounding tissue is depicted in [Fig. 1]. The strain ratio of DIE nodules (3.09, IQR 2.38–4.14) was significantly higher
than that of normal tissue (1.25, IQR 1.11–1.48) (p<0.001). Similarly, the CS ratio
of DIE nodules was significantly higher (4.62, IQR 3.83–6.94) than that of normal
tissue (1.13, IQR 1.06–1.29) (p<0.001). The distributions of strain and CS ratios
of DIE nodules was well distinguished from those of normal tissue ([Fig. 2]). ROC plots present strain and CS ratio accuracy in defining DIE implants ([Fig. 2]); the best threshold for strain ratio ROC was 1.68, with a sensitivity of 91.3%
(CI.95 82.61–97.83%) and a specificity of 82.61% (CI.95 71.74–93.48%), while the best
threshold for CS ratio ROC was 1.82, with a sensitivity of 97.83% (CI.95 93.48–100%)
and a specificity of 100% (CI.95 100–100%). CS ratio AUC (99.76%, CI.95 99.26–100%)
was higher than strain ratio AUC (91.35%, CI.95 85.23–97.47%) (p=0.007).
Fig. 2 Distribution of strain ratio (panel A) and color score (CS) ratio (Panel B) values in endometriosis versus normal tissue. ROC plot showing accuracy of strain
ratio (Panel C) and CS ratio (Panel D) values in distinguishing endometriosis nodule from non-endometriotic tissue.
Strain ratio and CS ratio of different DIE locations
Strain ratio and CS ratio were not significantly different among the different DIE
locations. Similarly, the difference between endometriosis and normal tissue strain
and CS ratios was consistently shown in all the locations investigated, even though,
for a low number of cases in the recto-vaginal septum, it did not reach statistical
significance ([Table 2]). Strain and CS ratios of nodules with histological diagnosis of endometriosis were
also calculated separately. Data were not different from those obtained when considering
all nodules ([Table 2]).
Table 2 Strain ratio and CS ratio divided by location and tissue type (endometriosis/normal
tissue). Analysis on nodules confirmed by histological diagnosis is also separately
reported.
|
(*) Differences between endometriosis and normal tissue in each location (paired Wilcoxon
test).
Acronyms: CS = color score.
|
|
A) All endometriotic nodules
|
Uterosacral ligaments (21)
|
Parametrium (7)
|
Rectum (14)
|
Recto-vaginal septum (4)
|
|
Strain ratio endometriosis
|
3.09 (2.10–4.00)
|
3.00 (2.42–3.50)
|
3.47 (2.64–4.83)
|
3.75 (3.08–4.38)
|
|
Strain ratio normal tissue
|
1.16 (1.07–1.32)
|
1.24 (1.17–1.26)
|
1.34 (1.15–1.56)
|
1.82 (1.44–2.65)
|
|
p-value (*)
|
<0.001
|
0.016
|
0.003
|
0.125
|
|
CS ratio endometriosis
|
5.17 (4.50–7.75)
|
3.83 (2.56–6.00)
|
4.50 (3.80–6.94)
|
4.00 (3.34–5.38)
|
|
CS ratio normal tissue
|
1.10 (1.05–1.26)
|
1.11 (1.07–1.16)
|
1.22 (1.09–1.30)
|
1.52 (1.36–1.64)
|
|
p-value (*)
|
<0.001
|
0.016
|
0.001
|
0.125
|
|
B) Nodules with histology
|
Uterosacral ligaments (6)
|
Parametrium (3)
|
Rectum (5)
|
Recto-vaginal septum (1)
|
|
Strain endometriosis
|
3.34 (2.14–4.00)
|
3.00 (2.50–3.46)
|
3.83 (2.67–4.30)
|
–
|
|
Strain control
|
1.30 (1.10–1.61)
|
1.24 (1.12–1.25)
|
1.15 (1.07–1.33)
|
–
|
|
p-value (*)
|
<0.001
|
0.016
|
0.003
|
–
|
|
CS endometriosis
|
4.69 (4.50–9.09)
|
6.25 (4.44–9.38)
|
4.50 (4.50–7.00)
|
–
|
|
CS control
|
1.18 (1.07–1.28)
|
1.17 (1.13–1.21)
|
1.17 (1.13–1.27)
|
–
|
|
p-value (*)
|
<0.001
|
0.016
|
0.001
|
–
|
The accuracy of the strain and CS ratios in distinguishing endometriotic nodules in
different locations was evaluated by ROC plots ([Fig. 3]). Data were similar among locations. At the uterosacral ligaments, the strain ratio
AUC was 89.91% (CI.95 81.49–98.33%) and the CS ratio AUC 100% (CI.95 100–100%), the
difference between the two being significant (p=0.019). At the parametrium, the strain
ratio AUC was 90.99% (CI.95 82.6–99.39%) and the CS ratio AUC 100% (CI.95 100–100%),
the difference between the two being significant (p=0.036). At the rectum, the strain
ratio AUC was 92.31% (CI.95 85.46–99.16%) and the CS ratio AUC 99.22% (CI.95 97.57–100%),
the difference between the two being close to significant (p=0.056). At the recto-vaginal
septum, the strain ratio AUC was 96.2% (CI.95 90.77–100%) while the CS ratio AUC was
100% (CI.95 100–100%); this difference was not significant, probably due to the limited
number of cases in the sample (p=0.169). Separate ROC curves for DIE nodules confirmed
by histology were also calculated. Data were comparable to those reported for all
nodules considered together (Supplemental Figure 1).
Fig. 3 ROC plots show the accuracy of strain and color score (CS) ratios in identifying endometriotic
nodules in different locations.
Discussion
Principal findings
At SE analysis, strain and CS ratios for DIE lesions and normal tissue were significantly
different, irrespective of nodule location. The accuracy of discriminating endometriotic
tissue from normal tissue was high in all areas examined.
Results in the context of what is known
The value of tissue stiffness obtained by SE is variable and operator dependent [23]
[24]. Accordingly, it is not feasible to appropriately define absolute stiffness by this
method. Usually, the stiffness ratio of sites of interest and adjacent tissue is more
appropriate to eliminate inter-operator variability, as the compression and decompression
exerted by the operator similarly affect the two areas [10]
[25]
[26]. In this study, the stiffness ratio of two areas of the same tissue surrounding
the nodule was compared with the stiffness of the nodule. DIE nodules were always
stiffer than surrounding tissue, and the ratio of non-endometriotic/DIE nodule was
consistently above 1. We aimed to evaluate whether SE analysis expressed in this way
can be used to differentiate DIE from surrounding tissue. To achieve this goal, we
calculated the normal tissue ratio as the ratio between the mean stiffness of the
two areas surrounding the nodule as the numerator, divided by the value of the stiffer
area of the two. In this case too, the ratio was consistently above 1. It emerged
that the distribution of both strain and CS ratios of DIE nodules was markedly different
from those of surrounding tissue. Cut-off values differentiating nodule from surrounding
tissue were similar for strain and CS ratios, and both showed high sensitivity and
specificity. For CS ratio, a cut-off value of 1.82 was associated with sensitivity
and specificity close to 100%. Results were similar for any location of DIE, as previously
reported for shear-wave elastography [27]. Similar results were obtained when considering DIE nodules with a histological
diagnosis separately.
Clinical implications
SE has previously been applied to DIE in Douglas’s cul-de-sac. Although greater tissue
stiffness was found in women with DIE, no attempt was performed to evaluate the sensitivity
of this technique [12]. In a recent article, share wave elastography proved to be capable of detecting
DIE nodules [27]. Share-wave elastography is more reproducible than SE and gives an absolute tissue
stiffness value. However, we demonstrate herein that SE detects DIE, from surrounding
non-endometriotic tissue, with high sensitivity and specificity when appropriate ROIs
are used for calculating strain or CS ratio. In particular, the distribution curves
of strain and CS ratios of normal versus endometriotic tissue had minimal if any superimposition.
In normal tissue the curve was narrow, while in DIE it was broader and flatter, indicating
a large variability of DIE nodule stiffness, probably due to the different proportions
of fibrotic tissue [27].
Strengths and limitations
One of the limitations of this study is that SE is operator-dependent. To minimize
this bias, we used three different strategies: a single experienced operator performed
all the elastography evaluations; strain values were analyzed at optimal tissue compression,
as indicated by the elastography software; strain or CS values were not considered
as an absolute value but as the ratio of ROIs values. Yet, we acknowledge that the
inter-observer reproducibility of our analysis needs to be tested in further studies.
Some women had undergone surgery before our evaluation and post-surgical scars may
have modified tissue stiffness. We did not observe relevant differences between women
with and without a previous surgery, but the number of subjects was limited, and additional
studies are necessary to investigate this possible confounding. Elastography was applied
after nodule identification, and in this condition, both strain and CS ratios discriminated
nodules from normal tissue. It remains to be seen whether elastography may help less-skilled
sonographers to distinguish DIE from surrounding tissue, becoming an additional tool
for achieving an accurate diagnosis of DIE.
Research implications
Prospective studies should help determine whether our method is a sensitive and specific
tool for diagnosing DIE, particularly in areas where ultrasonography is less accurate,
such as the uterosacral ligaments and the parametrium [5]. The stiffness of DIE nodules varies widely, and whether these differences are related
to a different DIE composition is suggestive but unproven. A different stiffness may
identify an active DIE, which can be more symptomatic and responsive to medical treatment.
Conclusions
SE expressed as both strain and CS ratio highly accurately distinguishes DIE nodules
at various locations from surrounding non-endometriotic tissue. Pending validation
and reproducibility testing of our findings by prospective studies, elastography may
represent an important tool for diagnosing DIE and possibly monitoring its evolution
either spontaneous or in response to treatment.
