Downloaded from www.ajronline.org by 190.181.12.21 on 11/08/15 from IP address 190.181.12.21. Copyright ARRS. For personal use only; all rights reserved
1191
Clinical Cervix:
Stage
The depth
invasion
I Carcinoma
of the
Value of MR Imaging in Determining Degree of Invasiveness
Hans H. Lien1 Viggo Blomlie1 Kjell Kj#{248}rstad2 Vera AbeIer OIav Kaalhus4
of tumor
measured
at histologic
examination
is the most important
prognostic factor in early-stage carcinoma of the uterine cervix. The ability of MR to estimate the depth of tumor invasion was studied in 47 patients who subsequently underwent radical hysterectomy. In two patients, MR failed to detect tumors with infiltration depths of 2 and 4 mm. The maximum sagittal tumor length along the axis of the cervix (D) and the maximum tumor area on sagittal (S) and axial (A) images were measured with MR. The depth of tumor infiltration determined histologically (I) correlated with the degree of infiltration noted on MR images. The univariate correlation coefficients for log I vs log D, log S, and log A were .87, .84, and .77, respectively. By stepwise inclusion of the variables in a multivanate analysis, the contributions to the coefficient of determination from including log S and log A after log D were less than 1%. The regression analysis showed that the best estimate for the depth of invasion was close to half of the sagittal tumor length measured on MR I = D/2. Our experience shows that MR is valuable in determining the degree of invasiveness in clinical stage I tumors. AJR
156:1191-1194,
June 1991
Early-stage cancer of the uterine cervix has an excellent prognosis and can be cured by surgery or radiotherapy. The most important prognostic factor is the size of the tumor, measured as depth of infiltration into the cervical stroma [1-6]. Accurate assessment of tumor size is important because it has therapeutic implications. Tumors with infiltration depths of less than 5 mm have a very limited metastatic potential, and patients with such tumors have a long-term survival rate of almost 1 00% [7]. Radical treatment procedures may not be needed in these patients. To date, the only reliable method to determine invasion depth has been histologic examination of the removed organ. Several studies have shown that MR imaging is useful for detecting and staging carcinoma of the cervix [8-1 1 ]. In the present prospective study, we compared the results of MR imaging in 47 patients treated for stage I cancer with the findings in 1990; accepted
revionDecember27l990. .
Department wegian Radium
.
after
.
of Diagnostic Radiology, The NorHospital, Montebello N-0310 Oslo
3, Norway. Address reprint requests to H. H. Lien. 2 Department of Gynecologic Oncology, The Norwegian 3
diurn
Radium
Department Hospital,
Hospital,
of Pathology,
the operative specimen. In particular, the correlation on MR and the depth of invasion noted on histologic
Subjects
between the size of the tumor examination was investigated.
and Methods
Oslo 3, Norway.
The Norwegian
Ra-
Oslo 3, Norway.
Fifty-nine a consecutive
patients and
4 Department of Biophysics, Cancer Research Institute, The Norwegian Radium Hospital, Oslo 3, Norway.
of a clinical staging biopsies, cystoscopy, radical hysterectomy
0361-803X/91/1566-1191
surgical
© American Roentgen Ray Society
by preoperative
specimen
with cervical unselected
cancer
manner.
examination under chest radiography, and lymphadenectomy. showed
conization,
no residual
large
of clinical
Each
patient
stage had
I entered
this prospective
a complete
clinical
workup
study
in
consisting
anesthesia, dilatation and curettage, multiple tumor and excretory urography. All patients underwent In eight patients, histologic examination of the tumor,
biopsies,
indicating
or curettage.
that
it had
been completely
MR images
were
of inferior
removed quality
1192
LIEN
in four
patients.
The remaining
Downloaded from www.ajronline.org by 190.181.12.21 on 11/08/15 from IP address 190.181.12.21. Copyright ARRS. For personal use only; all rights reserved
present report. The patients
47 patients
were 22-67
form
the basis
for the
years old (mean, 42).
Examinations were performed with a 1 .5-T Signa System (General Electric, Milwaukee, WI). Spin-echo (SE) pulse sequences were used and included a coronal localizer series followed by sagittal and axial T2-weighted series (2000/20-80 [TRITE]) and an axial Ti -weighted
Except for the Iocalizer series, all sequences were a 256 256 matrix and two excitations. Slice thickness was 5 mm with 2.5-mm gaps. The field of view was 36 cm. The patients did not fast before the examinations. Patients were given two glasses of water 30-45 mm before the study to fill the
ET
AL.
AJA:156,
June 1991
of the parametrium and upper part of the vagina as well as total pelvic lymphadenectomy. The cervix/isthmus was cut off and sliced in clockwise radial-length sections 2-3 mm thick. In each section the
maximum depth of tumor invasion into the cervical wall was measured by means
of a calibrated
Statistical
Methods
microscope.
series (600/20).
with
performed
bladder.
MR images clinical
stage
were evaluated of the disease.
by two radiologists
The findings
sus after the images were interpreted
who knew the
were assigned
in conference.
by consen-
The criterion for
tumor was a lesion of high or relatively high signal intensity replacing the low-signal-intensity fibrous cervical stroma on T2-weighted im-
ages. The following measurements were made (Fig. 1 ): the maximum length of the tumor in the sagittal plane along the long axis of the cervix (D), the angle at the intersection of the main axis of the cervix and the axial direction in the sagittal plane (u), the maximum tumor area on sagittal images (5), and the maximum tumor area on axial images (A). Distance and angle measurements were performed with the cross-hair cursor on the monitor. Areas were defined by tracing the tumor region with the cursor. All patients were subjected to radical
according
to the Wertheim/Meigs
technique,
The data were subjected to both univariate and multivariate correlation analysis, with the depth of infiltration as the dependent variable and the MR parameters as mutually independent variables. In a special model, the volume of the tumor as calculated from the
MR parameters was taken as the sole independent variable. In this model, the tumor was thought of as an ellipsoid with the distance D as a principal axis. The calculated tumor volume (V) was then
V
8xSxA =
X
ir
D
X
As an average semiaxis
[(ii-
D2/4S)2
-
i]sin2u.
length (L) of the ellipsoid, the variable L
was
X
=
used.
transforms of the variables were used to obtain approximately normally distributed values when the original data were not normally distributed, as assessed by normal score tests. Logarithmic
abdominal
hysterectomy
which includes
removal
Fig. 1.-T2-weighted
MR images
(SE 2000/80)
show high-signal tumor replacing low-signal fibrous stroma. Invasion depth was 15 mm at histologic examination. A, Maximal tumor length along axis of cervix (arrows). B, Angle (u) at intersection and axial plane.
of main axis of cervix
C, Maximal sagittal tumor area (arrows). D, Maximal axial tumor area (arrows).
Downloaded from www.ajronline.org by 190.181.12.21 on 11/08/15 from IP address 190.181.12.21. Copyright ARRS. For personal use only; all rights reserved
Fig. 2.-T2-weighted sagittal MR images (SE 2000/80) in patient 2 mm at histologic examination of left anterior cervical lip. A, Section B, Section
through cervical canal shows normal 7.5 mm to left of A. Small high-signal
low-signal
with tumor
fibrous
invasion
measuring
stroma.
lesion erodes fibrous stroma anteriorly (arrow).
Results
Fig. 3.-T2-weighted sagittal MR image (SE 2000/80) in patient with endocervical tumor and 5mm invasion depth at histologic examination. Highsignal tumor extends into fibrous stroma (arrows).
a large tumor, and I tumor of D = 25 mm.
MR demonstrated the lesion in 45 patients, five of whom tumors with invasion depths of 5 mm or less at histologic examination (Figs. 2 and 3). MR failed to show two tumors with invasion depths of 2 and 4 mm. The logarithmic transforms of the values of D, 5, and A in the 45 tumors found at both histologic and MR examinations were interrelated. The pairwise mutual correlation factors of log D, log S, and log A ranged from .89 to .94. The results of the linear regression analysis of log I vs each of these three MR parameters are shown in Table 1 The univariate correlation coefficients were .87, .84, and .77, respectively. As expected, the regression coefficients were close to 1 .0, 0.5, and 0.5, respectively, since D is a linear measure (mm) and S and A are areas (mm2). No significant dependence of the invasion depth on the angle u alone was found. The results of the multivariate analysis are shown in Table 2. By stepwise inclusion of the variables, the contributions to the coefficient of determination from including log S and log A after log D were less than 1 %. The correlation coefficient of log I vs the logarithm of the calculated MR volume (V) was .82, which was lower than for the variables used above. Finally, the simple dependence I = D/2 was examined and found to yield a correlation coefficient that was not significantly lower than that for the previous regression dependence. This line is shown in Figure 4, which contains the scatterplot of the individual D and I observations. A linear regression of (I D/2)2 vs D2 yielded a standard deviation of approximately 2.5 mm for the smallest tumors, 1 0% of D for
=
1 2.5
mm for a medium-sized
± 3.5
had
Discussion In previous MR studies of cervical carcinoma, much attention has been paid to the stage of the disease [8, 10-12]. Within each stage, tumor volume is of importance for the prognosis; small tumors have a better prognosis than large ones. The depth of invasion as measured at histologic examination has been found to be of particular importance [17]. Therefore, we focused on the value of MR in predicting the invasion depth of the tumor. Invasion depth was not measured directly, because conventional transverse MR images are usually not perpendicular to the long axis of the cervix. A true cross section through the cervix can be obtained by using an oblique imaging plane [1 3]. This may permit measurement of the invasion depth into the cervical wall, but this off-axis scanning sequence was not used in the present series. The fact that the tumor itself often distorts the cervical canal and thereby precludes its identification still poses a problem. The present results show that MR can give reliable information about depth of tumor invasion at histologic examination. The latter has a high linear correlation with the main diameter D and the square root of the sagittal and axial tumor areas determined from the MR images. Because the best estimate for the depth of invasion was close to half of the sagittal tumor length measured on MR, I = D/2, the more
.
-
TABLE Tumors:
1: Univariate Logarithmic
Linear Regression Transforms
Intercept
Variable Log I vs log D Log I vs log S Log I vs log A Note-I MA; S
=
=
±
SE
-0.37 -0.22 -0.20
maximum
maximum
tumor
± 0.13 ± 0.13 ± 0.16
Analysis
Antilog Intercept 0.43 0.60 0.63
images;
of Infiltration
Regression
Correlation
Coefficient SE
Coefficient (r)
1.02 ± 0.05 0.48 ± 0.05 0.46 ± 0.06
depth of tumor invasion measured area on sagittal
of Depth
A
=
histologically;
maximum
tumor
.87 .84 .77
vs MR Parameters
Relative
of 45
SD (%)
Significance Level (p)
35 40 45
1191
Clinical Cervix:
Stage
The depth
invasion
I Carcinoma
of the
Value of MR Imaging in Determining Degree of Invasiveness
Hans H. Lien1 Viggo Blomlie1 Kjell Kj#{248}rstad2 Vera AbeIer OIav Kaalhus4
of tumor
measured
at histologic
examination
is the most important
prognostic factor in early-stage carcinoma of the uterine cervix. The ability of MR to estimate the depth of tumor invasion was studied in 47 patients who subsequently underwent radical hysterectomy. In two patients, MR failed to detect tumors with infiltration depths of 2 and 4 mm. The maximum sagittal tumor length along the axis of the cervix (D) and the maximum tumor area on sagittal (S) and axial (A) images were measured with MR. The depth of tumor infiltration determined histologically (I) correlated with the degree of infiltration noted on MR images. The univariate correlation coefficients for log I vs log D, log S, and log A were .87, .84, and .77, respectively. By stepwise inclusion of the variables in a multivanate analysis, the contributions to the coefficient of determination from including log S and log A after log D were less than 1%. The regression analysis showed that the best estimate for the depth of invasion was close to half of the sagittal tumor length measured on MR I = D/2. Our experience shows that MR is valuable in determining the degree of invasiveness in clinical stage I tumors. AJR
156:1191-1194,
June 1991
Early-stage cancer of the uterine cervix has an excellent prognosis and can be cured by surgery or radiotherapy. The most important prognostic factor is the size of the tumor, measured as depth of infiltration into the cervical stroma [1-6]. Accurate assessment of tumor size is important because it has therapeutic implications. Tumors with infiltration depths of less than 5 mm have a very limited metastatic potential, and patients with such tumors have a long-term survival rate of almost 1 00% [7]. Radical treatment procedures may not be needed in these patients. To date, the only reliable method to determine invasion depth has been histologic examination of the removed organ. Several studies have shown that MR imaging is useful for detecting and staging carcinoma of the cervix [8-1 1 ]. In the present prospective study, we compared the results of MR imaging in 47 patients treated for stage I cancer with the findings in 1990; accepted
revionDecember27l990. .
Department wegian Radium
.
after
.
of Diagnostic Radiology, The NorHospital, Montebello N-0310 Oslo
3, Norway. Address reprint requests to H. H. Lien. 2 Department of Gynecologic Oncology, The Norwegian 3
diurn
Radium
Department Hospital,
Hospital,
of Pathology,
the operative specimen. In particular, the correlation on MR and the depth of invasion noted on histologic
Subjects
between the size of the tumor examination was investigated.
and Methods
Oslo 3, Norway.
The Norwegian
Ra-
Oslo 3, Norway.
Fifty-nine a consecutive
patients and
4 Department of Biophysics, Cancer Research Institute, The Norwegian Radium Hospital, Oslo 3, Norway.
of a clinical staging biopsies, cystoscopy, radical hysterectomy
0361-803X/91/1566-1191
surgical
© American Roentgen Ray Society
by preoperative
specimen
with cervical unselected
cancer
manner.
examination under chest radiography, and lymphadenectomy. showed
conization,
no residual
large
of clinical
Each
patient
stage had
I entered
this prospective
a complete
clinical
workup
study
in
consisting
anesthesia, dilatation and curettage, multiple tumor and excretory urography. All patients underwent In eight patients, histologic examination of the tumor,
biopsies,
indicating
or curettage.
that
it had
been completely
MR images
were
of inferior
removed quality
1192
LIEN
in four
patients.
The remaining
Downloaded from www.ajronline.org by 190.181.12.21 on 11/08/15 from IP address 190.181.12.21. Copyright ARRS. For personal use only; all rights reserved
present report. The patients
47 patients
were 22-67
form
the basis
for the
years old (mean, 42).
Examinations were performed with a 1 .5-T Signa System (General Electric, Milwaukee, WI). Spin-echo (SE) pulse sequences were used and included a coronal localizer series followed by sagittal and axial T2-weighted series (2000/20-80 [TRITE]) and an axial Ti -weighted
Except for the Iocalizer series, all sequences were a 256 256 matrix and two excitations. Slice thickness was 5 mm with 2.5-mm gaps. The field of view was 36 cm. The patients did not fast before the examinations. Patients were given two glasses of water 30-45 mm before the study to fill the
ET
AL.
AJA:156,
June 1991
of the parametrium and upper part of the vagina as well as total pelvic lymphadenectomy. The cervix/isthmus was cut off and sliced in clockwise radial-length sections 2-3 mm thick. In each section the
maximum depth of tumor invasion into the cervical wall was measured by means
of a calibrated
Statistical
Methods
microscope.
series (600/20).
with
performed
bladder.
MR images clinical
stage
were evaluated of the disease.
by two radiologists
The findings
sus after the images were interpreted
who knew the
were assigned
in conference.
by consen-
The criterion for
tumor was a lesion of high or relatively high signal intensity replacing the low-signal-intensity fibrous cervical stroma on T2-weighted im-
ages. The following measurements were made (Fig. 1 ): the maximum length of the tumor in the sagittal plane along the long axis of the cervix (D), the angle at the intersection of the main axis of the cervix and the axial direction in the sagittal plane (u), the maximum tumor area on sagittal images (5), and the maximum tumor area on axial images (A). Distance and angle measurements were performed with the cross-hair cursor on the monitor. Areas were defined by tracing the tumor region with the cursor. All patients were subjected to radical
according
to the Wertheim/Meigs
technique,
The data were subjected to both univariate and multivariate correlation analysis, with the depth of infiltration as the dependent variable and the MR parameters as mutually independent variables. In a special model, the volume of the tumor as calculated from the
MR parameters was taken as the sole independent variable. In this model, the tumor was thought of as an ellipsoid with the distance D as a principal axis. The calculated tumor volume (V) was then
V
8xSxA =
X
ir
D
X
As an average semiaxis
[(ii-
D2/4S)2
-
i]sin2u.
length (L) of the ellipsoid, the variable L
was
X
=
used.
transforms of the variables were used to obtain approximately normally distributed values when the original data were not normally distributed, as assessed by normal score tests. Logarithmic
abdominal
hysterectomy
which includes
removal
Fig. 1.-T2-weighted
MR images
(SE 2000/80)
show high-signal tumor replacing low-signal fibrous stroma. Invasion depth was 15 mm at histologic examination. A, Maximal tumor length along axis of cervix (arrows). B, Angle (u) at intersection and axial plane.
of main axis of cervix
C, Maximal sagittal tumor area (arrows). D, Maximal axial tumor area (arrows).
Downloaded from www.ajronline.org by 190.181.12.21 on 11/08/15 from IP address 190.181.12.21. Copyright ARRS. For personal use only; all rights reserved
Fig. 2.-T2-weighted sagittal MR images (SE 2000/80) in patient 2 mm at histologic examination of left anterior cervical lip. A, Section B, Section
through cervical canal shows normal 7.5 mm to left of A. Small high-signal
low-signal
with tumor
fibrous
invasion
measuring
stroma.
lesion erodes fibrous stroma anteriorly (arrow).
Results
Fig. 3.-T2-weighted sagittal MR image (SE 2000/80) in patient with endocervical tumor and 5mm invasion depth at histologic examination. Highsignal tumor extends into fibrous stroma (arrows).
a large tumor, and I tumor of D = 25 mm.
MR demonstrated the lesion in 45 patients, five of whom tumors with invasion depths of 5 mm or less at histologic examination (Figs. 2 and 3). MR failed to show two tumors with invasion depths of 2 and 4 mm. The logarithmic transforms of the values of D, 5, and A in the 45 tumors found at both histologic and MR examinations were interrelated. The pairwise mutual correlation factors of log D, log S, and log A ranged from .89 to .94. The results of the linear regression analysis of log I vs each of these three MR parameters are shown in Table 1 The univariate correlation coefficients were .87, .84, and .77, respectively. As expected, the regression coefficients were close to 1 .0, 0.5, and 0.5, respectively, since D is a linear measure (mm) and S and A are areas (mm2). No significant dependence of the invasion depth on the angle u alone was found. The results of the multivariate analysis are shown in Table 2. By stepwise inclusion of the variables, the contributions to the coefficient of determination from including log S and log A after log D were less than 1 %. The correlation coefficient of log I vs the logarithm of the calculated MR volume (V) was .82, which was lower than for the variables used above. Finally, the simple dependence I = D/2 was examined and found to yield a correlation coefficient that was not significantly lower than that for the previous regression dependence. This line is shown in Figure 4, which contains the scatterplot of the individual D and I observations. A linear regression of (I D/2)2 vs D2 yielded a standard deviation of approximately 2.5 mm for the smallest tumors, 1 0% of D for
=
1 2.5
mm for a medium-sized
± 3.5
had
Discussion In previous MR studies of cervical carcinoma, much attention has been paid to the stage of the disease [8, 10-12]. Within each stage, tumor volume is of importance for the prognosis; small tumors have a better prognosis than large ones. The depth of invasion as measured at histologic examination has been found to be of particular importance [17]. Therefore, we focused on the value of MR in predicting the invasion depth of the tumor. Invasion depth was not measured directly, because conventional transverse MR images are usually not perpendicular to the long axis of the cervix. A true cross section through the cervix can be obtained by using an oblique imaging plane [1 3]. This may permit measurement of the invasion depth into the cervical wall, but this off-axis scanning sequence was not used in the present series. The fact that the tumor itself often distorts the cervical canal and thereby precludes its identification still poses a problem. The present results show that MR can give reliable information about depth of tumor invasion at histologic examination. The latter has a high linear correlation with the main diameter D and the square root of the sagittal and axial tumor areas determined from the MR images. Because the best estimate for the depth of invasion was close to half of the sagittal tumor length measured on MR, I = D/2, the more
.
-
TABLE Tumors:
1: Univariate Logarithmic
Linear Regression Transforms
Intercept
Variable Log I vs log D Log I vs log S Log I vs log A Note-I MA; S
=
=
±
SE
-0.37 -0.22 -0.20
maximum
maximum
tumor
± 0.13 ± 0.13 ± 0.16
Analysis
Antilog Intercept 0.43 0.60 0.63
images;
of Infiltration
Regression
Correlation
Coefficient SE
Coefficient (r)
1.02 ± 0.05 0.48 ± 0.05 0.46 ± 0.06
depth of tumor invasion measured area on sagittal
of Depth
A
=
histologically;
maximum
tumor
.87 .84 .77
vs MR Parameters
Relative
of 45
SD (%)
Significance Level (p)
35 40 45
