1990, 77K? British Journal of Radiology, 63, 673-679
VOLUME 63 NUMBER 753
SEPTEMBER 1990
The British Journal of Radiology Magnetic resonance imaging, computed tomography and endosonography in the local staging of carcinoma of the cervix By Mark Cobby, MRCP, FRCR, tJames Browning, FRCS, MRCOG, tAngela Jones, MRCP, FRCR, ^Elizabeth Whipp, MRCP, FRCR and *Paul Goddard, MD, FRCR Departments of Radiodiagnosis, tGynaecology and ^Radiotherapy, Bristol Royal Infirmary, Bristol BS2 8HW (Received January 1990 and in revised form March 1990)
Abstract. Thirty-seven patients with carcinoma of the cervix were prospectively staged by examination under anaesthesia (EUA), transvaginal and transrectal ultrasound, computed tomography (CT) and magnetic resonance imaging (MRI). Pathological correlation was available for 20 patients. In the pathologically staged patients, EUA agreed with the staging in 17, understaging three patients. Endosonography agreed with the staging in 19, CT in 16 and MRI in 18 patients. For the remaining 17 patients, endosonography agreed with the EUA findings in 13, CT in 12 and MRI in 12. This study has shown that endosonography and MRI are more accurate than CT in the local staging of carcinoma of the cervix. Computed tomography was least accurate in staging early tumours and differentiating between Stage Ib and lib disease. Lymph node involvement was detected with equal frequency by both CT and MRI. Magnetic resonance imaging was useful in identifying vaginal and bladder wall involvement and in one patient showed features due to an unsuspected early pregnancy.
Carcinoma of the cervix is the second commonest gynaecological cancer. Its prognosis is determined by the stage of the disease and the volume of tumour at presentation, falling steeply and progressively from Stage I to Stage IV disease with 5-year survival rates of 90% for Stage I and approximately 15% for Stage IV disease (Shingleton & Orr, 1983). The stage of the disease is currently assessed clinically by examination under anaesthesia and influences the type of treatment offered. The findings are classified according to the definitions of the International Federation of Gynaecology and Obstetrics (FIGO) stages (Table I). Clinical assessment of parametrial involvement, spread to the pelvic side wall or regional lymph node involvement can be difficult and result in significant errors in under and over staging (van Nagell et al, 1971; Averette et al, 1975; Lagasse et al, 1980). Until recently computed tomography (CT) was the only imaging technique available for directly assessing parametrial and pelvic wall spread. Its ability to do so accurately though has been disappointing (Whitley et al, 1982). False positive diagnoses of parametrial involvement are common and, in prospective studies, CT has not been able to differentiate Stage Ib from early lib disease reliably (Walsh & Goplerud, 1981). Magnetic resonance imaging (MRI) has.been used in •Author for correspondence. Vol. 63, No. 753
the staging of cervical carcinoma with promising results (Powell et al, 1986; Hricak et al, 1988; Greco et al, 1989; Togashi et al, 1989). The pelvic organs are well shown by MRI with its high contrast resolution and direct multiplanar imaging. The problems with degradation of the images from respiratory motion seen in the upper abdomen do not occur. Cervical carcinoma can be differentiated from the normal cervix by its higher signal intensity on r2-weighted sequences. Parametrial, vaginal wall invasion and lymph node involvement can all be demonstrated and tumour volume calculated. Transrectal ultrasound is being used with increasing success in the diagnosis and staging of prostatic carcinoma (Griffiths et al, 1989). The technique can be Table I. FIGO staging of cervical carcinoma la Ib Ila lib
Microinvasive (only diagnosed histologically) Clinical invasive carcinoma confined to cervix Involvement of vagina (but not lower third) Parametrial involvement but not extending to pelvic side wall
Ilia Involvement of lower third of vagina Illb Extension to pelvic wall and/or obstruction of a ureter IVa Involvement of bladder mucosa or rectum and/or extension beyond the true pelvis IVb Spread to distant organs
673
M Cobby, J Browning, A. Jones E. Whipp and P. Goddard
(b) Figure 1. Endosonographic technique used to obtain axial and sagittal images with (a) the radial rotating transrectal transducer and (b) transvaginal offset sector probe.
adapted to demonstrate the cervix and parametria using a standard radial rotating transducer to produce axially orientated images (Fig. la) (Dragsted et al, 1987; Yuhara et al, 1987). A transvaginal transducer with an off-set sagittal sector probe will demonstrate the cervix and body of the uterus in the longitudinal plane allowing tumour volume to be calculated (Fig. lb). This study prospectively compared the findings of endosonography, CT and MRI in 37 patients with invasive carcinoma of the cervix. Patients and methods Thirty-seven patients with an average age of 38 years (range 24 to 65 years) with pathologically proven invasive carcinoma of the cervix were staged clinically by examination under anaesthesia (EUA) by a consultant radiotherapist or gynaegologist. Endosonographic examination was performed by one of two operators without knowledge of the clinical findings. The CT and MRI images were read independently by two observers, again without knowledge of the findings from the other investigations. Discrepancies in independent interpretation of the CT and MRI images were resolved by consensus. Endosonography was performed using a Bruel and Kjaer 1846 Module with a 1850 radial rotating 5.5 MHz transrectal transducer and an 8537 7 MHz offset sagittal sector transvaginal probe. Measurements made in three
Illb IV
674
Results The clinically and radiologically assigned stages of all 37 patients are shown in Table II and representative Table III. Pathological staging compared with examination under anaesthesia, endosonography, CT and MRI
Table II. Clinical and radiological staging
Ib Ha lib
dimensions were used to calculate tumour volume. The kidneys were examined with a transabdominal transducer to assess whether upper tract dilatation was present. The CT examinations were performed on a Siemens Somatom DRH or a Philips Tomoscan 350 body scanner with contiguous 8 mm slices through the true pelvis and alternate slices up to the aortic bifurcation. Magnetic resonance imaging was performed using a Picker Vista 2055 HP 0.5 tesla MR scanner. Contiguous multislice images 10 mm thick were obtained in the transverse and sagittal planes. Proton density (time to repeat (TR)=1800, time to echo (TE) = 30ms), T2- weighted (TR=1800, TE = 80ms), r, = weighted (TR = 500, TE = 26) and short tau inversion recovery (STIR) (TR=1500, inversion time (TI)=100, TE = 30ms) sequences were used. The total scanning time for each patient was less than 1 h. Eighteen patients subsequently had a radical hysterectomy and lymphadenectomy performed within 2 weeks of the staging investigations. The remaining patients were treated by intracavity and external beam irradiation. Post-mortem findings were available for two patients who died early in the course of their treatment.
Examination under anaesthesia
Endosonography
Computed Magnetic tomography resonance imaging
25 0 8 4 0
22 0 7 8 0
20 0 8 6 3
21 1 5 6 4
Examina- Endosono- Computed Magnetic tion under graphy tomography resonance anaesthesia imaging Correct 17 Overstaged 0 Understaged 3
19 0 1
16 2 2
18 1 1
Accuracy
95%
80%
90%
85%
The British Journal of Radiology, September 1990
Staging cervical carcinoma with MRI, CT and endosonography
Figure 2. Stage Ib cervical carcinoma, (a) Sagittal and (b) transverse r2-weighted magnetic resonance images. The high signal intensity cervical carcinoma (arrow) is completely surrounded by the normal low signal intensity cervical stroma. Normal high signal intensity parametrial tissue is shown on the axial image.
illustrations of the endosonographic and MRI findings are shown in Figs 2 to 6. The accuracy of each staging investigation compared with the pathological staging available for 20 of these patients is shown in Table III. In this latter group, 13 patients with pathologically Stage Ib disease were all correctly staged by EUA and endosonography; one patient was thought to have Stage Ha disease by MRI and two considered to have Stage lib on CT. The two patients with pathologically Stage lib disease were correctly assigned by all investigations. Four patients with Stage Hlb disease were correctly assigned by endosonography, but MRI and CT each down-graded one patient and EUA two patients. The one patient with Stage IV disease was correctly identified by CT and MRI but was thought to be Stage Hlb by endosonography and EUA. For the 17 patients without pathological correlation, endosonography agreed with the EUA findings in 13, CT in 12 and MRI in 12. Computed tomography downgraded one patient; for all other patients where there was discordance between the EUA and imaging findings the staging was up-graded by the radiological investigations. Pelvic lymphadenopathy was seen in three patients by MRI and CT. Bladder wall involvement was well demonstrated by MRI in one patient but was not apparent on the other imaging investigations. The MRI examination of one patient with Stage Ib disease showed an abnormal collection of fluid in the pouch of Douglas, a prominent endomentrial "reaction" and an apparent soft-tissue mass in the uterine Vol. 63, No. 753
fundus which was interpreted as an early pregnancy. At laparotomy a tubal ectopic, but no intrauterine pregnancy was found, which even in retrospect was not visible on the MR images; endosonography performed 2 weeks earlier was normal. The findings in this patient have been described in detail elsewhere (Browning et al, 1990). Discussion This study has shown that all of the techniques could accurately stage carcinoma of the cervix but that endosonography and MRI were marginally better. No comparison with lymphography was made. Relatively few pathologically confirmed Stage lib patients were examined, as in Western practice radiotherapy is the usual treatment for such tumours. For radical hysterectomy to be considered it is important to be able to distinguish between Stage Ib and lib disease reliably. Early parametrial involvement can be difficult to diagnosis with CT (Whitley et al, 1982; Vick et al, 1984). In this study a CT technique of 8 mm contiguous scans was used. Intravenous contrast media coupled with thinner slices and image reconstruction may potentially have improved the diagnostic accuracy. On CT, the cervix appears as a round or oval smooth soft-tissue structure outlined by parametrial fat. The cardinal ligaments are occasionally seen as smooth, laterally tapering, triangular structures extending from the side of the cervix with the distal ureters, surrounded by periureteral fat, lying anterolaterally (Vick et al, 1984). The margins of the normal fat containing parametria are not usually 675
M Cobby, J Browning, A. Jones E. Whipp and P. Goddard
Figure 3. Stage lib cervical carcinoma, (a) Sagittal and (b) transverse r2-weighted images. The large high signal intensity tumour is shown to extend into the left parametrium on the axial image (arrow). Full thickness disruption of the cervical stroma is shown on the right, although definite parametrial extension is not demonstrated.
defined but may contain a variable number of small soft-tissue dots and strands thought to represent parauterine vessels, nerves and lymphatics. These should not exceed 3^4 mm in diameter (Vick et al, 1984) but can be a source of false positive diagnosis of parametria invasion which is usually seen as a mass distorting the
cervical margins or as prominent, sometimes lobulated, parametrial soft-tissue densities. Obliteration of the periureteral fat planes may also be demonstrated (Whitley et al, 1982; Vick et al, 1984). With MRI, cervical carcinoma appears as a high signal intensity mass on r2-weighted images which may
Figure 4. Stage lib cervical carcinoma, (a) Transverse r2-weighted and (b) transverse STIR images. On the r2-weighted image, full thickness disruption of the cervical stroma on the right is shown with subtle signal change in the adjacent parametrium. The extent of parametrial involvement, however, is more easily appreciated on the STIR image. 676
The British Journal of Radiology, September 1990
Staging cervical carcinoma with MRI, CT and endosonography
Figure 5. Stage IV cervical carcinoma, (a) Sagittal proton-density- and (b) r r weighted images. The tumour extends into the body of the uterus, partially obstructing the cervical canal. The upper two-thirds of the vagina and the posterior wall of the bladder are involved. The involved thickened bladder wall and vagina have a high signal intensity on the r2-weighted image. A small quantity of blood is present in the posterior vaginal fornix (arrow).
expand the cervical canal or disrupt the normal low signal intensity cervical stroma (Powell et al, 1986; Hricak et al, 1988). If the low signal intensity ring of cervical stroma is preserved then the tumour is confined to the cervix (Fig. 2) (Hricak et al, 1988; Togashi et al, 1989) but in the presence of full thickness stroma invasion it is not possible to exclude microscopic parametrial spread. As with CT, the margins of the normal parametria are not defined, but the vascular connective tissue produces a medium signal intensity on Tr and protondensity-weighted images and high signal intensity on r2-weighted and STIR images. Parametrial involvement is shown by irregularity of the cervical margins associated with full thickness disruption of the low signal intensity stroma and abnormal signal or mass in the parametrial region (Figs 3 and 4) (Hricak et al, 1988; Togashi et al, 1989). If morphological changes alone are used to diagnose parametrial involvement, MRI is no more accurate than CT (Williams et al, 1989). Magnetic resonance imaging in the sagittal plane is useful for showing vaginal wall involvement, extension into the body of the uterus or forwards to the bladder wall (Williams et al, 1989). These are most easily demonstrated with r2-weighted images which show the internal anatomy of these organs and the presence of tumour as a high signal intensity area (Fig. 5). As in previous studies, MRI showed no distinct advantage over CT in the detection of lymph node metastases using transverse and sagittal images (Lee et Vol. 63, No. 753
al, 1984; Williams et al, 1989) although this may not be so if the coronal plane is also included. Previous endosonographic studies have suggested that the width of the normal parametrium at its mid-point should not exceed 1 cm (Yuhara et al, 1987). It can be difficult to separate the beginning of the parametrium from the edge of the cervix and variations in parametrial width with parity have also been described. This measurement should therefore be used as a guide rather than an absolute value of parametrial thickness. Instead, the morphological appearance may be more valuable in determining whether parametrial involvement is present (Browning, 1989a). On the axial scans the normal parametria appear as sharply defined relatively hypoechoic crescents interrupted centrally by the cervix (Fig. 6a) (Dragsted et al, 1987; Yuhara et al, 1987). Loss of the crescentic outline, irregularity of the outline or mass formation are seen with parametrial invasion (Figs 6b and c) although enlargement of the parametria can be seen following cervical biopsy (Browning, 1989b) and other inflammatory conditions. Shift of the cervix to the side of the parametrial involvement may be demonstrated (Yuhara et al, 1987) and paracervical lymph nodes identified as separate low echogenic areas adjacent to the cervix. Changes in the echogenicity of the cervix or parametria have previously been used to detect cervical carcinoma and its spread into the parametria (Dragsted et al, 1987; Yuhara et al, 1987) but in this study were not consistently shown. 677
M Cobby, J Browning, A. Jones E. Whipp and P. Goddard
Figure 6. Axial transrectal ultrasound images of the cervix and parametria. (a) Normal cervix. The cervix lies centrally between the well defined crescentic outlines of the parametria (arrows). Curved arrow: rectum; long straight arrow: bladder, (b) Cervical carcinoma. Enlarged cervix with heterogeneous echogenic pattern and irregular outline. Early parametrial spread is demonstrated on the left (arrows). From the position of the bladder (curved arrow) it is evident that the cervix is shifted to the left, (c) Cervical carcinoma. Extensive left parametrial nvolvement (arrows) at the level of the upper vagina and
With combined radial and sagittal sector scanning, tumour volume can be calculated which is important for treatment planning and prognosis (Shingleton & Orr, 1983). The sagittal image is the optimal view for detecting possible bladder wall involvement and extension into the uterine body. It can also be used to perform transrectal ultrasound-guided transperineal fine needle aspiration of the parametrium when these demonstrate indeterminate features on endosonography (Browning & Halliwell, 1989). This study has shown that endosonography and MRI were marginally more accurate than CT in the local 678
staging or carcinoma of the cervix. They appeared to be particularly useful in distinguishing Stage Ib from Stage lib disease. Computed tomography tended to overstage early carcinomas but was more accurate in assessing larger tumours. Lymph node involvement was detected with equal frequency by both CT and MRI, but MRI was useful in identifying vaginal and bladder wall involvement. Acknowledgements
The authors would like to thank the Trustees of the Bristol MRI Scanner Fund, the Dawn James Trust, the The British Journal of Radiology, September 1990
Staging cervical carcinoma with MRI, CT and endosonography
staff of the Bristol MRI Centre, the Department of Medical Illustration (Bristol Royal Infirmary) and the Frances and Augustus Newman Foundation who gave grants supporting Mark Cobby and James Browning. References AVERETTE, H. E., FORD, J. H., DUDAN, R. C , GIRTANNER, R.
J. H. & BLESSING, J. A., 1980. Results and complications of operative staging in cervical cancer: experience of the Gynecologic Oncolgy Group. Gynecologic Oncology, 9, 90-98. LEE, J. K. T., HEIKEN, J. P., LING, D., GLAZER, H. S., BALFE, D. M., LEVITT, R. G., DIXON, W. T. & MURPHY, W. M., 1984.
Magnetic resonance imaging of abdominal and pelvic lymphadenopathy. Radiology, 153, 181-188. POWELL, M. C , WORTHINGTON, B. S., SOKAL, M., WAISTIE, M.,
E., HOSKINS, W. J. & LUTZ, M. H., 1975. Staging of cervical
BUCKLEY, J. & SYMONDS, E. M., 1986. Magnetic resonance
cancer. Clinical Obstetrics and Gynecology, 18, 215-232. BROWNING, J., 1989a. Endosonography of the parametrium—towards scientific staging. Proceedings of the Silver Jubilee Congress of Obstetrics and Gynaecology, London, July (Royal College of Obstetrics and Gynaecology, London). 1989b. Parametrial changes after cervical biopsy. Proceedings of the Silver Jubilee Congress of Obstetrics and Gynaecology, London, July (Royal College of Obstetrics and Gynaecology, London).
imaging—its application to cervical carcinoma. British Journal of Obstetrics and Gynaecology, 93, 1276-1285. SHINGLETON, H. M. & ORR, J. W., 1983. In Cancer of the
Cervix: Diagnosis and Treatment (Churchill Livingsone, London), pp. 114-117. TOGASHI, K., NISHIMURA, K., SAGOH, T., MINAMI, S., NOMA, S., FUFISAWA, I., NAKANO, Y., KONISHI, J., OZASA, H., KONISHI,
I. & MORI, T., 1989. Carcinoma of the cervix: Staging with MR imaging. Radiology, 171, 245-251. VAN NAGELL, J. R., RODDICK, N. W. & LOWIN, D. M., 1971.
guided transperineal fine needle aspiration of the parametrium. British Journal of Radiology, 62, 739.
The staging of cervical cancer: inevitable discrepancies between clinical staging and pathologic findings. American Journal of Obstretics and Gynaecology, 110, 973-978.
BROWNING, J., GODDARD, P., BAMFORD, D. J. & SANGALA, V.,
VICK, C. W., WALSH, J. W., WHEELOCK, J. B. & BREWER, W.
1990. Tubal ectopic pregnancy: Coincidental magnetic resonance imaging. Journal of Obstetrics and Gynaecology 10, 341-342.
H., 1984. CT of the normal and abnormal parametria in cervical cancer. American Journal of Roentgenology, 143, 597-603.
BROWNING, J. & HALLIWELL, M., 1989. Transrectal ultrasound-
DRAGSTED, J., ASMUSSEN, M., GAMMELGAAD, J. & BOCK, J. E.,
1987. Endoluminal ultrasonic scanning in the staging of cervical carcinomas. Gynecologic Oncology, 28, 1-7. GRECO, A., MASON, P., LEUNG, A. W. L., DISCHE, S., MCINDOE, G. A. J. & ANDERSON, M. C , 1989. Staging of
carcinoma of the uterine cervix: MRI-surgical correlation. Clinical Radiology, 40, 401-405. GRIFFITHS, G. J., CLEMENTS, R. & PEELING, W. B., 1989. The
current status of transrectal ultrasonography in the diagnosis and management of prostatic carcinoma. Clinical Radiology, 40, 337-340. HRICAK, H., LACEY, C. G., SANDLES, L. G., CHANG, Y. C. F.,
WINKLER, M. L. & STERN, J. F., 1988. Invasive cervical
carcinoma: Comparison of MR imaging and surgical findings. Radiology, 166, 623-631. LAGGASSE, L. D., CREASMAN, W. T., SHINGLETON, H. M., FORD,
Vol. 63, No. 753
WALSH,
J.
W.
&
GOPLERUD,
D.
R.,
1981. Prospective
comparison between clinical and CT staging in primary cervical carcimona. American Journal of Roentgenology, 137, 997-1003. WHITLEY, N. O., BRENNER, D. E., FRANCIS, A., VILLASANTU, U., ASSNER, J., WIERNICK, P. H. & WHITLEY, J., 1982.
Computed tomographic evaluation of carcinoma of the cervix. Radiology, 142, 439^46. WILLIAMS, M. P., HUSBAND, J. E., HERON, C. W., CHERRYMAN,
G. R. & KOSLIN, D. B., 1989. Magnetic resonance imaging in recurrent carcinoma of the cervix. British Journal of Radiology, 62, 544-550. YUHARA, A., AKAMATSU, N. & SEKIBA, K.,
1987. Use of
transrectal radial scan ultrasonography in evaluating the extent of uterine cervical cancer. Journal of Clinical Ultrasound, 15, 507-517.
679
VOLUME 63 NUMBER 753
SEPTEMBER 1990
The British Journal of Radiology Magnetic resonance imaging, computed tomography and endosonography in the local staging of carcinoma of the cervix By Mark Cobby, MRCP, FRCR, tJames Browning, FRCS, MRCOG, tAngela Jones, MRCP, FRCR, ^Elizabeth Whipp, MRCP, FRCR and *Paul Goddard, MD, FRCR Departments of Radiodiagnosis, tGynaecology and ^Radiotherapy, Bristol Royal Infirmary, Bristol BS2 8HW (Received January 1990 and in revised form March 1990)
Abstract. Thirty-seven patients with carcinoma of the cervix were prospectively staged by examination under anaesthesia (EUA), transvaginal and transrectal ultrasound, computed tomography (CT) and magnetic resonance imaging (MRI). Pathological correlation was available for 20 patients. In the pathologically staged patients, EUA agreed with the staging in 17, understaging three patients. Endosonography agreed with the staging in 19, CT in 16 and MRI in 18 patients. For the remaining 17 patients, endosonography agreed with the EUA findings in 13, CT in 12 and MRI in 12. This study has shown that endosonography and MRI are more accurate than CT in the local staging of carcinoma of the cervix. Computed tomography was least accurate in staging early tumours and differentiating between Stage Ib and lib disease. Lymph node involvement was detected with equal frequency by both CT and MRI. Magnetic resonance imaging was useful in identifying vaginal and bladder wall involvement and in one patient showed features due to an unsuspected early pregnancy.
Carcinoma of the cervix is the second commonest gynaecological cancer. Its prognosis is determined by the stage of the disease and the volume of tumour at presentation, falling steeply and progressively from Stage I to Stage IV disease with 5-year survival rates of 90% for Stage I and approximately 15% for Stage IV disease (Shingleton & Orr, 1983). The stage of the disease is currently assessed clinically by examination under anaesthesia and influences the type of treatment offered. The findings are classified according to the definitions of the International Federation of Gynaecology and Obstetrics (FIGO) stages (Table I). Clinical assessment of parametrial involvement, spread to the pelvic side wall or regional lymph node involvement can be difficult and result in significant errors in under and over staging (van Nagell et al, 1971; Averette et al, 1975; Lagasse et al, 1980). Until recently computed tomography (CT) was the only imaging technique available for directly assessing parametrial and pelvic wall spread. Its ability to do so accurately though has been disappointing (Whitley et al, 1982). False positive diagnoses of parametrial involvement are common and, in prospective studies, CT has not been able to differentiate Stage Ib from early lib disease reliably (Walsh & Goplerud, 1981). Magnetic resonance imaging (MRI) has.been used in •Author for correspondence. Vol. 63, No. 753
the staging of cervical carcinoma with promising results (Powell et al, 1986; Hricak et al, 1988; Greco et al, 1989; Togashi et al, 1989). The pelvic organs are well shown by MRI with its high contrast resolution and direct multiplanar imaging. The problems with degradation of the images from respiratory motion seen in the upper abdomen do not occur. Cervical carcinoma can be differentiated from the normal cervix by its higher signal intensity on r2-weighted sequences. Parametrial, vaginal wall invasion and lymph node involvement can all be demonstrated and tumour volume calculated. Transrectal ultrasound is being used with increasing success in the diagnosis and staging of prostatic carcinoma (Griffiths et al, 1989). The technique can be Table I. FIGO staging of cervical carcinoma la Ib Ila lib
Microinvasive (only diagnosed histologically) Clinical invasive carcinoma confined to cervix Involvement of vagina (but not lower third) Parametrial involvement but not extending to pelvic side wall
Ilia Involvement of lower third of vagina Illb Extension to pelvic wall and/or obstruction of a ureter IVa Involvement of bladder mucosa or rectum and/or extension beyond the true pelvis IVb Spread to distant organs
673
M Cobby, J Browning, A. Jones E. Whipp and P. Goddard
(b) Figure 1. Endosonographic technique used to obtain axial and sagittal images with (a) the radial rotating transrectal transducer and (b) transvaginal offset sector probe.
adapted to demonstrate the cervix and parametria using a standard radial rotating transducer to produce axially orientated images (Fig. la) (Dragsted et al, 1987; Yuhara et al, 1987). A transvaginal transducer with an off-set sagittal sector probe will demonstrate the cervix and body of the uterus in the longitudinal plane allowing tumour volume to be calculated (Fig. lb). This study prospectively compared the findings of endosonography, CT and MRI in 37 patients with invasive carcinoma of the cervix. Patients and methods Thirty-seven patients with an average age of 38 years (range 24 to 65 years) with pathologically proven invasive carcinoma of the cervix were staged clinically by examination under anaesthesia (EUA) by a consultant radiotherapist or gynaegologist. Endosonographic examination was performed by one of two operators without knowledge of the clinical findings. The CT and MRI images were read independently by two observers, again without knowledge of the findings from the other investigations. Discrepancies in independent interpretation of the CT and MRI images were resolved by consensus. Endosonography was performed using a Bruel and Kjaer 1846 Module with a 1850 radial rotating 5.5 MHz transrectal transducer and an 8537 7 MHz offset sagittal sector transvaginal probe. Measurements made in three
Illb IV
674
Results The clinically and radiologically assigned stages of all 37 patients are shown in Table II and representative Table III. Pathological staging compared with examination under anaesthesia, endosonography, CT and MRI
Table II. Clinical and radiological staging
Ib Ha lib
dimensions were used to calculate tumour volume. The kidneys were examined with a transabdominal transducer to assess whether upper tract dilatation was present. The CT examinations were performed on a Siemens Somatom DRH or a Philips Tomoscan 350 body scanner with contiguous 8 mm slices through the true pelvis and alternate slices up to the aortic bifurcation. Magnetic resonance imaging was performed using a Picker Vista 2055 HP 0.5 tesla MR scanner. Contiguous multislice images 10 mm thick were obtained in the transverse and sagittal planes. Proton density (time to repeat (TR)=1800, time to echo (TE) = 30ms), T2- weighted (TR=1800, TE = 80ms), r, = weighted (TR = 500, TE = 26) and short tau inversion recovery (STIR) (TR=1500, inversion time (TI)=100, TE = 30ms) sequences were used. The total scanning time for each patient was less than 1 h. Eighteen patients subsequently had a radical hysterectomy and lymphadenectomy performed within 2 weeks of the staging investigations. The remaining patients were treated by intracavity and external beam irradiation. Post-mortem findings were available for two patients who died early in the course of their treatment.
Examination under anaesthesia
Endosonography
Computed Magnetic tomography resonance imaging
25 0 8 4 0
22 0 7 8 0
20 0 8 6 3
21 1 5 6 4
Examina- Endosono- Computed Magnetic tion under graphy tomography resonance anaesthesia imaging Correct 17 Overstaged 0 Understaged 3
19 0 1
16 2 2
18 1 1
Accuracy
95%
80%
90%
85%
The British Journal of Radiology, September 1990
Staging cervical carcinoma with MRI, CT and endosonography
Figure 2. Stage Ib cervical carcinoma, (a) Sagittal and (b) transverse r2-weighted magnetic resonance images. The high signal intensity cervical carcinoma (arrow) is completely surrounded by the normal low signal intensity cervical stroma. Normal high signal intensity parametrial tissue is shown on the axial image.
illustrations of the endosonographic and MRI findings are shown in Figs 2 to 6. The accuracy of each staging investigation compared with the pathological staging available for 20 of these patients is shown in Table III. In this latter group, 13 patients with pathologically Stage Ib disease were all correctly staged by EUA and endosonography; one patient was thought to have Stage Ha disease by MRI and two considered to have Stage lib on CT. The two patients with pathologically Stage lib disease were correctly assigned by all investigations. Four patients with Stage Hlb disease were correctly assigned by endosonography, but MRI and CT each down-graded one patient and EUA two patients. The one patient with Stage IV disease was correctly identified by CT and MRI but was thought to be Stage Hlb by endosonography and EUA. For the 17 patients without pathological correlation, endosonography agreed with the EUA findings in 13, CT in 12 and MRI in 12. Computed tomography downgraded one patient; for all other patients where there was discordance between the EUA and imaging findings the staging was up-graded by the radiological investigations. Pelvic lymphadenopathy was seen in three patients by MRI and CT. Bladder wall involvement was well demonstrated by MRI in one patient but was not apparent on the other imaging investigations. The MRI examination of one patient with Stage Ib disease showed an abnormal collection of fluid in the pouch of Douglas, a prominent endomentrial "reaction" and an apparent soft-tissue mass in the uterine Vol. 63, No. 753
fundus which was interpreted as an early pregnancy. At laparotomy a tubal ectopic, but no intrauterine pregnancy was found, which even in retrospect was not visible on the MR images; endosonography performed 2 weeks earlier was normal. The findings in this patient have been described in detail elsewhere (Browning et al, 1990). Discussion This study has shown that all of the techniques could accurately stage carcinoma of the cervix but that endosonography and MRI were marginally better. No comparison with lymphography was made. Relatively few pathologically confirmed Stage lib patients were examined, as in Western practice radiotherapy is the usual treatment for such tumours. For radical hysterectomy to be considered it is important to be able to distinguish between Stage Ib and lib disease reliably. Early parametrial involvement can be difficult to diagnosis with CT (Whitley et al, 1982; Vick et al, 1984). In this study a CT technique of 8 mm contiguous scans was used. Intravenous contrast media coupled with thinner slices and image reconstruction may potentially have improved the diagnostic accuracy. On CT, the cervix appears as a round or oval smooth soft-tissue structure outlined by parametrial fat. The cardinal ligaments are occasionally seen as smooth, laterally tapering, triangular structures extending from the side of the cervix with the distal ureters, surrounded by periureteral fat, lying anterolaterally (Vick et al, 1984). The margins of the normal fat containing parametria are not usually 675
M Cobby, J Browning, A. Jones E. Whipp and P. Goddard
Figure 3. Stage lib cervical carcinoma, (a) Sagittal and (b) transverse r2-weighted images. The large high signal intensity tumour is shown to extend into the left parametrium on the axial image (arrow). Full thickness disruption of the cervical stroma is shown on the right, although definite parametrial extension is not demonstrated.
defined but may contain a variable number of small soft-tissue dots and strands thought to represent parauterine vessels, nerves and lymphatics. These should not exceed 3^4 mm in diameter (Vick et al, 1984) but can be a source of false positive diagnosis of parametria invasion which is usually seen as a mass distorting the
cervical margins or as prominent, sometimes lobulated, parametrial soft-tissue densities. Obliteration of the periureteral fat planes may also be demonstrated (Whitley et al, 1982; Vick et al, 1984). With MRI, cervical carcinoma appears as a high signal intensity mass on r2-weighted images which may
Figure 4. Stage lib cervical carcinoma, (a) Transverse r2-weighted and (b) transverse STIR images. On the r2-weighted image, full thickness disruption of the cervical stroma on the right is shown with subtle signal change in the adjacent parametrium. The extent of parametrial involvement, however, is more easily appreciated on the STIR image. 676
The British Journal of Radiology, September 1990
Staging cervical carcinoma with MRI, CT and endosonography
Figure 5. Stage IV cervical carcinoma, (a) Sagittal proton-density- and (b) r r weighted images. The tumour extends into the body of the uterus, partially obstructing the cervical canal. The upper two-thirds of the vagina and the posterior wall of the bladder are involved. The involved thickened bladder wall and vagina have a high signal intensity on the r2-weighted image. A small quantity of blood is present in the posterior vaginal fornix (arrow).
expand the cervical canal or disrupt the normal low signal intensity cervical stroma (Powell et al, 1986; Hricak et al, 1988). If the low signal intensity ring of cervical stroma is preserved then the tumour is confined to the cervix (Fig. 2) (Hricak et al, 1988; Togashi et al, 1989) but in the presence of full thickness stroma invasion it is not possible to exclude microscopic parametrial spread. As with CT, the margins of the normal parametria are not defined, but the vascular connective tissue produces a medium signal intensity on Tr and protondensity-weighted images and high signal intensity on r2-weighted and STIR images. Parametrial involvement is shown by irregularity of the cervical margins associated with full thickness disruption of the low signal intensity stroma and abnormal signal or mass in the parametrial region (Figs 3 and 4) (Hricak et al, 1988; Togashi et al, 1989). If morphological changes alone are used to diagnose parametrial involvement, MRI is no more accurate than CT (Williams et al, 1989). Magnetic resonance imaging in the sagittal plane is useful for showing vaginal wall involvement, extension into the body of the uterus or forwards to the bladder wall (Williams et al, 1989). These are most easily demonstrated with r2-weighted images which show the internal anatomy of these organs and the presence of tumour as a high signal intensity area (Fig. 5). As in previous studies, MRI showed no distinct advantage over CT in the detection of lymph node metastases using transverse and sagittal images (Lee et Vol. 63, No. 753
al, 1984; Williams et al, 1989) although this may not be so if the coronal plane is also included. Previous endosonographic studies have suggested that the width of the normal parametrium at its mid-point should not exceed 1 cm (Yuhara et al, 1987). It can be difficult to separate the beginning of the parametrium from the edge of the cervix and variations in parametrial width with parity have also been described. This measurement should therefore be used as a guide rather than an absolute value of parametrial thickness. Instead, the morphological appearance may be more valuable in determining whether parametrial involvement is present (Browning, 1989a). On the axial scans the normal parametria appear as sharply defined relatively hypoechoic crescents interrupted centrally by the cervix (Fig. 6a) (Dragsted et al, 1987; Yuhara et al, 1987). Loss of the crescentic outline, irregularity of the outline or mass formation are seen with parametrial invasion (Figs 6b and c) although enlargement of the parametria can be seen following cervical biopsy (Browning, 1989b) and other inflammatory conditions. Shift of the cervix to the side of the parametrial involvement may be demonstrated (Yuhara et al, 1987) and paracervical lymph nodes identified as separate low echogenic areas adjacent to the cervix. Changes in the echogenicity of the cervix or parametria have previously been used to detect cervical carcinoma and its spread into the parametria (Dragsted et al, 1987; Yuhara et al, 1987) but in this study were not consistently shown. 677
M Cobby, J Browning, A. Jones E. Whipp and P. Goddard
Figure 6. Axial transrectal ultrasound images of the cervix and parametria. (a) Normal cervix. The cervix lies centrally between the well defined crescentic outlines of the parametria (arrows). Curved arrow: rectum; long straight arrow: bladder, (b) Cervical carcinoma. Enlarged cervix with heterogeneous echogenic pattern and irregular outline. Early parametrial spread is demonstrated on the left (arrows). From the position of the bladder (curved arrow) it is evident that the cervix is shifted to the left, (c) Cervical carcinoma. Extensive left parametrial nvolvement (arrows) at the level of the upper vagina and
With combined radial and sagittal sector scanning, tumour volume can be calculated which is important for treatment planning and prognosis (Shingleton & Orr, 1983). The sagittal image is the optimal view for detecting possible bladder wall involvement and extension into the uterine body. It can also be used to perform transrectal ultrasound-guided transperineal fine needle aspiration of the parametrium when these demonstrate indeterminate features on endosonography (Browning & Halliwell, 1989). This study has shown that endosonography and MRI were marginally more accurate than CT in the local 678
staging or carcinoma of the cervix. They appeared to be particularly useful in distinguishing Stage Ib from Stage lib disease. Computed tomography tended to overstage early carcinomas but was more accurate in assessing larger tumours. Lymph node involvement was detected with equal frequency by both CT and MRI, but MRI was useful in identifying vaginal and bladder wall involvement. Acknowledgements
The authors would like to thank the Trustees of the Bristol MRI Scanner Fund, the Dawn James Trust, the The British Journal of Radiology, September 1990
Staging cervical carcinoma with MRI, CT and endosonography
staff of the Bristol MRI Centre, the Department of Medical Illustration (Bristol Royal Infirmary) and the Frances and Augustus Newman Foundation who gave grants supporting Mark Cobby and James Browning. References AVERETTE, H. E., FORD, J. H., DUDAN, R. C , GIRTANNER, R.
J. H. & BLESSING, J. A., 1980. Results and complications of operative staging in cervical cancer: experience of the Gynecologic Oncolgy Group. Gynecologic Oncology, 9, 90-98. LEE, J. K. T., HEIKEN, J. P., LING, D., GLAZER, H. S., BALFE, D. M., LEVITT, R. G., DIXON, W. T. & MURPHY, W. M., 1984.
Magnetic resonance imaging of abdominal and pelvic lymphadenopathy. Radiology, 153, 181-188. POWELL, M. C , WORTHINGTON, B. S., SOKAL, M., WAISTIE, M.,
E., HOSKINS, W. J. & LUTZ, M. H., 1975. Staging of cervical
BUCKLEY, J. & SYMONDS, E. M., 1986. Magnetic resonance
cancer. Clinical Obstetrics and Gynecology, 18, 215-232. BROWNING, J., 1989a. Endosonography of the parametrium—towards scientific staging. Proceedings of the Silver Jubilee Congress of Obstetrics and Gynaecology, London, July (Royal College of Obstetrics and Gynaecology, London). 1989b. Parametrial changes after cervical biopsy. Proceedings of the Silver Jubilee Congress of Obstetrics and Gynaecology, London, July (Royal College of Obstetrics and Gynaecology, London).
imaging—its application to cervical carcinoma. British Journal of Obstetrics and Gynaecology, 93, 1276-1285. SHINGLETON, H. M. & ORR, J. W., 1983. In Cancer of the
Cervix: Diagnosis and Treatment (Churchill Livingsone, London), pp. 114-117. TOGASHI, K., NISHIMURA, K., SAGOH, T., MINAMI, S., NOMA, S., FUFISAWA, I., NAKANO, Y., KONISHI, J., OZASA, H., KONISHI,
I. & MORI, T., 1989. Carcinoma of the cervix: Staging with MR imaging. Radiology, 171, 245-251. VAN NAGELL, J. R., RODDICK, N. W. & LOWIN, D. M., 1971.
guided transperineal fine needle aspiration of the parametrium. British Journal of Radiology, 62, 739.
The staging of cervical cancer: inevitable discrepancies between clinical staging and pathologic findings. American Journal of Obstretics and Gynaecology, 110, 973-978.
BROWNING, J., GODDARD, P., BAMFORD, D. J. & SANGALA, V.,
VICK, C. W., WALSH, J. W., WHEELOCK, J. B. & BREWER, W.
1990. Tubal ectopic pregnancy: Coincidental magnetic resonance imaging. Journal of Obstetrics and Gynaecology 10, 341-342.
H., 1984. CT of the normal and abnormal parametria in cervical cancer. American Journal of Roentgenology, 143, 597-603.
BROWNING, J. & HALLIWELL, M., 1989. Transrectal ultrasound-
DRAGSTED, J., ASMUSSEN, M., GAMMELGAAD, J. & BOCK, J. E.,
1987. Endoluminal ultrasonic scanning in the staging of cervical carcinomas. Gynecologic Oncology, 28, 1-7. GRECO, A., MASON, P., LEUNG, A. W. L., DISCHE, S., MCINDOE, G. A. J. & ANDERSON, M. C , 1989. Staging of
carcinoma of the uterine cervix: MRI-surgical correlation. Clinical Radiology, 40, 401-405. GRIFFITHS, G. J., CLEMENTS, R. & PEELING, W. B., 1989. The
current status of transrectal ultrasonography in the diagnosis and management of prostatic carcinoma. Clinical Radiology, 40, 337-340. HRICAK, H., LACEY, C. G., SANDLES, L. G., CHANG, Y. C. F.,
WINKLER, M. L. & STERN, J. F., 1988. Invasive cervical
carcinoma: Comparison of MR imaging and surgical findings. Radiology, 166, 623-631. LAGGASSE, L. D., CREASMAN, W. T., SHINGLETON, H. M., FORD,
Vol. 63, No. 753
WALSH,
J.
W.
&
GOPLERUD,
D.
R.,
1981. Prospective
comparison between clinical and CT staging in primary cervical carcimona. American Journal of Roentgenology, 137, 997-1003. WHITLEY, N. O., BRENNER, D. E., FRANCIS, A., VILLASANTU, U., ASSNER, J., WIERNICK, P. H. & WHITLEY, J., 1982.
Computed tomographic evaluation of carcinoma of the cervix. Radiology, 142, 439^46. WILLIAMS, M. P., HUSBAND, J. E., HERON, C. W., CHERRYMAN,
G. R. & KOSLIN, D. B., 1989. Magnetic resonance imaging in recurrent carcinoma of the cervix. British Journal of Radiology, 62, 544-550. YUHARA, A., AKAMATSU, N. & SEKIBA, K.,
1987. Use of
transrectal radial scan ultrasonography in evaluating the extent of uterine cervical cancer. Journal of Clinical Ultrasound, 15, 507-517.
679
