6 Magnetic resonance imaging in clinical cervical cancer: pretherapeutic tumour volumetry

6 Magnetic resonance imaging in clinical cervical cancer: pretherapeutic tumour volumetry H. M. H. HOFMANN F. E B N E R J. H A A S R. EINSPIELER E. JUSTICH M. LAHOUSEN H. PICKEL E. BURGHARDT

Magnetic resonance spectroscopy is an established method in solid-body physics and analytical chemistry. Based on the phenomenon of magnetic nuclear resonance, it produces two- and three-dimensional images. Damadian (1971) introduced magnetic resonance imaging (MRI) as an imaging technique. The advantages of MRI over ultrasound and computed tomography (CT) are the excellent soft-tissue contrast and the topographic anatomic imaging in the standard planes of the human body. To date there have been no reports of health hazards associated with modern MRI--which uses field strengths of less than 2 teslas (Rinck et al, 1986; Vogl et al, 1986).

CLASSIFICATION OF CERVICAL CANCER Cervical cancer patients are classified before treatment so that results can be compared afterwards. The conventional FIGO classification (Chapter 2) is based on palpation and is subjective. Staging would be better if it were based on the findings at surgery, but the FIGO classification persists because it is the only one possible for the radiotherapist. A staging laparotomy before radiotherapy gives the radiologist a better idea of the extent of the disease. Lymph node and, in some cases, parametrial involvement can be assessed. However, staging procedures rely on biopsies that evaluate only gross abnormalities. Small tumour deposits, especially micrometastases in the parametria, and unenlarged nodes can escape detection. Additionally, a staging procedure can only estimate the important factor of tumour size. Bailli~re's ClinicalObstetrics and Gynaecology--Vol. 2, No. 4, December 1988

789

790

H.M.H.

HOFMANN ET AL

Tumour size is a major prognostic factor and thus the best basis for a classification of cervical cancer. The more precise the measurement the more reliable the classification. Tumour size and volume can be measured exactly in the surgical specimen. Classification by size is far superior to clinical staging, and it produces results that can be assessed objectively (Figure 1). The disadvantage of classification by size is that it can be done only after surgery and can thus evaluate only surgical results. PURPOSE Attempts to measure pretherapeutic tumour size by high-resolution ultrasound or by CT have failed. We hoped that MRI, which allows better differentiation and contrast of the tumour, would provide better results. The differentiation between tumour and the healthy surrounding tissue does not guarantee that an imaging technique can demonstrate the entire extent or the true volume of the tumour. Infiltrates around the tumour can cause overestimation of its volume, and partial volume effects--which depend on the turnout size and the chosen slice thickness--can give rise to estimates which are too high or too low. We addressed the question: how precisely can MRI preoperatively determine the volume of a cervical cancer? The goal was precise pretherapeutic measurements for classifying cervical cancer. A method had to be found to evaluate the MRI measurements and to compare the actual turnout spread with the MRI contours. Since 1971 surgical specimens obtained at radical hysterectomy for cervical cancer have been processed as serial giant sections (Burghardt and Pickel, 1978). This technique determines exactly both the size and volume of the tumour, its contours, and its spread. Except for errors due to fixation, the giant sections constitute the ideal reference for tumormetry by imaging techniques. Further comparisons can be drawn by MRI of the unfixed surgical specimen. METHODS MRI technique We used a superconducting MRI unit (Gyroscan, Philips AG, Eindhoven, Netherlands) with a field strength of 1.5 T. Only spin-echo and partialsaturation-spin-echo pulse sequences were used. Pulse parameters We used spin-echo sequences with repetition times of 400-600 ms and echo times of 20-30 ms (mostly T1 weighting), spin-echo sequences with repetition times of 2000-2500 ms and echo times of 20-40 ms (with proton and slight T2 weighting) or of 60-100 ms (mostly Ta weighting).

PRETHERAPEUTIC TUMOUR VOLUMETRY BY MRI

791

Scans were done in transverse and sagittal planes in all patients. Coronal scans were done only in isolated cases. The slice thickness was initially 8-10 mm, then 5.3 mm or 6 mm. Scans were done at 0.5-1.0 mm intervals. The picture matrix comprised 256 x 256 points or 256 points along the x-axis and 128 points along the y-axis. Depending on the patient's size, the field of view was 35 cm x 35 cm or 40 c m x 40 cm. For tumour volumetric analysis of cervical cancers, only T2-weighted pictures in transverse planes were used. Just before MRI the patients received glucagon (0.014 mg kg -1 bodyweight) by vein to decrease intestinal peristalsis. There was no other preparation. Most scans were done with the patient in a prone position. An average of 40 min elapsed between the start of positioning and the end of the study. MRI anatomy of the female pelvis Familiarity with MRI anatomy of the internal genitalia is a prerequisite to the recognition of abnormalities (McCarthy et al, 1986; Heiken and Lee, 1988). Sagittal images demonstrate the position of the uterus in relation to the bladder, the rectum, and the vaginal fornices. T2-weighted sagittal and transverse planes show the three-layered uterine body. A bright, hyperintense outer zone, a darker intermediate zone, and a brighter inner zone can be distinguished. The darkness of the intermediate zone is attributed to the high fibre density and to a different fibre direction in the spirally-

Figure 1. Sagittal section of the pelvis with the uterus in R V F L (T~-wcightcd image), l, Cervical canal and endocervix. 2. Intermediate zone. 3. Outer mvometriai laver. V - vagina. U = bladder.

792

n . M. H. HOFMANN ET AL

arranged myometrial muscle fibres. The hyperintense inner zone represents the varying width of the endometrium (depending on the age of the patient and her menstrual cycle) and secretions in the uterine cavity. As visualized by MRI, the endometrium of a fertile woman in the proliferative phase is 1-3 mm thick and reaches 5-7 mm in the secretory phase. In the secretory phase the uterine volume is slightly increased, the junctional zone is better defined, and the signal intensity of the myometrium is increased. The cervix differs from the body of the uterus in that usually only two zones can be distinguished. Increased signal density along the cervical canal is reflected as a bright line in the MR image. The cervical connective tissue, because of its high connective fibre content, has a low signal intensity (Figure 2). Normal-sized ovaries can be demonstrated in 60-70% of women. Having found the fossa ovarica, the adnexae can be discerned at moderate signal intensity in Ta-weighted sequences and at increased intensity in Tz-weighted sequences. The vagina is well demonstrated in sagittal and transverse planes. Under abnormal conditions such as colpitis, the walls can be thickened and hyperintense. Enlarged lymph nodes in the parametria or at the pelvic wall are seen as nodular structures of different sizes and signal intensities. It is not possible to distinguish inflammatory or reactive enlargement from metastatic involvement.

Figure 2. Normal cervical anatomy in a transverse plane (T--weighted image). I. Cervix (low signal intensity). 2. Cervical canal. 3. Internal obturator muscle. 4. ParametriaI vessels (high intensity). U = bladder. SB = small bowel.

PRETHERAPEUTIC TUMOUR VOLUMETRY BY MRI

793

Tumour assessment by MRI

In comparison with the normal cervix, cervical cancers show a markedly increased signal intensity in the T2-weighted spin-echo-pulse sequences (Togashi et al, 1986; Worthington et al, 1986; Hricak et al, 1988). It is not difficult to demarcate small tumours by their hyperintense signals (Figures 3 and 4). If the turnout extends beyond the uterine isthmus it can be difficult to

Figure 3(a)

Figure 3(b)

794

H.M.H.

HOFMANN ET AL

Figure3(c) Figure 3. (a) Sagittal plane showing a small, hyperintense cervical tumour (T2-weighted image). U = bladder. R = rectum. (b) Coronal section of the surgical specimen (T2-weighted image). T = tumour. O = ovary. C = cavum uteri of the uterus arcuatus. (c) Corresponding giant section of the same plane as (b),

Figure 4. Sagittal section showing a hyperintense turnout (T) in the anterior cervix (T1weighted image). Arrows indicate the connective tissue phme between bladder wall and tumour.

PRETHERAPEUTIC TUMOUR VOLUMETRY BY MRI

795

distinguish from the myometrium of the uterine body, since both are hyperintense. Direct extension of the hyperintense lesion from the ectocervix to the vaginal fornix (Figure 5), the paracolpium, or the parametrium is considered infiltration of these structures. An exophytic tumour is imaged as a hyperintense signal inside the vaginal walls (Figure 6). When a tumour abuts an organ, MRI cannot state whether the tumour only abuts or whether it has infiltrated the wall of the organ. If the tumour encroaches on the bladder or the rectum, or if the layer demarcating these organs is missing, the findings are considered indicative of infiltration only if the mucosa of these organs is thickened. Extension of the tumour to the pelvic wall (Figure 7), and extension to pelvic soft tissue or bone, can be demonstrated unequivocally (Figure 8). Enlarged, positive lymph nodes are seen as nodular densities in the corresponding node regions.

Tumour volumetry by MRI and by measurements of giant sections MRI images were displayed on the monitor and photographed in a proper window setting. The tumour was outlined on the hard copy by an experienced radiologist and by a gynaecologist. The photos were digitized and the tumour area in the transverse sections was calculated taking into account image demagnification.

Figure 5. Sagittal section (Tl-weighted image) showing a hyperintense, exophytic (E) cervical tumour (T). Arrow indicates missing demarcation of the tumour from the bladder (U). V = vagina.

796

I-I. M. H. HOFMANN ET AL

Figure 6. Transverse section at the level of the cervix (C) (mixed sequence), U = bladder, T = tumour. P = spread to the parametrium.

Figure 7, Transverse section at the level of the cervix (C) (mixed sequence) showing invasion of the bladder (U) and extension of the tumour (T, arrows) to the pelvic wall. M - piriform muscle (open arrow),

PRETHERAPEUTIC TUMOUR VOLUMETRY BY MRI

797

Figure 8. Transverse section at the level of the uterine fundus showing turnout infiltration of the corpus reaching the fundus. The arrow points to continuous turnout extension of the turnour (T) to the pelvic bone and the right acetabulum (A).

Morphometric analysis was done with a semi-automatic image analyser consisting of a microcomputer and a digitizer tablet. Using stereological methods, tumour volume was calculated by adding the approximated volumes of the tumour slices. Several geometric models were used to obtain reliable section volumes; each slice was composed of two adjacent sections using geometric bodies. Immediately following its removal, and before fixation, the surgical specimen was remeasured by MRI with the same method as used preoperatively. The fixed specimen was then processed as serial giant sections (Burghardt and Pickel, 1978). On each section, the perimeter of both the tumour and the cervix were outlined by an experienced pathologist. The contours of the respective structures were digitized and the areas calculated. MATERIAL

Between February 1987 and January 1988 all patients with clinical cervical cancer underwent MRI studies. To compare the volumes obtained by in vivo MRI, by MRI of the surgical specimen, and by measurements of the histological giant section, 25 patients scheduled for radical hysterectomy for Stage Ib or IIb cervical cancer were studied. Thirteen patients with Stage IIb to IV disease who did not undergo surgery were studied to assess the spread of the tumour and its behaviour towards the neighbouring organs, and for classification by volume.

798

H . M . H . HOFMANN ET AL

RESULTS Table 1 summarizes the m e a s u r e m e n t s in the 25 patients who underwent surgery. By preoperative M R I the volume of the largest tumour exceeded 750 cm 3. The discrepancy to the volume determined from the giant sections was a result of the size of the specimen, which could not be processed in toto. Table 1. Clinical (FIGO) stage and tumour volumetry (preoperative MRI, surgical specimen MRI, histological giant section) in 25 cases. MRI tumour volume MRI tumour volume Tumour volume Clinical (preoperative) (surgical specimen) (giant section) stage (mm3) (mm3) (mm3) Ib IIb IIb IIIb IIb Ib Ib Ib Ib IIb IIb Ib Ib IIb Ib Ib Ib lib IIb Ib Ib IIb Ib IIb Ib

15600 20053 759563 49047 21016 2204 3600 3008 19500 4400 60141 74600 7 876 53600 7781 66915 12411 77640 22845 32 t00 3536 17675

4347 4700 3258 5565 39800* 41717" 8970 69000 11199 7300 15547 36971

14916 19584 458649 46700 29082 900 691 3036 2780 18006 2423 2951 52133 64000 7 300 49864 7333 62800 11280 5 623 74926 15752 29000 1956 17190

*The exophytic part of the tumour was not measurable. The M R I volumes and the volumes determined from the giant sections showed a statistically significant correlation with a coefficient of 0.983. E v e n the extremely large turnout (case 3) fell into the 95% confidence interval. Only four cases lay outside it. All data were prospective and coded (Figure 9). The volumes determined by M R I of 13 surgical specimens also correlated significantly, even though in two cases the exophytic part of the tumour could not be measured since it could not be distinguished by M R I from the vaginal wall (Figure 10). The M R I volumes were, on average, 8.8% larger than the volumes determined from the giant sections. The shrinking of the specimen during fixation contributed to this difference. Table 2 stratifies the volume differences.

799

PRETHERAPEUTIC TUMOUR VOLUMETRY BY MR[

8°i

'

'

e..-".

701 60-

Q-"

"'"

0

'~

50.

~

40.

o

>

30, 20.

10. -10

10

20

30

40

50

60

70

80

Histomorphometric Volume (cc)

Figure 9. Correlation of MRI volume with histomorphometric volume in 22 cervical cancers. (In three cases MRI studies were carried out on the surgical specimen only.) y = 1.066x + 0.543, R-squared = 0.983, 8o

g

70.

"i

60. 50

40 30

-~ >

20 lO

o 10

20

30 40 50 Histomorphometric Volume (cc)

60

70

Figure 10. Correlation of MRI turnout volume of the surgical specimen with that of the histomorphometric volume, y = 0.826x + 2.663, R-squared = 0.894. Table 2. Comparison of preoperative MRI tumour volume with volume determined from giant sections. The MRI volumes were an average of 8.8% larger than the histological volumes. MRI < Histology MRI > Histology under 5% 5-10% over 10% Total

1 5 7 9 22

800

H. M. H. HOFMANN ET AL

In most cases the volumes calculated by the five geometric models showed similar results. Irregular geometric models were approximated by spherical and cylindrical models since these corresponded best to the M R image of the turnout. In only one case was the M R I t u m o u r volume smaller than the volume determined from the giant sections. Four of the nine turnouts with a discrepancy of less than 10% were smaller than 5.0 cm 3. The five mathematical models correlated significantly with each other, even when the extremely large turnout (case 3) had been excluded.

Classification by volume For comparison with the F I G O classification, the 25 tumours that underwent surgery were classified by volume (Table 3); Table 4 includes the 13 patients who did not have surgery. In both tables a scattered distribution a m o n g the F I G O stages is evident. For example, one Stage I I b t u m o u r was in the smallest volume class ( < 5 cm 3) and three Stage Ib tumours were in the 50-100 cm 3 class.

Table 3. Classification by MRI turnout volume and FIGO stage in 25 patients undergoing surgery. VoLume (MRI) (cc) Stage Ib Stage IIb Stage IIIb <5 5 2 0 5-10 3 0 0 10-30 2 5 0 30-50 1 0 1 50-100 3 2 0 >100 0 1 0 Total 14 10 1 Table 4. MRI tumour volume by FIGO classification (n = 38). Volume (MRI) (cc) Stage Ib Stage lib Stage IIIb Stage IVb <5 5 2 1 0 5-10 3 0 0 0 10-30 2 7 1 0 30-50 1 0 3 0 50-100 3 4 0 1 > 100 0 1 2 2 Total 14 14 7 3

DISCUSSION M R I studies of cervical cancer have been c o m p a r e d with clinical stage (Walsh and Goplerud, 1981; Togashi et al, 1986; Hricak et al, 1988; Z a p f et al, 1988). Correlations between M R I findings and precise biometric data have not been reported.

PRETHERAPEUTIC TUMOUR VOLUMETRY BY MRI

801

Experimental studies have shown that MRI can distinguish tumour tissue from normal tissue only at the appropriate pulse parameters (Weisman and Bennett, 1972; Hollis et al, 1973; Hazlewood et al, 1974; Inch et al, 1974; Saryan et al, 1974; Fruchter et al, 1978; Moore et al, 1983; Lohmann and Haase, 1987). Tumour tissue is characterized by a higher water content. Optimal imaging of cervical tumours and their demarcation from normal cervical tissue requires spin-echo sequences with long repetition times and long echo times. Limiting factors are the prolonged examination time of a series, the increasing influence of motion artefacts, and the worsening signal-to-noise ratio if echo times are too long. In our experience, and that of others (Togashi et al, 1986; Zapf et al, 1987, 1988; Hricak et al, 1988), exclusively Tl-weighted or only slightly Tz-weighted sequences generally cannot discern tumour tissue. Because of fundamental differences between the techniques, MRI results cannot be correlated with those of ultrasound or CT. Since the introduction of the clinical classification of cervical cancer, ways have been sought to improve it. Neither ultrasound nor CT was able to delineate an invasive cervical cancer precisely enough to determine its spread, size, or volume. Our results show that MRI can discern even small cancers and can measure their volume. The in vivo MRI volume of the tumour correlated significantly with the actual volume determined from histological giant sections. It has long been an open question whether the results of radiotherapy can be compared with those of surgery. Surgery produces a specimen that can be measured (Burghardt and Pickel, 1978); with radiotherapy this is not possible, M R / t u r n o u t volumetry before treatment will make it possible to compare objectively the results of primary radiotherapy and surgery. Tumour volume is presently the most significant prognostic factor. To place treatment results on an objective footing every patient with invasive cervical cancer should undergo MRI before treatment. SUMMARY

MRI can define the spread, size, and volume of clinical cervical cancers. Appropriate pulse sequences and slice thicknesses are necessary. Twenty-five patients underwent MRI tumour volumetry before radical hysterectomy. The volume obtained by MRI was compared with that obtained from the histological giant sections; the volumes agreed at a statistically significant correlation coefficient of 0.983. The volumes obtained by MRI of 13 unfixed surgical specimens correlated with their histological volumes with a statistically significant coefficient of 0.894. Tumour volumes were compared with the respective clinical stages. Clinical stage did not correlate with tumour volume. Three very large tumours were in clinical Stage Ib. Tumour size is a major prognostic factor, can be measured easily, and, as the basis for classification, is superior to FIGO staging. MRI can measure tumour volume before treatment.

802

H . M . H . HOFMANN ET AL

REFERENCES Burghardt E & Pickel H (1978) Local spread and lymph node involvement in cervical cancer. Obstetrics and Gynecology 52: 138-145. Damadian R (1971) Tumor detection by nuclear magnetic resonance. Science 171: 1151-1153. Fruchter RG, Goldsmith M, Boyce JG, Nicastri AD, Koutcher J & Damadian R (1978) Nuclear magnetic resonance properties of gynecologicaltissues. Gynecologic Oncology 6: 243-255. Hazlewood CF, Clevel G & Medina D (1974) Relationship between hydration and proton nuclear magnetic resonance relaxation times in tissues of tumor-bearing and non-tumorbearing mice. Implications for cancer detection. Journal of the National Cancer Institute 52: 184%1853. Heiken JP & Lee JKT (1988) MR Imaging of the pelvis. Radiology 166: 11-16. Hollis DP, Economou JS, Parks LC, Eggleston JC, Saryan LA & Czeisler JL (1973) Nuclear magnetic resonance studies of several experimental and human malignant tumors. Cancer Research 33: 2156-2160. Hricak H, Lacey GC, Sandles GL, Chang YCF, Winkler ML & Stern LJ (1988) Invasive cervical carcinoma: comparison of MR imaging and surgical findings. Radiology 166: 623-631. Inch WR, McCredie J, Knispel RR, Thompson RT & Pintar MM (1974) Water content and proton spin relaxation time of neoplastic and non-neoplastic tissues from mice and humans. Journal of the National Cancer Institute 52: 353-356. Lohmann W & Haase A (1987) Kernspintomographie in der Geburtshilfe und Gyn~ikologie. Gyniikologe 20: 197-200. McCarthy S, Tauber C & Gore J (1986) Female pelvic anatomy. MR assessment of variations during the menstrual cycle and with use of oral contraceptives. Radiology 160: 11%123. Moore FA, Todd LE & Lara de Jimenez R (1983) Ti relaxation times in vivo of human normal pelvic tissues and gynecological malignancies. Physiological Chemistry and Physics and Medical Nuclear Magnetic Resonance 15: 23--25. Rinck PA, Pedersen SP & Muller RN (1986) Sicherheit yon Patienten und Personal. In Rinck PA, Pedersen SP & Muller RN (eds) Magnetresonanz Imaging und Spectroskopie in der Medizin, pp 35-36. Stuttgart, New York: Georg Thieme Verlag. Saryan LA, Hollis DP, Economou JS & Eggleston JC (1974) Brief communication. Nuclear magnetic resonance studies of cancer. IV. Correlation of water content with tissue relaxation times. Journal of the National Cancer Institute 52: 599-602. Togashi K, Nishimura K, Itoh K, Fujisawa I, Asato R, Nakano Y, Itoh H, Torizuka K, Ozasa H & Mori T (1986) Uterine cervical cancer: assessment with high-field MR imaging. Radiology 160: 431-435. Vogl Th, Lissner J, Seiderer M, Krimmel K & Sandner H (1986) Einflul3der bei der Kernspintomographie verwendeten Feldarten auf die Kern-und Oberfl~ichentemperatur im menschlichen Organismus. Fortschritte auf dem Gebiete der Rdntgenstrahlen und der Nuklearmedizin 144: 591-596. Walsh JW & Goplerud DR (1981) Prospective comparison between clinical and CT staging in primary cervical carcinoma. American Journal of Radiology 137: 997-1003. Weisman ID & Bennett LH (1972) Recognition of cancer in vivo by nuclear magnetic resonance. Science 178: 1288-1290. Worthington LJ, Balfe MD, Lee TKJ, Gersell JD, Heiken PJ, Ling D, Glazer SH, Jacobs JA, Kao MS & McClennan LB (1986) Uterine neoplasms MR imaging. Radiology 159: 725-730. Zapf S, Halbsguth A, Schweden F, Klose K, Lochner B, Beck Th, Friese K & Kreienberg R (1987) Zum Problem des pdttherapeutischen Stagings des carcinoma coli uteri. Geburtshilfe und Frauenheilkunde 47: 838-842. Zapf S, Halbsguth A, Schweden F, Klose K, Lochner B, Beck Th, Friese K & Kreienberg R (1988) Magnetresonanztomographie in tier Diagnostik des Kollumkarzinoms. Computertomographische und histologische Korrelation. Fortschritte auf dem Gebiete der ROntgenstrahlen und der Nuklearmedizin 148: 34-37.