British Journal of Radiology 74 (2001),805-810 © 2001 The British Institute of Radiology
Detection of nasopharyngeal carcinoma: fast short time inversion recovery images compared with fat suppression, contrast enhanced T1 weighted spin echo images
S Yamamoto, MD
1
H Takano, MD
1
K Motoori, MD
1
T Ueda, MD
1
M Ikeda, MD
1
S Kimura, MD
2
T Uno, MD
1
S Yasuda, MD
1
H Ito, MD
1
R Hara, MD
1 and
K Isobe, MD
1
1Department of Radiology, Chiba University, 1-8-1 Inohana, Chuou-ku, Chiba City, Chiba 260-8670 and 2Department of Radiology, Inoue Memorial Hospital, 1-16 Shindencho, Chuou-ku, Chiba City, Japan
 |
Abstract
|
|---|
The aim was to compare fast short time inversion recovery (FSTIR) images and fat suppression, contrast enhanced T1 weighted (FSCE T1W) spin echo images in the diagnosis of nasopharyngeal carcinoma (NPC). 102 MR studies were obtained with a 1.0 T or a 1.5 T system in 28 patients with NPC. The MR studies comprised both FSTIR and FSCE T1W images. FSTIR and FSCE T1W images were compared for detection of NPC by means of a receiver operating characteristic (ROC) analysis. The areas under the ROC curves of FSTIR and FSCE T1W images showed no statistical difference (0.87 vs 0.87). There was also no statistical difference in the sensitivity, specificity and accuracy of each sequence (0.74 vs 0.77, 0.81 vs 0.77 and 0.79 vs 0.77, respectively). Both sequences had the same performance for detection of NPC. FSTIR is as useful as FSCE T1W images, especially in the detection of recurrent tumours, but without the cost of contrast medium.
|
Introduction
|
|---|
In head and neck imaging, T2 weighted fast spin echo (FSE) sequences have the advantage ofreduced cost, short scanning time and no requirement for contrast medium. However, fat still remains as a relatively high signal on FSE sequences, potentially reducing tumour conspicuity. This is a particular problem at the skull base owing to the abundance of fat in the bone marrow and parapharyngeal space. Fat suppression techniques have therefore been advocated for T2 weighted as well as T1 weighted imaging of tumours of the head and neck [1–6]. However, chemical shift fat suppression techniques also have several disadvantages. They are prone to magnetic susceptibility artefacts from air–tissue interfaces, which are particularly a problem around the nasal cavity, paranasal sinuses and petrous temporal bones. The short time inversion recovery (STIR) sequence is believed to have an equivalent contrast to the fat suppression T2 weighted image, but with less susceptibility to artefacts. Conventional STIR requires a longer scanning time and is therefore inappropriate in modern clinical practice. Nowadays, fast short time inversion recovery (FSTIR) can be performed in a short time period, almost as quickly as fat suppression, contrast enhanced T1 weighted (FSCE T1W) spin echo imaging.
The purpose of this study was to determine whether FSTIR could replace FSCE T1W imaging and whether it could remove the need for contrast medium in the detection of nasopharyngeal carcinoma (NPC).
|
Materials and methods
|
|---|
MR images were retrospectively reviewed in 28 patients (19 males and 9 females; age range 27–70 years, mean 56 years) with either newly diagnosed NPC, residual tumour after therapy or tumour recurrence. These studies contained at least transverse images of both FSTIR and FSCE T1W sequences. Patients were initially subcategorized into four T-stage subgroups, as outlined in Table 1
. Stage T1 was found in 11 patients, T2 in 7 patients, T3 in 4 patients and T4 in 6 patients.
Serial studies were performed on each patient, with a total of 102 studies with at least both sequences (Table 2).
Images were obtained on a 1.0 T system (Magnetom Impact; Siemens, Erlangen, Germany) or a 1.5 T system (Sigma; GE Medical Systems, Milwaukee, WI) using a head coil. FSTIR images (Figures 1a and 2a) were obtained in transverse planes with an inversion time of 150 ms, repetition time of 3500–4000 ms, echo time of 28–60 ms, echo train length of 8–11, field-of-view of 22 cm x 22 cm, slice thickness of 3–5 mm, interslice gap of 1–1.5 mm, acquisition matrix of 256 x 192 and two signal averages. FSCE T1W images were obtained in transverse planes with a repetition time of 400–484 ms, echo time of 14–15 ms, field-of-view of 22 cm x 22 cm, slice thickness of 3–5 mm, interslice gap of 1–1.5 mm, acquisition matrix of 256 x 192 and two signal averages, with a bolus injection of 0.2 mmol kg-1 body weight of gadopentetate dimeglumine (Magnevist; Japan-Schering, Osaka, Japan).
View larger version (81K):
[in this window]
[in a new window]
|
Figure 1. Nasopharyngeal carcinoma in a 62-year-old male before radiation therapy. Transverse MR images of the nasopharynx. Tumour has spread on the posterior wall of the nasopharynx. There is a high intensity area on fast short time inversion recovery (FSTIR) images (a) and an enhanced area on fat suppression, contrast enhanced T1 weighted (FSCE T1W) images (b).
|
|
View larger version (79K):
[in this window]
[in a new window]
|
Figure 2. Nasopharyngeal carcinoma in a 40-year-old female before radiation therapy. Transverse MR images of the nasopharyx show a soft tissue mass on the right nasopharynx. Tumour extends into the anterior vertebral muscle. There is a high intensity area on fast short time inversion recovery (FSTIR) images (a) and an enhanced area on fat suppression, contrast enhanced T1 weighted (FSCE T1W) images (b).
|
|
Comparison was made with results from both endoscopy and needle biopsy, performed within a week before or after the MRI study. 26 results were positive, including 3 recurrences, and 76 results were negative. These numbers were used as actual positive and actual negative data in receiver operating characteristic (ROC) analysis.
ROC analysis was performed to compare FSTIR with FSCE T1W imaging. The FSTIR and FSCE T1W images were analysed separately, at least 2 weeks apart. All MR images were independently reviewed by four experienced radiologists (HT, KM, TU, MI). Data from the results of this scoring were entered into the ROC analysis [7–9]. The reviewers scored each image for the presence or absence of NPC and assigned confidence levels to their observations (1, definitely or almost definitely absent; 2, probably absent; 3, possibly present; 4, probably present; 5, definitely or almost definitely present). On FSTIR, mass lesions with intensities between that of mucosa and muscle were assigned as tumour. On FSCE T1W imaging, mass lesions with contrast enhancement were assigned as tumour.
For each sequence evaluation, a binomial ROC curve was drawn in accordance with each observer's confidence rating data by means of maximum likelihood estimation.
Differences between ROC curves of individual reviewers were tested for statistical significance. These evaluations were done by means of the ROCKIT 0.9 B program (freely available, courtesy of Metz CE, University of Chicago, 1998). After drawing each ROC curve, the areas under the ROC curves (Az) were calculated for each observer and MR pulse sequence. Composite ROC curves were used to represent the performance of the four readers as a group and were calculated by averaging the binomial parameter values of the individual curves. Az areas were also evaluated. Sensitivity, specificity and accuracy were calculated for each sequence. The number of lesions correctly determined as probably present (a score of 4 in the ROC analysis) or definitely present (a score of 5) by each reviewer was regarded as the number of correctly diagnosed lesions.
|
Results
|
|---|
ROC analysis
The Az area for the four reviewers for each FSTIR and FSCE T1W image were 0.9 vs 0.91, 0.87 vs 0.87, 0.88 vs 0.89 and 0.92 vs 0.89, respectively (Table 3). The difference between FSTIR and FSCE T1W imaging did not achieve statistical significance in any observer's curve (p>0.5). The composite ROC curves were also drawn on the basis of pooled data from the four reviewers for each sequence (Figure 3). The mean areas under the curves showed no significant difference (mean Az at FSTIR 0.87; mean Az at FSCE T1W imaging 0.87).
View this table:
[in this window]
[in a new window]
|
Table 3. Area under the alternative free response receiver operating characteristic curve (Az) for each observer and MR pulse sequence Az value
|
|
View larger version (17K):
[in this window]
[in a new window]
|
Figure 3. Composite receiver operating characteristic curves from pooled data from the four reviewers for each sequence. FSCE T1WI, fat suppression, contrast enhanced T1 weighted imaging; FSTIR, fast short time inversion recovery. Az, area under curve.
|
|
Sensitivity, specificity and accuracy
The mean sensitivities, specificities and accuracies of FSTIR and FSCE T1W imaging were 0.74 vs 0.76, 0.79 vs 0.76 and 0.77 vs 0.76, respectively; there were no significant differences (p>0.05) (Table 4).
Three recurrent tumours were found, but recurrence was indicated before biopsy in only one case. In that case, on a MRI study 22 months after radiation therapy, three out of four reviewers pointed out the recurrent tumour on both sequences, although the biopsy result was negative (Figure 4). 10 months after this MRI the following biopsy was positive.
View larger version (74K):
[in this window]
[in a new window]
|
Figure 4. Nasopharyngeal carcinoma in a 52-year-old male just after radiation therapy, needle biopsy having shown no residual tumour. Transverse MR images of the nasopharynx. Mucosa of the left lateral wall shows a high intensity area (long arrow) on fast short time inversion recovery (FSTIR) images (a) and an enhanced area (short arrow) on fat suppression, contrast enhanced T1 weighted (FSCE T1W) images (b).
|
|
|
Discussion
|
|---|
The usefulness of MRI in the evaluation of NPC has been established [2, 10–13]. Diagnostic evaluation of NPC by MRI commonly includes both T1 weighted and T2 weighted spin echo sequences in the axial plane and either the coronal or sagittal plane. Contrast enhanced T1 weighted images are useful both in tumour detection and the evaluation of tumour extent. Fat in the head and neck region is a particular problem on T1 and T2 weighted images. Fat lowers the detection rate of tumours on T2 weighted images. This problem can be solved by fat suppression techniques [14], several of which are currently available [14, 15]. One such technique is STIR. In this technique, the fat signal is nulled by virtue of its T1 relaxation time. Specifically, the signal from that tissue is nulled when the inversion time is set to 0.69 of the T1 of the tissue. A recent study compared fat suppression T2 weighted, inversion recovery, fast spin echo (T2W-IR-FSE) sequences with T2 weighted, fast spin echo sequences for conspicuity of tumour margin and tumour extent [13]. The authors concluded that T2W-IR-FSE improved conspicuity in 48% of tumours and provided equal conspicuity in 22% of tumours. Conspicuity of small tumours with long T2 relaxation times that were surrounded by fat was improved most by the T2W-IR-FSE sequence. Another method is chemical saturation. There are some reports of the use of this method to evaluate head and neck lesions. However, chemical saturation has a problem with artefacts resulting in inhomogeneous fat suppression [14, 15]. Such artefacts are most pronounced at air–tissue interfaces; that is, where the imaging volume changes abruptly, such as at the junction of the skull base. The artefacts are thought to arise from resonant frequency shifts due to focal magnetic field inhomogeneity at the air–tissue interfaces, resulting in partial local failure of the pre-satuation pulse. Fat suppression failures are seen in areas of changing tissue geometry such as the nasopharynx and orbital region with the frequency selective pre-saturation technique, and thus these regions are most troublesome and could lead to difficulties in clinical diagnosis [16].
We conclude that both fat suppression T1 weighted imaging and FSTIR seem to be useful methods for evaluating head and neck tumours.
FSTIR reduces the cost and shortens the study time, since no contrast medium is used. One study compared a T2 weighted fat suppression, turbo spin echo with fat suppression, contrast enhanced T1 weighted, spin echo MRI in NPC [12]. The authors concluded that the T2 weighted fat suppression turbo spin echo was unable to replace contrast enhanced T1 weighted spin echo because of poor delineation of anatomical detail and reduced tumour conspicuity.
There was statistically no significant difference between FSTIR and FSCE T1W imaging in our study. As a pre-treatment evaluation, it seems to be useful to combine these imaging methods. FSCE T1W imaging should be used to determine perineural, skull base or meningeal invasion. However, as a follow-up, T1 weighted imaging and FSTIR may be sufficient. There may be a role for FSCE T1W imaging as a complementary study for the clarification of tumour progress, when FSTIR shows a suspicion of recurrence. Detecting residual tumours after radiotherapy is difficult because of oedematous changes of the mucosa, which may cause misinterpretation. We experienced three recurrent tumours, but only in one case could we indicate the lesion before the biopsy on both sequences. Early invasion was sometimes missed or was difficult to distinguish from high signal from normal mucosa on STIR sequences and normal mucosal enhancement after gadolinium enhancement. As the tumours were enhanced moderately on gadolinium T1 weighted imaging and showed high intensity areas on STIR sequences, this differentiation is a problem that needs further investigation.
|
Acknowledgments
|
|---|
We thank Dr Ghulam Farid Memon for his useful suggestions and critical discussions for the manuscript.
Received for publication November 20, 2000.
Revision received May 2, 2001.
Accepted for publication May 11, 2001.
|
References
|
|---|
-
Tien RD, Hesselink JR, Chu PK, Szumowski J. Improved detection and delineation of head and neck lesions with fat suppression spin-echo imaging. Am J Neuroradiol 1991;12:19–24.[Abstract]
-
Tien RD, Robbins KT. Correlation of clinical, surgical, pathologic, and MR fat suppression results for head and neck cancer. Head Neck 1992;14:278–84.[Medline]
-
Panus D, Fulbright R, Sze G, Smith RC, Constable RT. Inversion-recovery fast spin-echo MR imaging: efficacy in the evaluation of head and neck lesions. Radiology 1993;187:421–6.[Abstract/Free Full Text]
-
Hasso AN, Brown KD. Use of gadolinium chelate in MR imaging of lesions of the extracranial head and neck. J Magn Reson Imaging 1993;3:247–63.[Medline]
-
Barakos JA, Dillon WP, Chew WM. Orbit, skull base, and pharynx: contrast-enhanced fat suppression MR imaging. Radiology 1991;179:191–8.[Abstract/Free Full Text]
-
Vogl T, Dresel S, Bilaniuk LT, Grevers G, Kang K, Lissner J. Tumors of the nasopharynx and adjacent areas: MR imaging with Gd-DTPA. AJR 1990;154:585–92.[Abstract/Free Full Text]
-
Metz CE. ROC methodology in radiologic imaging. Invest Radiol 1986;21:720–33.[Medline]
-
Hanley JA, McNeil BJ. The meaning and use of the area under receiver operating characteristic (ROC) curve. Radiology 1982;143:29–36.[Abstract/Free Full Text]
-
Metz CE. Some practical issues of experimental design and data analysis in radiological ROC studies. Invest Radiol 1998;24:234–45.
-
Dubin MD, Teresi LM, Bradley WG Jr, Jordan JE, Pema PJ, Goergen SK, et al. Conspicuity of tumorsof the head and neck on fat-suppressed MR images: T2-weighted fast-spin-echo versus contrast-enhanced T1-weighted conventional spin-echo sequences. AJR 1995;164:1213–21.[Abstract/Free Full Text]
-
Michael R, Don S, Phylliss C, Dieter E. MR imaging of head and neck tumors: comparison of T1-weighted contrast-enhanced fat-suppressed images with conventional T2-weighted and fast spin-echo T2-weighted images. AJR 1994;163:173–8.[Abstract/Free Full Text]
-
King AD, Lam WW, Leung SF, Chan YL, Metreweli C. Comparison of T2 weighted fat suppressed turbo spin echo and contrast enhanced T1 weighted spin echo MRI in nasopharyngeal carcinoma. Br J Radiol 1997;70:1208–14.[Abstract]
-
Panush D, Fulbright R, Sze G, Smith RC, Constable RT. Inversion-recovery fast spin-echo MR imaging: efficacy in the evaluation of head and neck lesions. Radiology 1993;187:421–6.
-
Tien RD. Fat-suppression MR imaging in neuroradiology: techniques and clinical application. AJR 1992;158:369–79.[Abstract/Free Full Text]
-
Chan TW, Listerud J, Kressel HY. Combined chemical-shift and phase selective imaging for fat suppression: theory and initial clinical experience. Radiology 1999;181:141–7.[Abstract/Free Full Text]
-
Yoshimi A, Robert BL, Bradly AJ, William NH. Fat-suppression failure artefacts simulating pathology on frequency-selective fat-suppression MR images of the head and neck. Am J Neuroradiol 1992;13:879–84.[Abstract]
This article has been cited by other articles:
|
|
|
|
 
K. Y. Lau, W. K. Kan, W. M. Sze, A. W.M. Lee, J. K.W. Chan, T. K. Yau, R. M. Yeung, L. Tan, P. O. Chan, and A. S.L. Lee
Magnetic Resonance for T-staging of Nasopharyngeal Carcinoma--The Most Informative Pair of Sequences
Jpn. J. Clin. Oncol.,
April 1, 2004;
34(4):
171 - 175.
[Abstract]
[Full Text]
[PDF]
|
|
|
Top
Abstract
Introduction
