localized magnetic resonance spectroscopy of head and neck tumors-Preliminary findings

31p

ROBERT A. HENDRIX, MD, ROBERT E. LENKINSKI, PhD, KATHERINE VOGELE, PETER BLOCH, PhD, and W. GILLIES McKENNA. MD, PhD, Philadelphia, Pennsylvania

Magnetic resonance Imaging (MRI) Is a powerful tool for accurate assessment of the anatomic extent of head and neck neoplasms. Thedevelopment of methods for spatial localization by useof multiply tuned radio frequency colis that permit the measurement of multiple nuclear MR spectra rH and 31PI from precisely defined volumes of Interest has provided a basis for Integrating spectroscopy Into the clinical MRI examination. This offers a means for noninvasive monitoring of relative concentrations of mobile metabOlites within a tumor. Withthe useof Imaging to determine proper coli placement. a test·retest variance of about 17% Is seen on MR spectroscopy. Data are presented from MRI/MRS studies for four head and neck lesions: (1) a squamous cell carcinoma of the lip; (2) a Juvenile angiofibroma extending Into the nasal cavity; (3) a massive chondrosarcoma of the nasal septum; and (4) a cervical nodal metastasis of a squa. mous cell carcinoma of the pharynx. Spectra are evaluated by comparison of relative concentrations of phosphorus compounds. The concentrations of phosphomonoesters and phosphodlesters are significantly higher In the neoplasms studied than In normal skeletal muscle. The developing role of Integrated MRI/MRS to monitor the response of malignant neoplasm to radiation therapy Is discussed. (OTOLARYNGOL HEAD NECK SURG 1990;103: ne:

Magnetic resonance (MR), which was first reported in 1946, has been developed into two different but complementary techniques: magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS). In MRI, anatomic images are produced by means of spatial encoding using switched magnetic field gradients. The signal intensities in these images reflect a number of intrinsic parameters of the tissue, such as water content, fat content, and magnetic resonance relaxation properties. The acquisition parameters of MRI can be adjusted in order to optimize contrast between soft tissues present in the region of interest. This has

From the Departments of Otorhinolaryngology and Human Communication (Dr. Hendrix), Radiology (Drs. Lenkinski and Bloch), Biophysics (Dr. Vogele), and Radiation Oncology (Dr. McKenna), Hospital of the University of Pennsylvania. Presented at the Annual Meeting of the American Academy of Otolaryngology-Head and Neck Surgery. Washington. D.C .• Sept. 25-29, 1988. Received for publication Sept. 27. 1988; revision received Jan. 31. 1990; accepted June 25, 1990. Reprint requests: Robert A. Hendrix MD. Department of Otorhinolaryngology and Human Communications. University of Pennsylvania School of Medicine. 3400 Spruce SI.• Philadelphia, PA 19104.

23/1/23462

led to increasing use of MRI as an imaging modality in head and neck tumors. MRS provides a noninvasive, potentially risk-free method with which to monitor the biochemistry of tissue.!? In the clinical setting the major motivation in using MRS is the hypothesis that diseased tissue exhibits altered metabolism and that these alterations can be identified through an interpretation of MRS data. In each spectrum, peaks OCcur at resonant frequencies related to the chemical environment of each moiety. The area under each peak is proportional to the concentration of the particular compound. This provides the basis for measurement of relative tissue concentrations of molecular components present. In the 31-phosphorus e1p) spectra, the alpha-, beta-, and gamma-nucleoside5'-triphosphate (NTP), creatine phosphate (CP), and inorganic phosphorus (Pi), as well as components of phospholipid metabolism and degradation (phosphomonoesters [PME] and phosphodiesters [PDE]), are detected." In addition, the pH can be determined from the difference in frequency between the Per and Pi peaks. In its early stage, in vivo MRS was limited by the relative low tissue concentration of 31p (5 to 20 mmol! L) compared to that of MR imaging of protons ('H) (110 moIlL), as well as the lack of suitable techniques for selection of a volume of choice without contamination by surrounding tissue." To improve local771

776

HENDRlX at 01.

OfolaryngologyHead and Neck Surgery

Fig. 1. MRI of a large cervical metastasis of squamous cell carcinoma arising from the left tonsillar fossa.

ization and sensitivity, the use of surface coils was introduced in 1980. 6 The Tumor Metabolism-Soft Tumor MRI/MRS Working Group at the University of Pennsylvania has used surface coils for signal acquisition from accessible masses . A variety of head and neck neoplasms has been studied. The visualization of the tumor is made by use of 'H MRI. The position of the surface coil is then determ ined from these images. Thi s permits the integration of spectroscopy into the clini cal MRI to supplement anatomic information of MRI with biochemical information as measured by thel'P spectrum. Examples of head and neck lesion s are presented to demonstrate the range of results obtained with integrated MRI and lip MRS. The reproducibility of the method is illus trated by example . Further discussed are problems arising from sampling overlying muscle tissue in the sensitive volume of the surface coil.

METHODS All MRI and MRS stud ies to locali ze and characteri ze the neoplasms were carried out on a standard General Electric Sigma I.ST whole body MR scanner equipped

with the standard spectroscopy acces sory (commercially available software and hardware). A 7-cm diameter surface coil double-tuned to I Hand 1) P was used . The homogeneity over the sensitive volume of the surface coil was adjusted by shimming on the water resonance . In all of the cases reported here, the width at half-heights of the water resonance was adju sted to less than 40 hertz (Hz) before the\lP spectral acquisition was initiated. The flip angle for lip was adju sted to 90 degrees at the region most proximal to the surface coil (constructed on site) by using a small vial of concentrated H., P0 4 • The vial was placed on the coil when the coil was being positioned on the patient and subsequently removed for the spectral acqui sition . Typical acquisition parameters were 2000-H z sweep width, 1024 data points. 4-second repetition rate, and 32 to 128 averages per spectrum. Because all the tumors reported here were located superficially, surface coil localized acquisitions were performed . The data were proce ssed on an AT&T PC, using so ftware written by Tim Allman at the University of Pennsylvania, includ ing a spectral fitting program based on the simplex method and a deconvolution program to account for any

Volume 103 Number 5 November 1990

Part 1

31p localized magnetic resonance spectroscopy of head and neck tumors 777

PC

d A B

M

PC

b

c

a

b

A B

10

0 -10 Che.lcel Shl ft (pplI)

-20

10

Fig. 2. Pretreatment spectra a and b were obtained at a 14day Interval on the neck mass shown In Fig. 1.Identified peaks InclUde alpha-, beta-, and gamma-trinucleotide phosphate (labeled A B, and C respectively), phosphocreatine (PC), phosphomonoesters and phosphodlesters (M and D, respectively), and Inorganic phophorus (P).

0 -10 Chelilcel Shl ft (PPIll)

-20

FIg. 3. The spectrum b Isthe experimental spectrum b of Fig. 1. Spectrum c, plotted Immediately above, Is the computergenerated plot derived by the 'best fit' algorithm. At the top Isplot d. the mathematical difference between spectra band c. (see Fig. 2 for letter codes).

Table 1. Summary of data Study

1a* 1b* 1c* 2e 21* 3* (case 1) 4* (case 2) 5* (case 3) 6* (case 4) 7 (muscle) Mean Standard deviation

PME/b·NTP 1.1 1.3 1.3 0.9 1.3 16 1.5 10 3.3

t 1.55 0.733

P1/b·NTP 1.1 1.0 0.8 0.5 1.2 1.1 1.4 1.2 29 0.5 1.337 0.655

PDE/b·NTP 1.2 1.4 1.1 06 1.5 0.5 30 3.1 22

t 175 093

PCr/b·NTP 20 20 4.1 2.8 2.5 2.9 3.1 1.2 4.5 2.062 1.036

pH 7.2 7.3 7.2 7.3 73 7.4 7.4 7.3 73 7.2 7.3 0076

'Valid studies of neoplasms. tToo low to measure. Study la and lb are derived from the spectra a and b shown in Fig. 2. These represent two separate pretreatment examinations (14-day interval) of a large cervical mass of squamous cell carcinoma and demonstrate reproducibility by consistency. StUdy lc is derived from the computergenerated 'best fit' spectrum c shown in Fig. 3. Study 2e and 2f are derived from spectra e and' of Fig. 5. The large differences in metabolic ratios reflect the effect of mispositioning the surface coil. Studies 3 through 6 are derived from the spectra for the four case histories shown in Figs 8. 10, 12, and 14, respectively, Study 7 is derived from the spectrum of normal skeletal muscle shown in Fig. 6. The mean and standard deviation under each column are derived only from the values for studies indicated by * (la. lb, 2f. 3. 4, 5, and 6). These are studies believed to represent reproducible and accurate measurements of neoplasm.

771

OlolaryngologyHead and Neck Surgery

HENDRIX et 0 1.

PC

10 Fig .... MRI of a squamous cell carcinoma of the maxillary sinus.

0 -10 Che.lcal ShUt (PPIl)

-20

Fig. 6. Spectrum obtained from normal skeletal muscle. (See Fig. 2 for letter codes).

PC

unusuable points at the start of the free induction decay (FID) .7.8 Imaging studies were used to determine op-

f

A

e

10

0 -10 Che.lcal Shlft (pp.>

-20

Fig. 5. Spectrum f Is an MRS study of the lesion shown In Fig. 4. Spectrum e was acquired on the same patient with an Intenttonal mlsposlttonlng of the surface call by 5 em, resulttng in masseter muscle contamlnatton .This demonstrates the need for spectroscopy to be Integrated with Imaging for proper call placement. (see Fig. 2 for letter codes).

timal placement of the surface coil s to minimize contamination . The images were made with the standard GE multi-echo. multiplanar, pulse sequences . Limitations. Because of the low concentrations of biochemical compounds studied by spectroscopy. poor sensitivity with a resultant low signal/noise ratio presents a fundamental limitation to the technique of MRS. Thus MRS requires a longer period of signal acquisition and a much larger volume of interest (VOl) than that used for MRI (which is based on acquisition of signals from hydrogen in cellular water and lipids). Further limitations associated with using surface coils include the inability to completely remove undesired signals from the surface and the nonlinear spatial weighing of the contribution of the metabolites to the spectrum resulting from the response of the coils .

RESULTS Reproducibility. The reproducibility of the method was tested by carrying out an integrated MRI / MRS examination of the same patient on different days. Shown in Fig . I are axial images of a 62-year-old. debilitated. male patient with a stage IV, T 4N JM o squamous cell carcinoma arising from the left tonsillar fossa with a 6- x 8-cm cervical metastasis . The spectra obtained before treatment from two separate examinations (I4-day interval) are shown in Fig. 2. The se spectra were fit using a simplex algorithm (see Methods) that was part of the analysis software. The results of the fit

Volume 103 Number 5 Port 1 November 1990

3,p localized magnetic resonance spectroscopy of head and neck tumors 779

Fig. 7. MRI of a patlent with a squamous celi carcinoma of t~ lip.

obtained on spectrum b of Fig. 2 are shown in Fig. 3, Spectrum C , along with the difference, d. between the experimental spectrum b of Fig. 3 and the computergenerated , best-fit spectrum C of Fig. 3. There is good agreement between the experimental and best-fit spectra. The relative area of the metabolites present in the two spectra shown in Fig. 2 may be expressed as ratios of metabolite/beta-NTP as given in Table I . The largest difference observed in a given ratio is about 0.2 units . This repre sents a 17% variation in both the PME/betaNTP and PDE /beta-NTP ratios . PosItioning the coli. The effects of mispositioning the coil were examined by purposely moving the co il to a location 5 cm away from its optimum position (as judged by MRI location) in a 64-year-old, male patient with a right maxillary sinus carcinoma (Fig. 4) . The two spectra obtained are shown in Fig. 5; spectrum e is 5 ern mispositioned; and spectrum j was obtained with the coil in the correct position. These spectra were fit as described above and the resulting metabolite ratios are given in Table I . There are dramatic changes observed in these ratio s that confirm the importance of using the MR images in choosing the correct position of the coil. For comparison purposes , a spectrum of normal muscle is shown in Fig . 6 . It is clear that when the coil is mispositioned. a substantial fraction of the tissue sampled is muscle. This conclusion can be drawn by comparing the ratios derived from spectra e and j,

PC

10

0

-10

-20

Chemical Shift (ppm) Fig. 8. Spectrum of the lesion shown In Fig. 7. (see Fig. 2 for letter codes).

shown in Fig. 5, with those derived from the spectrum in Fig. 6. Spectrum j, displayed in Fig . 5, exhibits ratios that support the argument that most of the volume being sampled contains neoplasm . CASE REPORTS

Case 1. A 68-year-old man with gastric carcinoma and a 3-cm squamous cell carcinoma of the lower lip underwent MRI and MRS (Figs. 7 and 8, respectively).

710

OtolaryngologyHead and Neck Surgery

HENDRIX at 01.

Fig. 9. MRI of a patient with a large Juvenile angiofibroma.

Fig . 11. MRI of a patient with a chondrosarcoma of the nasal septum.

PC PC

o

10

0

-10

-20

Che.1cal Shift (ppm)

10

0

-10

-20

Chemical Shift (ppm) Fig . 10. Spectrum of the Juvenile angiofibroma. (see Fig. 2 for letter codes).

Case 2. A 19-year-old man , with nasal obstruction and anosmia of I year duration , was found to have a large mass by cr scan , leading to biopsy, proving the diagnosis of juvenile angiofibroma . MR! was performed , revealing a large nasopharyngeal and nasal cavity mass (Fig. 9). The results of MRS are presented in Fig. 10. Case 3. A 28-year-old Cambodian woman was discovered to have total nasal obstruction during preparation for open heart surgery. Biopsy led to the diagnosis of grade II chon-

Flg .12. Spectrum of the chondrosarcoma of the nasal septum . (see Fig. 2 for letter codes).

drosarcoma of the nasal septum. MR! showed the mass filling the nasal cav ity, involving the orb it, and extending into the base of the skull ( Fig. II) . MRS is shown in Fig . 12. Case 4. A 42-year-old man with a stage IV, T,N'AMo poorly differentiated, invasive . squamous cell carcinoma of the anterior floor of mouth and mandible with a large neck metastasis underwent MRI and MRS (Figs. 13 and 14, re-

VOlume 103 Numbe r 5 Port 1 November 1990

" P loc alized magnetic resonance spectroscopy of head and neck tumors 781

SquI.OUI Clreino.1 Mltlltie to CervSell Nodi

Ly.ph

p

i

10

i -10

o

Che-Sell ShUt (pp_)

Flg .13. MRIof a patient with a cervical metastasisof a tonsillar squamous cell carcinoma.

Fig . 14. Spectrum of the neck mass shown In Fig. 13 after two unsuccessful cycles of chemotherapy. (See Fig. 2 for letter codes).

NORMALIZED RATIOS OF PMEI B·ATP AND PME/PDE VERSUS THE DAY IN RADIATION THERAPY 10

10 --

nPMEJPDE n PMEJ B ATP

8

8

6

6

4

4

2

2

o+-..........----......-..,...............-r--r---r-...--.........~-...- .........__,.-....__+o -1

i

-20

4

9

14

Oay In Rad iation Therapy

Fig. 15. Changes during rad iation therapy In the relative concentrations of low-ene rgy phosphate metabolites (phosphoesters) In the MR spectra for case 4 can be demonstrated by p lots of ratios of PME and POE to beta-nucleotide triphosphate.

OtolaryngologyHead and Neck Surgery

712 HENDRIX at 01.

M P

x

2b

10

0 -10 Ch••leal Shl tt (pp.)

-20

Fig. 16. Labeled 2b Isthe experimental spectrum shown as b In Fig. 2. Spectrum X Is a 'corrected' spectrum obtained by subtracting a scaled muscle spectrum, as In Fig. 6, from the experimental spectrum 2b.

spectively) after two unsuccessful cycles of chemotherapy with cis-platinum and 5-fluorouricil. This spectrum shows an increase in phospholipid moieties and inorganic phosphorus. with a decrease in high energy phosphocreatine. This pattern has been observed in a number of malignancies. The patient subsequently underwent radical radiotherapy. with a 20% reduction in tumor volume noted on treatment day 7 (after a total dose of 1600 cOy). The spectrum was repeated, showing a significant change in the areas of the peaks present. A third spectrum, performed at 14 days (after a total dose of 3200 cOy). showed further change in peak morphology. This is graphically demonstrated in Fig. 15, showing plots of the ratios of PME/PDM, PME/bcta-NTP, and PC/P, over time. All of the spectra from these cases were fit using the algorithms described earlier, and the resulting ratios are given in Table I.

DISCUSSION From the results presented here, it is clear that the test-retest reliability of this method is approximately 17%. It is also clear that the MR images must be used

to arrive at the optimum placement of the coil used for liP-MRS. From the data presented in Table 1, it is evident that the PME/beta-NTP ratio (mean, 1.6; standard deviation, 0.7) and PDE/beta-NTP ratio (mean, 1.3; standard deviation, 0.7) are significantly higher than in normal muscle. The largest squamous cell carcinoma (case 4) exhibited ratios that were significantly higher than the rest. These higher PME and PDE ratios may indicate an impairment in energy metabolism in the disordered biochemistry of neoplasms. It is interesting to note that even in these large tumors, the pH is in the normal-to-alkaline range. Similar observations have been made by others in other tumor systems." The issue of contamination of the tumor spectrum by contribution from surrounding muscle tissue can be addressed if we assumed that the tumor contains little or no Per. As an example, spectrum 2b of Fig. 16 shows the experimental spectrum b of Fig. 2 together with a 'corrected' spectrum X in Fig. 16. This corrected spectrum was obtained by subtracting a scaled muscle spectrum from the experimental spectrum. The corrected spectrum was fit and the resulting metabolite ratios are given in Table 1. These ratios are only slightly different from those derived from the uncorrected spectrum. For this reason, we conclude that comparisons of the PME/beta-NTP, Pi/beta-NTP, and PDE/beta-NTP ratios are valid, even if there are small amounts of muscle tissue present. The concept of a spectroscopic fingerprint for a neoplasm-Le., a spectral features characteristic of a neoplasm that are stable over time-does not appear to be likely. The metabolism of neoplasms is dynamic and inherently unstable. However, the spectral data provide information about tissue metabolism at a specific point in time. Further, changes in the spectra may provide valuable information about the response of a tumor to therapy. Although the inherent problem of low sensitivity of MR spectroscopy will require development of new techniques to improve signal/noise ratio and minimize contamination from surrounding tissue, the use of surface coils and a magnetic field of 1.5 T in a whole body MR scanner allows for superficial lesions (within 4 ern of the surface) of adequate volume (at least 20 ml) to be studied with MRS.

CONCLUSIONS With the availability of whole-body MR scanners with spectroscopic accessories, and the use of surface coils for localization, in vivo lip MR spectroscopy may be integrated with MR imaging to evaluate superficial tumors of sufficient volume. Further study is necessary to define the significance of spectral data to head and neck oncology; however, this developing role of MRS

Volume 103 Number 5 Part 1 November 1990

:tip localized magnetic resonance spectroscopy Of head and neck tumors

will enable clinicians to make decisions based in part on the cell physiology and biochemistry of a specific

neoplasm. REFERENCES 1. Weiner NW. The promise of magnetic resonance spectroscopy for medical diagnosis. Invest Radiol 1988;23:253-6l. 2. Bottomley PA. Human in vivo NMR spectoscopy in diagnostic medicine: clinical tool for or research probe? Radiology 1989; 170:1-15. 3. Radda OK. Rajagopalan B, Taylor OJ. Biochemistry in vivo: an appraisal of clinical magnetic resonance spectroscopy. Magn Reson Imaging 1989;5:122-51. 4. Cohen JS. Phospholipid and energy metabolism of cancer cells monitored by "P magnetic resonance spectroscopy: possible clinical significance. Mayo Clin Proc 1988:1199-207.

783

5. Aue WP. Muller S. Cross TA. Seelig J. Volume-selective excitation: a novel approach to topical NMR. Magn Reson Med 1984;56:350-4. 6. Ackerman JJH. Orove TH. Wong GO, Oadikan DO, Radda OK. Mapping of metabolites in whole animals by lip NMR using surface coils. Nature 1980;283:167-70. 7. Wehrli FW. Principles of magnetic resonance. In: Stark DO. Bradley WG. eds. Magnetic resonance imaging. St. Louis: The CV Mosby Co.• 1988:3-23. 8. Lenkinski RE. Allman T. Scheiner 10, Deminig SN. An automated iterative algorithm for the quantitative analysis of in vivo spectra based on the simplex optimization method. Magn Reson Med 1989;10:338-48. 9. Negendank WO, Crowley MO. Ryan JR. Keller NA. Evelhoch JL. Bone and soft-tissue lesions: diagnosis with combined H-l MR imaging and P-31 MR spectroscopy. Radiology 1989; 173:181-8.